Research For Nurses - Methods and Interpretation

Copyright 2002 F.A.
Davis Company
RESEARCH FOR NURSES:

METHODS AND INTERPRETATION
Copyright 2002 F.A. Davis Company
This page intentionally left blank

RESEARCH FOR
NURSES:
METHODS AND
INTERPRETATION
Angela Gillis, BSN, PhD

Professor and Chair
Department of Nursing
St. Francis Xavier University
Antigonish, Nova Scotia
Winston Jackson, PhD

Professor of Sociology
St. Francis Xavier University
Antigonish, Nova Scotia
F.A. DAVIS COMPANY PHILADELPHIA

F. A. Davis Company
1915 Arch Street
Philadelphia, PA 19103
www. fadavis.com
Copyright 2002 by F. A. Davis Company
All rights reserved. This book is protected by copyright. No part of it may be reproduced, stored in
a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopy-
ing, recording, or otherwise, without written permission from the publisher.
Printed in the United States of America
Last digit indicates print number: 10 9 8 7 6 5 4 3 2 1
Acquisitions Editor: Melanie Freely
Developmental Editor: Laura Bonazzoli
Production Editor: Nwakaego Fletcher-Perry
Cover Designer: Louis Forgione
As new scientic information becomes available through basic and clinical research, recom-
mended treatments and drug therapies undergo changes. The author(s) and publisher have done
everything possible to make this book accurate, up to date, and in accord with accepted standards
at the time of publication. The author(s), editors, and publisher are not responsible for errors or
omissions or for consequences from application of the book, and make no warranty, expressed or
implied, in regard to the contents of the book. Any practice described in this book should be ap-
plied by the reader in accordance with professional standards of care used in regard to the unique
circumstances that may apply in each situation. The reader is advised always to check product in-
formation (package inserts) for changes and new information regarding dose and contraindica-
tions before administering any drug. Caution is especially urged when using new or infrequently
ordered drugs.
Library of Congress Cataloging-in-Publication Data
Gillis, Angela, 1953
Research for nurses : methods and interpretation / Angela Gillis,
Winston Jackson.1 st ed.
p. cm.
Includes bibliographical references and index.
ISBN 0-8036-0896-9 (pbk.)
1. NursingResearch. 2. NursingResearchMethodology. I.
Jackson, Winston. II. Title.
RT81.5 .G53 2001
610.73072dc21
Authorization to photocopy items for internal or personal use, or the internal or personal use of
specic clients, is granted by F. A. Davis Company for users registered with the Copyright Clear-
ance Center (CCC) Transactional Reporting Service, provided that the fee of $.10 per copy is paid
directly to CCC, 222 Rosewood Drive, Danvers, MA 01923. For those organizations that have been
granted a photocopy license by CCC, a separate system of payment has been arranged. The fee
code for users of the Transactional Reporting Service is: 8036-0896/02 0 $.10.
T
o my mother, father, and sisters, Liza and Catherine, who taught me love. To
my husband, Phonse, and daughters, Rita and Catherine, who tolerated my ab-
sence when they would have preferred my presence, and who continually re-
mind me that love is what is really important. AG
. . . and to all the nursing students, past and present, for their commitment and
enthusiasmthanks for the memories and all your wonderful projects. WJ
Preface
The authors of this text set out with one dergraduate student with no research ex-
major goalto write a book that would perience can grasp the concepts dis-
demystify the research process as much cussed. It is also appropriate for post-reg-
as possible and to present the material in istered nurse degree students returning to
an exciting, fresh, and innovative man- pursue a baccalaureate degree.
ner. The book examines both qualitative
and quantitative approaches available to Part 1 focuses on basic concepts of re-
the nurse researcher because no single search, the fundamental importance of
theory, research design, data collection research to nursing, and the process
method, measurement procedure, or sta- of identifying important researchable
tistical approach can provide a satisfac- problems.
tory answer to all research questions. Part 2 presents ve chapters that address
The text prepares students to be both the most popular quantitative and qual-
knowledgeable users and doers of research. itative designs in contemporary usage.
We believe students learn most by active Part 3 examines bias and ethical issues in
involvement. Where possible, the book research.
presents rules and steps for understanding Part 4 discusses basic statistical concepts.
each stage in evaluating and conducting re- This material provides the foundation
search. In order to empower and encour- for the next part.
age the student, the book provides many Part 5 discusses the actual planning and
ideas for feasible projects that have been implementation of a research project.
undertaken by graduate and undergradu- Part 6 deals with analyzing the ndings of
ate nursing students. In short, the book is a study.
student-friendly, useful, and relevant to Part 7 focuses on reporting and dissemi-
contemporary nursing. nating the ndings, critiquing research,
Research for Nurses: Methods and Inter- and utilizing the ndings to enhance
pretation does not assume that the student practice.
has prior knowledge of statistics or of re- Part 8, a section not typically featured in
search design. Indeed the strength of the other nursing research methods texts,
text is that it is comprehensive enough to explains how to process data using Sta-
be used as a graduate text, yet written in a tistical Package for the Social Sciences
simple and straightforward manner with (SPSS). This section walks the rst-time
numerous examples so that the senior un- user through the various steps in ana-
vi
PREFACE vii
lyzing research data and in reading A chapter devoted to the problem of

SPSS output. bias in nursing research. This is a
Other features are: unique feature of the text on a topic
that is not sufciently discussed in
Easy-to-follow guidelines for reading most nursing research texts, yet it is a
and critically appraising quantitative problem that plagues every aspect of
and qualitative research projects. the research process.
A section devoted to new-wave ap- A two-chapter statistics primer cover-
plied designs such as action research, ing basic statistical material. Ad-
health promotion studies, feminist, vanced techniques such as regression
and evaluation research. analysis, discriminant analysis, multi-
Down-to-earth steps to guide the stu- variate analysis of variance, and factor
dent through a project. This feature analysis are presented in Chapter 18.
helps the student to realize that he or An overview of what NUD*IST means,
she can engage in rigorous research to and a discussion of how it can be used
answer questions that emerge in the to analyze qualitative data.
practice setting. A series of 10 laboratory assignments to
Case studies and sample projects pre- train students in the use of SPSS (SPSSx,
sented in Nursing Researchers at SPSS PC, and SPSS for Windows ver-
Work boxes to illustrate key ap- sion 10.0) are included in the instruc-
proaches, methods, and ndings. This tors manual along with a sample data
is an important feature that helps on attitudes toward euthanasia.
ground nursing research in the real Numerous Web site addresses, includ-
world of nursing. ing the authors Web site, which in-
Step-by-step instructions for question- cludes 160 sample questionnaires de-
naire design and administration and signed by student researchers. Other
for sample size determination. sites will be identied to assist students.
Acknowledgements
Many individuals have freely given their to Donna Beiswanger, Gary Brooks, Lisa
advice on different aspects of this text as Keeping, Tanya Greencorn, Marlies Jack-
it was being developed. We would espe- son, Shirley Laffrey, Ellen MacFarlane,
cially like to thank the many students at Jean MacInnis, Jason Ryan MacLean, Deb-
St. Francis Xavier University who chal- bie Murphy, Dorothy Payne, Barbara
lenged us with interesting research prob- Phillips, Maria Sangster, Judith Shaw, and
lems and who taught us so much. To the Saru Sony.
chapter reviewers at F. A. Davis: thank We wish to acknowledge the contribu-
you for your expertise, promptness, and tion of St. Francis Xavier University in
many suggestions. We hope you can rec- providing the sabbatical leave that gave
ognize the places where your suggestions us the opportunity to work on the manu-
improved the text, and perhaps not no- script. The support of Hugh Gillis, Ron
tice where we did not follow your coun- Johnson, and Ed McAlduff is most appre-
sel. To the F. A. Davis staff: thank you for ciated.
your professionalism, and your editorial We are also indebted to SPSS Inc. for
and production work. In particular, we permission to present the commands for
would like to thank our acquisitions edi- SPSS. As always, SPSS personnel have
tor, Melanie Freely, for her enthusiasm, been most helpful. In particular, we
encouragement, creative ideas, and be- would like to thank Marcus Hearne of
lief in this project; our development edi- SPSS Inc., Chicago, Illinois, who was most
tor, Laura Bonazzoli, for her clarity of generous in providing a copy of version
thought, spirited advise, patience, and 10.0 of SPSS for Windows to help in the
skilled editing; our production editor, development of Appendix A.
Nwaka Fletcher-Perry, for her construc- Finally we wish to acknowledge our
tive suggestions and attention to last- families for their love, patience, under-
minute details. To our friends and col- standing, and support throughout this in-
leagues: thank you for the insightful tensive but exciting endeavor.
comments and the welcomed coffee
breakswithout which we could not Angela Gillis
have written the text as well as we have. Winston Jackson
We also give our thanks and our gratitude February, 2001
viii
Consultants
Elizabeth Black, MSN, BSN Catherine Blackwell Holland, BS, MS, PhD
Assistant Professor Assistant Professor
Director ASN Program Southeastern Louisiana University
Gwynedd-Mercy College Baton Rouge Campus
Gwynedd Valley, Pennsylvania Baton Rouge, Louisiana
E. Joyce Black, RN, EdD Mary Taylor Martof, RN, EdD

Education Consultant Associate Professor
Registered Nurses Association of British Louisiana State University Health Sciences
Columbia Center
Vancouver, British Columbia New Orleans, Louisiana
Dr. Janet Brookman, DSN, RN Martha MacLeod, RN, PhD

Clinical Associate Professor Associate Professor
The University of Alabama in Huntsville University of British Columbia,
Huntsville, Alabama British Columbia, Canada
Elizabeth S. Carlson, PhD, RNC Sharon Moore, RN, PhD

Assistant Professor College Research Ofcer
School of Nursing Faculty of Health and Community Studies
Loyola University-Chicago Mount Royal College
Maywood, Illinois Calgary, Alberta
Ann M. Dollins, AD, BSN, MPH, MSN, PhD Mary Jo Gorney-Moreno, PhD, RN
Director BSN Program School of Nursing
Northern Kentucky University San Jose State University
Highland Heights, Kentucky San Jose, California
Karen Fernengel, PhD, FNP, CS Carla Mueller, PhD, RN

Chair Graduate Nursing Program Associate Professor
Graceland University University of St. Francis
Leavenworth, Kansas Fort Wayne, Indiana
Rebecca L. Hartman, EdD, RN Kathleen Poindexter, RN, DNSc

Coordinator, Allied Health Professions Assistant Professor
Assistant Professor, Nursing College of Allied Health
Indiana University of Pennsylvania Temple University
Indiana, Pennsylvania Philadelphia, Pennsylvania
ix
x CONSULTANTS
Rosalee J. Seymour BSN, MSN, EdD Maria A. Smith, DSN, RUN, CCRN
Associate Professor Professor
School of Nursing School of Nursing
East Tennessee University Middle Tennessee State University
Johnson City, Tennessee Chattanooga, Tennessee
Teresa Shellenbarger, DNSc, MSN, BS Susan Urbanski, MSN, RN

Professor of Nursing President /Executive Director
Indiana University of Pennsylvania CIGNA Behavioral Health of California
Indiana, Pennsylvania Glendale, California
Contents
PART 1 GETTING STARTED 1 Predictable Setbacks and Challenges 96

Determining Your Readiness to Start
Chapter 1 the Project 98
Introduction to Nursing Research 3

Research and Nursing: Parallel Processes 4
PART 2 RESEARCH DESIGNS AND
Conceptualizing Research in Nursing 6
APPROACHES 103
The Varieties of Nursing Knowledge 8
Acquiring Nursing Knowledge
Chapter 4
through Research 12
Key Methodological Approaches in Experimental and Quasi-experimental
Nursing Science 17 Designs 105
Types of Research: Some Important Basics of Experimental Designs 106
Distinctions 23
Pre-experimental Designs 108
Research as a Professional Mandate 34
Classic Experimental Designs 112
Quasi-experimental Designs 126
Chapter 2 Advantages and Limitations of
Experimental Designs 131
Understanding Theory 37
Explanation 38
Chapter 5
Theory 44
Concepts and Statements: Building Survey Designs 135
Blocks of Theory 48 A Short History of the Survey 136
Theoretical Frameworks 53 The Rationale for Surveys 137
Testing and Using Theory in Research 58 Steps in Survey Research, 141
Points of Agreement on Theory Guidelines for the Administration of
and Research 65 Questionnaires 141
Comparative Studies 156
Chapter 3 Secondary Data Analysis 164
Advantages and Limitations of
Deciding What to Study 68 Survey Designs 173
Choosing a Topic and Stating the Problem 69
Reviewing the Literature 76
Chapter 6
Specifying the Model 84
Specifying Hypotheses and Procedures Qualitative Research Designs 177
of Analysis 90 Entering the World of Qualitative Research 179
Understanding Research Design 92 Qualitative Research Designs 186
xi
xii CONTENTS
Scientific Adequacy of Qualitative PART 4 A STATISTICS PRIMER 349

Research 214
Advantages and Limitations of Chapter 11
Qualitative Approaches 219
A Primer for Descriptive Statistics 351
Chapter 7 Levels of Measurement 352
Describing an Individual Variable 354
Field Study Approaches 224 Describing Relationships Between Variables 370
Participant Observation Studies 225
In-depth Interviews 232 Chapter 12
Focus Group Interviews 234
A Primer for Inferential Statistics 389
Field Experiments 239
Key Elements in Tests of Statistical
Naturalistic Observational Studies 241
Significance 391
Analyzing Qualitative Data 248
The Chi-Square Test of Statistical
Conclusion 256 Significance (X2) 396
The t Test of Statistical Significance 403
Chapter 8 The F Test of Statistical Signficance 408
New Wave Research: Contemporary When Tests of Statistical Significance
are not Appropriate 416
Applied Approaches 263
Action Research 264
Evaluation Research 269
PART 5 ISSUES IN DATA
Health Promotion Research 275
COLLECTION 421
Feminist Research 281
Advantages and Limitations of Contemporary Chapter 13
Applied Approaches 289
Post Script to Part 2 289
Data Collection and Measurement 423
Levels of a Research Project 424
Linkages between Levels 427
PART 3 EXPLORING BIAS AND Measurement Error 440
ETHICAL ISSUES IN RESEARCH 295 Levels of Measurement 443
The Effects of Reduced Levels
Chapter 9 of Measurement 446
Procedures for Data Collection 450
Understanding Bias 297 Development of an Instrument 466
The Nature of Bias 298
How Bias Affects the Research Process 302 Chapter 14
Advocacy versus Pure Research 313
Questionnaire Development 471
Guidelines for Minimizing Bias 316
General Guidelines for Questionnaire
Conclusion: The Gap between Myth Development 472
and Reality 318
Types of Questions Illustrated 474
Steps in Developing a Questionnaire 480
Chapter 10
Ordering Questions, Formatting,
Ethical Issues In Nursing Research 322 and Presentation 486
What Is Ethics? 323 Evaluation of Questionnaires 492
Ethical Decision-Making 324
Studies Illustrating Ethical Chapter 15
Dilemmas 336 Sampling and Sample Size 495
Rules for Conducting Ethical Research 341 The Terminology of Sampling 496
CONTENTS xiii
Fundamental Sampling Techniques 498 Organization of Quantitative Reports 595

Sample Size Determination 508 Organization of Qualitative Reports 606
Other Sampling Issues 515
Chapter 20
PART 6 QUANTITATIVE DATA The Research Critique 611
ANALYSIS 519 Understanding the Critiquing Process 612
Critically Appraising Research Reports 613
Chapter 16 Examples of Research Critiques 620
Starting the Data Analysis 521
Analyzing Data without Using a Computer 522 Chapter 21
Analyzing Data Using a Computer 526
Research Utilization in Nursing 632
Understanding Research Utilization 633
Chapter 17
Barriers to Research Utilization 637
Testing Simple Causal Models 544 Strategies to Facilitate Research-Based
Testing Three-Variable Causal Models 545 Practice 641
Testing Four-Variable Causal Models 556 Example of Research Utilization 644
Future of Research Utilization 647
Chapter 18
Four Multivariate Techniques 559
Multiple Regression 560 PART 8 APPENDICES 651
Discriminant Function Analysis 570
Multivariate Analysis of Variance 573 Appendix A
Factor Analysis 577 Using SPSS for Windows 653
The Four Techniques Compared 583 Methods of Issuing SPSS Commands 654
Basic SPSS Procedures 658
SPSS Manuals 691
PART 7 REPORTING, CRITIQUING,
AND USING RESEARCH Appendix B
FINDINGS IN NURSING 587 A Student-designed Questionnaire 692
Chapter 19 GLOSSARY 699
The Research Report 589 BIBLIOGRAPHY 717

General Orientation 590 INDEX 733

Part 1
GETTING
STARTED
C
hapter 1 introduces students to a variety of nursing research
approaches. In addition to describing research and its utiliza-
tion as a key to the progress of nursing science, the chapter
outlines the major methodological approaches in nursing
scholarship. Chapter 2 describes the role of theory in understanding
human behavior. Various nursing theoretical frameworks are dis-
cussed, along with approaches to testing them. Chapter 3 begins our
exploration of how research projects are designed.
1

Chapter 1
Introduction to
Nursing Research
CHAPTER OUTLINE
A. Research and Nursing: Parallel Processes E. Key Methodological Perspectives in

B. Conceptualizing Research in Nursing Nursing Science
1. What Is Research? 1. The Positivist Perspective
2. What Is Nursing Research? 2. The Interpretive Perspective
3. What Are the Goals of Nursing 3. The Critical Perspective
Research? F. Types of Research: Some Important
C. The Varieties of Nursing Knowledge Distinctions
1. Sources of Nonresearch-Based Nursing 1. Quantitative versus Qualitative
Knowledge Approaches
2. Four Ways of Knowing 2. Descriptive versus Explanatory versus
D. Acquiring Nursing Knowledge Through Predictive Research
Research 3. Pure versus Applied Research
4. Units of Analysis: Individuals versus
1. Role of Philosophy in Nursing Research
Aggregations
2. Stages of a Research Project
3. A Preview of Research Designs G. Research as a Professional Mandate
4. Use of Computers in Nursing Research
KEY TERMS
Aesthetics Explanatory research Positivist perspective

Applied research Field studies Predictive research
Conceptual hypothesis Hypotheses Pure research
Conceptual level Interpretive perspective Quantitative research
Conceptual variable Macrovariables Qualitative research
Contemporary applied Mechanistic model Reliability
research Methodological triangulation Research
Critical perspective Microvariables Research process
Descriptive research NUD*IST Scientific method
Empirical knowledge Operational level SPSS
Epistemology Operationalization Theoretical level
Ethical knowing Participatory action research Validity
Evidence-based practice Personal knowledge
3
4 GETTING STARTED
R
esearch has been given a bad rap. enhance the quality of your care, further
Many people would have you think your career goals, and advance the status
of it as dull, intimidating, and even of the entire nursing profession.
immobilizing, when in reality, conducting Nursing research includes both quali-
research can be fun and exciting as well tative and quantitative approaches. There
as crucial to the development of nursing are fundamental differences between
as a profession and the achievement of these approaches on virtually every as-
excellence in practice. For those who ap- pect of research. Throughout this text,
proach research with trepidation, this we will attempt to alert the reader to such
text will help you to realize that the differences and to point out how both ap-
essence of research is reected in nurses proaches are legitimate in their own right
everyday practice. It is not an obstacle to and both will make signicant contribu-
be avoided but an essential tool that we tions to nursing research in the years
hope you will come rst to understand, ahead. Later in this chapter, we explore
then to use, and nally, to embrace. some of the differences between the two
Nurses have a responsibility to ensure approaches. Chapters 6, 7, and 8 are de-
that their clients are the recipients of care voted to exploring qualitative research
that reects the most current and relevant strategies. We will begin our journey on
knowledge available. Thus, research has familiar ground, with a comparative view
become a highly valued activity for nurses. of the nursing process.
As you yourself are experiencing, bac-
calaureate and graduate-level nursing pro-
grams now prepare students with courses A. RESEARCH AND NURSING:
in research methods. You also probably PARALLEL PROCESSES
have noticed that content in nursing text-
books reects current research ndings The steps of the research process should
as an evidence base for practice. On the not be totally new to you because you
professional level, nursing leaders for perform similar activities every day when
decades have been challenging nurses to you engage in critical thinking, reect
use research ndings to improve care. Re- on your practice, and apply the nursing
search utilization has long been a priority process. For example, you engage in data
in the nursing literature (Lindeman, 1975) collection about your client when you read
and remains a priority today (Moore, 1995; the clients chart; receive reports; and en-
Robinson, 1997). Finally, hospitals, clinics, gage in dialogue with colleagues, the
agencies, and clients expect nurses to en- client, and the family. Based on this in-
gage in evidence-based practicethat is, formation and your observations, you for-
nursing practice should be informed and mulate a diagnosis of the clients health
modied in the light of systematic re- needs. This stage is similar to naming,
search that evaluates the effectiveness of taxonomy, or problem identication. Next,
alternative interventions. you develop a plan of care specic to the
The authors of this text believe every clients identied needs. This is similar to
nurse can develop the skills needed to be- the methodology plan that the researcher
come actively engaged in the research develops after the problem statement is
process. We therefore invite you to take clearly dened. Your plan is based on
up the challenge of nursing research, and an expectation that certain interventions
to do so with condence! In return, we produce intended outcomes for the client.
promise that you will soon come to re- This is similar to prediction in the quanti-
spect the power of nursing research to tative research process. Finally, you eval-
INTRODUCTION TO NURSING RESEARCH 5
uate the effect of the interventions on the soning to guide logical decisions. Issues
clients outcomes and recommend modi- of concern to both include:
cations to the plan as new evidence be-
comes available. This is similar to interpre-
Are data about the client valid and
reliable?
tation of ndings and the recommendation
stage of the research process.
Have the relevant problems (diagnoses)
been accurately identied?
Table 1.1 compares the steps of the
nursing process with those of the re-
Is the clients perspective truly repre-
sented in the diagnosis and plan of
search process, which will be described
care?
in detail later in this chapter and through-
out this text. For now, it is important to re-
What relationships, among which vari-
ables, are responsible for the clients
member that both the nursing process
outcomes?
and the research process are simply
series of actions taken by the nurse to
What other variables need to be con-
sidered by the nurse?
achieve a specic goal. Ultimately, the
general goal of both processes is to im-
What other alternative explanations
could account for the intervention
prove the quality of care provided to our
results?
clients and their families, although spe-
cic differences do exist in the means that As we have stated, although the nurs-
each process uses to achieve this end. ing process and the research process
In effective application of the nursing contain parallel elements, their intended
process, the staff nurse is concerned with purposes are quite different. Quantitative
similar issues as the researcher. In both research is conducted to produce nd-
processes, the nurse uses problem solv- ings that are generalizable and meaning-
ing, critical thinking, and complex rea- ful to others. Findings of the nursing
Table 1.1 Comparison of Steps of Nursing Process and Stages of Nursing Research
Nursing Process Qualitative Research Process Quantitative Research Process
Assessment and Selection of a social context Problem identication
interpretation of data Literature review
Selection of conceptual
framework or theory
Diagnosis of problem Formulation of research question Formulation of research question
Explication of researchers beliefs or hypothesis
Development and Development and implementation Development and implementation
implementation of of methodology plan of methodology plan
plan of care Cyclical process of data collection,
analysis, and concept formation
and modication
Evaluation of plan of care Interpretation of ndings Interpretation of the ndings
May involve review of the literature
at this stage
Validation of ndings with
participants
Revision of plan based Communication of the ndings Communication of the ndings,
on evaluation Implications and recommendations implications, and
recommendations
6 GETTING STARTED
process and of qualitative research, in tions about the type and nature of knowl-
contrast, are specic to a particular client edge you need to practice nursing. These
and setting and are not necessarily gen- questions may include:
eralizable to others. Also, in applying
the nursing process and in qualitative re- What is nursing knowledge?
search, the nurse may not necessarily What knowledge do nurses need to
practice nursing?
have reviewed the literature about a client
problem or situated the problem within a How is nursing knowledge created?
theoretical framework. The nurse also Is nursing knowledge different from
that required by other health-care pro-
may not describe the various factors in a
fessionals?
client situation precisely enough that any
other nurse could take the same inter- What is nursing research?
vention and effectively demonstrate the What are the important questions to
be researched in nursing?
same results. There is no attempt in the
nursing process or in qualitative research When should research results be used
to change practice?
to control all of the variables in the envi-
ronment to the extent that the nurse can The world of nursing research you are
be sure that the action taken really ac- about to explore will help you reect on
counts for the results. responses to these questions, spark your
In addition, the research process is curiosity for research possibilities, and
quite simply more complex than the nurs- challenge your intellectual creativity. Let
ing process. It requires that the nurse us begin our exploration by reecting on
learn a new vocabulary and a specic set the nature of research itself.
of rules and methods that must be rigor-
ously applied. Without this new knowl-
edge, you will not be able to perform 1. What Is Research?
nursing research effectively.
As a nursing student, you apply the Researchers frequently attach different
nursing process on a consistent basis meanings to the term research; however,
with ease and condence. Through work- there is consensus that research implies
ing with this text, you will learn to apply a formal, rigorous, and precise process.
the research process with the same ease These characteristics require research to
and to build a body of scientic knowl- be a planned and systematic activity. A
edge that will enhance your practice and second point on which there is consen-
increase the accountability of your pro- sus in the literature is that the aim of re-
fession to the public you serve. search is to discover new knowledge and
relationships and nd solutions to prob-
lems or questions. Box 1.1 provides com-
B. CONCEPTUALIZING mon denitions of research with varied
RESEARCH IN NURSING meanings.
You will notice that whereas some def-
The term research generally implies a sys- initions call for a single approach to de-
tematic inquiry into a subject to discover velopment of knowledge, requiring objec-
new knowledge or to validate or rene ex- tivity of the researcher, others allow more
isting knowledge. Such evidence-based diversity in the method. The authors of
knowledge is the foundation of profes- this text believe a broad denition inclu-
sional nursing practice. As an inquisitive sive of both quantitative and qualitative
student of nursing, you may already have approaches is most useful to advancing
asked yourself some fundamental ques- the development of nursing knowledge.
BOX 1.1 Definitions of Research munity highlights the differences in nursing

practice and contexts across rst-world,
Research is a systematic, controlled, em- second-world, and developing countries.
pirical, and critical investigation of hypo- The American Nurses Association provides
thetical propositions about the presumed a denition that is broad enough to include
relations among natural phenomena (Ker- this variety within nursing practice:
linger, 1973, p. 11).
Research is a systematic, formal, rigor- Nursing research is investigation into
ous, and precise process used to gain solu- the area of knowledge in which the
tions to problems or discover and interpret physical and behavioral sciences meet
new facts and relationships (Waltz & and inuence one another in an effort
Bausell, 1981, p. 1). to study how health problems relate to
Research is a process in which observ- human behavior and how behavior re-
able, verifiable data are systematically col- lates to health and illness. It addresses
lected from the empirical world (i.e., the the human and behavioral questions
world we know through our senses) to de-
that arise in the treatment of disease, in
scribe, explain, predict, or control events
(Seaman, 1987, p. 3).
the prevention of illness, and in the
Research is an attempt to increase the promotion and maintenance of health
sum of what is known, usually referred to (ANA, 1976, p.1).
as a body of knowledge, by the discovery Dening nursing research requires de-
of new facts or relationships through a
termining the relevant knowledge for
process of systematic scientific enquiry
(Macleod & Hockey, 1989).
nursing. Since Florence Nightingales era,
Research is application of formalized nurses have been expanding their profes-
methods of obtaining reliable and valid sional practice foundation through re-
knowledge about empirical experience search efforts designed to develop and
(Chinn & Kramer, 1995, p. 75). test knowledge. This text denes nursing
research as research into phenomena
that are predominantly and appropriately
the responsibility of nurses in their pro-
2. What Is Nursing Research? fessional practice. This denition implies
that we as nurse researchers need to ana-
Nursing research is similar to research lyze the work that nurses do as a basis for
conducted by other professional groups determining what needs to be known and
or disciplines in that the same rigorous what gaps exist in nursings knowledge
steps must be followed and the research base. Nursing is a professional as well as
design and method of investigation must an academic discipline; hence, it incorpo-
be appropriate to the problem or ques- rates a clinical practice component as an
tion studied. What, then, distinguishes integral part of its activities. Nurses are
research in nursing from research con- responsible for provision of client care,
ducted by other health-care disciplines? administration of nursing and health-care
The past decade has witnessed a lively services, and education of practitioners.
debate in the literature in response to Therefore, research in nursing encom-
this question. The diversity of the human passes systematic investigation into each
experience and the nature of nursing prac- of these areas. We believe that this com-
tice require exibility in our denition, prehensive view has considerable poten-
which must make provision for the varia- tial to contribute to advancing the profes-
tions that occur across professional prac- sional identity of nursing and promoting
tice settings. This is particularly true excellence in practice through knowledge
today as globalization of our world com- development.
8 GETTING STARTED
3. What Are the Goals of practice empowers the nurse. At the

Nursing Research? level of the profession, research know-
ledge justies nursing services by es-
A goal is something toward which ones
tablishing their validity to clients and
efforts are directed. The goals of nursing
other important stakeholders.
research reect its importance to the dis-
cipline. Although they may vary some- To do research in nursing requires in-
what from researcher to researcher, most tellectual integrity, curiosity, knowledge
would agree on the following ve goals: of the scientic method, and a belief that
nursing is important enough to warrant
1. To produce an understanding of human
investigation. If we want to argue that
responses across the life span to health
nursing makes a difference, that our in-
and illness states in ever-changing envi-
terventions are cost effective and efca-
ronments. In relation to this goal, we
cious and our services socially relevant,
ask, Why do we nd regularities, or pat-
then we need to produce evidence
terns, in human responses? Which pat-
to support our claims. The research
terns are unique to partic-ular health or
process and scientic knowledge gener-
illness states and environments, and
ated through research are important
which are common across settings? We
tools to help us accomplish these goals.
want not only to document these regu-
larities but also to understand the com-
plexity of human responses to thera-
C. THE VARIETIES OF NURSING
peutic measures to promote health and
KNOWLEDGE
minimize the negative impact of illness
states.
1. Sources of Nonresearch-Based
2. To improve the quality of nursing
Nursing Knowledge
care and promote health through the
application of research methods. To A great deal of our nursing knowledge is
do this, nurses must ask questions not generated through research. Rather,
about their practice and seek system- nurses typically draw on knowledge from a
atic answers. variety of sources, each offering some-
3. To enhance the recognition of nursing thing unique to holistic nursing practice.
as a science. This recognition is im- These include tradition, authority, trial
portant in attracting to nursing the and error, personal experience, intuition,
brightest and best minds and in in- and logical reasoning. The difculty with
creasing the respectability of nursing nonresearch-based knowledge is that it
in the research community. has limited scientic predictability and
4. To enable nursing, as a practice disci- cannot be used to justify nursing actions or
pline, to control the provision of its provide new scientic information to di-
services. To do so, research on those rect practice. The following discussion of
services and the knowledge research alternative sources of nursing knowledge
can yield are essential. Knowledge pro- highlights the strengths and limitations of
vides a basis for professional power acquiring knowledge in this manner, creat-
and autonomy. ing the context for discussing the process
5. To empower nurses. Research efforts of acquiring knowledge through research.
help to empower nurses and distin-
guish nursings body of knowledge a. Tradition
from those of other sciences. On an in-
dividual level, the use of knowledge Folklore and tradition abound in nurs-
generated through research to change ing practice and are powerful sources
of knowledge. Many established customs to them. Conformity was expected and re-
and methods of procedure in nursing are warded. Today, nursing authorities more
embedded in our culture and passed down often include nursing professors, instruc-
from one generation of nurses to the next tors, clinical associates, and nurse man-
through verbal and written communica- agers. Although these individuals often
tion. The advantage of this source of possess useful information, it is never-
knowledge is that it is usually grounded theless important to question and chal-
in past experiences that were successful lenge their knowledge through empirical
and effective; therefore, it is usually an ef- testing. Many times the information is
cient way of solving a nursing problem. of value; however, prudent nurses will
Many traditions, however, have not be aware that knowledge from authority
been tested for accuracy or effectiveness. sources is not infallible.
For example, the tradition of restricted
visiting hours on childrens units was
thought useful in helping children deal c. Trial and Error
with the separation from home and family.
The trial and error approach to nursing
Research into separation anxiety and cop-
knowledge usually involves multiple at-
ing in hospitalized children revealed, how-
tempts to solve a particular problem un-
ever, that children benet greatly from
til a satisfactory solution is found. For ex-
the presence of a parent, family-centered
ample, a pediatric nurse who is consoling
care approaches, and unrestricted visit-
an immobilized child in the hospital may
ing privileges. Although tradition may of-
try a number of different play techniques
fer useful solutions to practice problems,
based on the childs developmental stage
nurses need to be aware that such knowl-
until nding one that is effective. This
edge has not withstood the test of em-
method of knowing is inefcient because
pirical validation. Nurses should be cog-
a number of attempts are usually required
nizant that new research information
before a solution is found. Depending on
may invalidate knowledge and practices
the situation, trial and error could be haz-
derived from tradition.
ardous to client safety, and it is often fal-
lible. Additionally, the solution that is
b. Authority found may be specic to a particular client
and setting and not generalizable to a
Authority becomes a source of nursing
broad population. A more structured ap-
knowledge when one accepts something
proach similar to the scientic process
an authoritative person says as truth.
would be more practical, efcacious, and
Historically, nursing has accepted an or-
frugal in terms of human and nancial re-
ganizational hierarchy in health care that
sources.
gave a great deal of authority to those
who had the highest level of education or
experience. For example, expert nurses d. Personal Experience
who passed on their knowledge and skills
to student nurses were considered au- Personal experience is a rich source of
thorities, as were physicians. Indeed, un- knowing in nursing. Extensive clinical ex-
til the latter half of the 20th century, physi- perience provides nurses with opportu-
cians were often the instructors of nursing nities to recognize patterns of responses
students, and they selected topics that in practice and then make predictions
perpetuated nurses role as their hand- about future responses based on obser-
maidens. In that era, nurses did not ques- vations of the practice. This important
tion the knowledge that was passed down source of knowledge allows nurses to
10 GETTING STARTED
transfer learning from one situation to a to develop an argument to support their

similar situation with another client. conclusions about a particular clinical sit-
The work of Benner (1984) has been uation. Two types of reasoning exist, de-
revolutionary in highlighting the im- ductive and inductive.
portant relationship between clinical ex- Common sense deductive reasoning is
perience and nursing knowledge. Benner that by which a nurse reaches a conclu-
identied ve levels of clinical expertise: sion by moving from the general to the
novice, advanced beginner, competent, specic. It is the approach used to test
procient, and expert practitioner. Al- predictions and validate existing relation-
though personal experience is a func- ships. For example, an obstetrics nurse
tional source of knowledge, it can be lim- has noticed over many years of practice
iting if the eld of experience is not that women who have no support person
sufciently diverse to allow an adequate with them while giving birth require more
breadth of rich experiences. Also, the val- supportive nursing care than those ac-
ues and beliefs of the knower can bias it. companied by a partner, mother, friend,
or professional labor coach. When this
nurse then admits a 22-year-old woman
e. Intuition who is in labor who arrives at the hospi-
tal alone by cab, the nurse anticipates
Intuitive knowing is the ability to under-
that additional care measures will be ap-
stand a situation or phenomenon as a
propriate for this client.
whole without reasoning or previous
Common sense inductive reasoning, by
study. It has been described as a gut feel-
contrast, moves from the specic to the
ing or hunch that causes a nurse to re-
general. Specic situations are observed
spond appropriately in unfamiliar situa-
and then combined into a larger, more gen-
tions. Nurses often seem to possess a
eral statement that can be tested through
tacit knowledge within themselves that is
research. The obstetrics nurse described
deeply rooted in the subconscious until it
observed hundreds of individual clients in
is required in a practice situation (Meer-
labor over many years. From these obser-
abeau, 1992). Intuition plays a critical
vations of individual clients, the nurse was
role in the reective process and, there-
able to propose a general theory about
fore, is central to nursing practice. It has
women who are in labor that could then be
been cited as the exclusive gift of expert
tested through formal research.
practitioners, but others see it as a uni-
The validity of nursing knowledge gen-
versal human experience (Mitchell, 1994).
erated through reasoning depends on the
However, controversy exists regarding its
accuracy of the information or premises
validity because it does not conform to the
with which one is working. The conclu-
recent requirements for evidence-based
sions are valid only if the premises (state-
practice in health carethat is, intuitive
ments) on which they are based are valid.
knowing is not predictable, measurable,
Research is used to conrm or refute
and generalizable. At the same time, it has
premises so nurses in practice may use
unquestionably enabled nurses to make
them.
signicant decisions in client care.
2. Four Ways of Knowing

f. Common Sense Reasoning
Epistemology is the study of the nature of
Reasoning is the ability to understand phe- knowledge. It seeks to determine how we
nomena and draw conclusions through know what we know. A number of authors
logical thinking. Reasoning enables nurses have studied ways of knowing in nursing.
A seminal piece of work by Carper (1978) c. Personal Knowledge as a Pattern

suggested that there are four fundamen- of Knowing
tal ways of knowing in nursing:
Personal knowledge involves the inter-
personal interactions and relationships
a. Aesthetics or the Art of Nursing between a nurse and a client. It is con-
cerned with the knowing and encounter-
Aesthetics is the study of the nature of ing of the individual self. Through the
beauty or art. In nursing, aesthetic knowl- component of personal knowledge, a
edge allows nurses to perceive the beauty nurse using therapeutic use of self re-
underlying each encounter with each jects approaching the client as an object
client. The art of nursing is best expressed and strives to establish an authentic per-
through the virtue of empathy, which is sonal relationship. This requires an ac-
the capacity to vicariously experience the ceptance that each client is not a xed en-
feelings of another. The more skilled a tity but someone constantly engaged in
nurse becomes in empathizing with the process of becoming. In therapeutic
clients, the greater the knowledge and un- encounters, nurses develop authentic per-
derstanding that will be gained of alter- sonal relationships with others, a process
nate modes of reality. Orem (1995) speaks possible only because they have learned
of the art of nursing as the individual about themselves. Personal knowledge is
nurses creativity and style in designing of extreme importance to all areas of
nursing care that is effective and satisfy- nursing practice. It is difcult to imagine
ing. Aesthetic knowledge involves percep- the development of practice without this
tion, understanding, empathy, and intu- essential element that is implicit in every-
ition, and acknowledges the value of thing a nurse does for a client. It is the
everyday experiences lived by individu- personal knowledge and opportunity to
als. In this respect, it is associated with the reect on personal feelings that enables
interpretive research perspective that will nurses to recognize the most personal as-
be discussed later in this chapter. pects of a situation and respond to the
needs of the particular client (Barragan,
b. Ethics or the Moral Component of 1998).
Nursing Knowledge
d. Empirics or the Science of Nursing
Ethical knowing in nursing focuses on
obligation, on what should be done. This Empirical knowledge refers to knowledge
pattern of knowing in nursing requires an of the experienced or empirical world.
understanding of different philosophical This type of knowledge is generated
positions regarding what is good, what is through scientic methods and is usually
right, and what is desired. The ethical organized into laws and theories that help
component involves more than knowl- to describe, predict, and explain phenom-
edge of morality and ethical codes; it in- ena. This knowledge is embedded within
volves judgment of moral value in relation the positivist perspective that is dis-
to nursing actions and intentions. Nursing cussed later in this chapter. In recent
involves carrying out deliberate actions years, much emphasis has been placed on
that are subject to interpretation and judg- developing scientic knowledge for nurs-
ment. The moral component of knowledge ing practice.
inuences difcult decisions that nurses Although the focus of this text is an ex-
must make in the context of increasingly ploration of the methods that create the
complex health care services. science of nursing, it is important to keep
12 GETTING STARTED
in mind that empirical knowledge is only the research process, some common re-
one component. A broad scope of knowl- search designs, and the use of computers
edge that emanates from both humani- in nursing research.
tarian and scientic perspectives is re-
quired to understand the profession of 1. Role of Philosophy in Nursing
nursing, the value orientation of the pro- Research
fession, and the nature and philosophical
foundations of practice. A philosophy of nursing provides an ex-
In summary, the nonresearch-based planation of the nursing world. It is con-
sources of nursing knowledge and the hu- cerned with the values and beliefs of
manitarian ways of knowing in nursing the discipline and with the values and
should not be judged against research- beliefs held by members of the discipline
based knowledge; rather, they should be (Meleis, 1997). According to Kimchi and
valued in their own right as legitimate Simmons (1994), a nursing philosophy
and useful. Nurses often experience situ- accomplishes the following four functions:
ations that require action and for which
there are no scientic answers. In such
It provides a framework for identifying
central concepts of the discipline.
circumstances, nurses draw on other
sources and ways of knowing that pro-
It provides assumptions that guide
theory development in nursing.
vide insight and understanding into that
particular situation.
It relates nursing to a particular world-
view.
It provides information on how one
may come to learn about the world.
D. ACQUIRING NURSING
KNOWLEDGE THROUGH A philosophy of nursing, therefore, pro-
RESEARCH vides a focus for nursing research. It is
similar to a camera lens: as you adjust the
Western cultures share a great condence lens, you rene your view of the world.
in and respect for science as the way to Similarly, as you adjust your philosophy of
truth. The scientic method is viewed in nursing, your view of the nursing world
western cultures as the most sophisti- will change. Nursings philosophical posi-
cated method of acquiring knowledge. It tion, in particular, inuences the type of
combines the process of logical reasoning questions posed by the researcher, and
with systematic planned investigation, these questions, in turn, inuence the re-
data collection, analysis, and evaluation. search method selected to study the
Quantitative nursing research involves problem. Thus, it is critical to examine
application of the scientic method to the relationship between the underlying
problems, questions, and phenomena of nursing philosophy and any particular
interest to the discipline. Although not in- nursing research study.
fallible, nursing knowledge generated The eld of nursing is in the midst of
through research is much more reliable an exciting debate on precisely how
than knowledge generated through tradi- to articulate its philosophy. To be useful,
tion, authority, and other sources. This the philosophy must be able to direct not
section begins by examining the relation- only research but also practice, educa-
ship between nursing philosophy and tion, and administration. Although much
nursing research, an important considera- diversity of thought exists, important ar-
tion before evaluating research or begin- eas of unity are emerging. The following
ning to conduct research of your own. We examples represent common ground in
then provide an overview of the stages of nursings philosophical positions:
Nurses value health over illness. guides the selection of variables, the
Nursings focus is the promotion of development of the research question
health and well being of persons across or hypotheses, and the interpretation
their life span. of ndings. A review of the literature
Nurses view humans as holistic, ratio- provides direction to the researcher in
nal beings searching for meaning in life. selecting the most appropriate frame-
Nurses believe caring is foundational work to guide the study development.
to practice. 4. Formulation of the research question or hy-
Nurses promote the protection of indi- pothesis. The research question is an
vidual rights and client autonomy. interrogative statement that directs
Nurses value respect for human dignity. the study. It contains a description of
Nurses believe truth is relative. the variables and the relationships
Nurses believe reality varies with per- among the variables. A hypothesis
ception. provides a predicted answer to the
Nurses believe quality of life is as im- research question. It predicts the re-
portant as quantity of life. lationship between two or more vari-
ables. The researcher refines the re-
search question based on the current
2. Stages of a Research Project state of knowledge about the research
topic.
Qualitative research follows a cyclical
5. Development of methodology plan. This is
process. Chapter 6 introduces each of the
the plan for how the study will actually
qualitative steps and provides guidelines
be conducted. It involves a series of
for performing each of them. Most quanti-
steps including selection of a research
tative research involves distinct stages
design, identication of the study pop-
that follow a linear sequence. We will in-
ulation, selection of the sample, imple-
troduce each of the quantitative stages
mentation of a pilot study, collection
now.
of data, and analysis of data. Each step
1. Problem identification. First, the researcher is discussed in detail in later chapters
determines the purpose of the research of the text.
project. Is a theory being tested, is a 6. Interpretation of the findings. The researcher
known relationship being explored in discusses the studys ndings in light of
detail, or is a solution being sought to the framework of the study and the re-
an applied problem? Usually, a general search question or hypotheses. How do
topic is chosen, and then the re- the results answer the research ques-
searcher proceeds to narrow the topic tion? Do they support the research hy-
to a one-sentence statement about a potheses and theoretical framework?
specic problem. The study results are interpreted in re-
2. Literature review. Thorough researchers lation to other ndings reported in the
review the scientic literature to nd literature. In this way, the research
out what has been discovered by other study contributes to building a body of
researchers working in the selected knowledge.
area. What kind of design is appropri- 7. Communication of the findings. The re-
ate for the problem that is being researcher writes a nal report on the
searched? Which variables are to be project. Media available to communi-
measured? Which hypotheses are to cate study ndings include colorful
be tested? poster presentations and oral presen-
3. Selection of conceptual framework or theory. tations at scientic and professional
The conceptual framework or theory meetings and conferences, publication
14 GETTING STARTED
in professional and scientic journals, making related to prophylactic use of con-

and establishment of a Web page with traceptives. In this case, the nurse would
regular updates about the research probably use an experimental design. Two
study. groups would be established randomly
and both would be measured on their in-
tention to use contraceptive methods.
3. A Preview of Research Designs One group (the experimental group)
would then be shown a video about family
Nurse researchers are attempting to
planning, and the other (the control
understand phenomena that are ex-
group) would be shown a video unrelated
tremely complex and subject to many
to the issue at hand. Finally, both groups
interpretationsnamely, the responses
would be measured again on their inten-
of the holistic person in various health-
tion to use contraceptives. In this case, the
related situations interacting with an ever-
experiment is designed to test whether a
changing environment. Although some
video has an impact on adolescents pro-
elements of human behavior may be
phylactic use of contraceptives.
simple, much of human behavior cannot
A quasi-experiment lacks either ran-
be understood satisfactorily with sim-
domization of subjects or the formation
ple, mechanistic models. (A mechanistic
of a control group. It is particularly useful
model is one that asserts that a particu-
when it is practically or ethically difcult
lar behavior is completely controlled by
to provide an intervention for only one
one or more external actions or events.)
half of a group. For example, a nurse re-
Many factors are usually involved, and
searcher might observe the parenting
they are involved differently, depending
skills of a group of new mothers before
on the patients health situation and on
and after they receive a series of post-
the environment. Thus, nurse scientists
partum home visits in an attempt to eval-
come to know, to understand the com-
uate the effectiveness of home visits on
plexity of human responses to health and
increasing new mothers parenting skills.
illness situations, by using a variety of re-
search designs, such as the four intro-
duced below. Later chapters explore b. Survey Designs
both quantitative and qualitative designs
A survey typically uses a data-collection
in more detail: experiments and quasi-
method with a series of questions admin-
experiments (Chapter 4), surveys (Chap-
istered to a particular population in order
ter 5), various qualitative designs (Chapter
to gain information about that popula-
6), eld studies (Chapter 7), and new wave,
tion. For example, a nurse researcher
contemporary applied approaches (Chap-
may wish to explore changes in parent
ter 8). Each approach is legitimate in its
satisfaction level with new ways of deliv-
own way, and no one design covers all
ering maternity nursing services that
problems. Thus, the selection of a design
have resulted from health care reform.
should be guided by what is most appro-
This researcher would probably survey
priate for the problem under investigation.
parents of newborn infants drawn from
the catchment area of a local maternity
a. Experimental and Quasi- hospital. The questions would address
Experimental Designs satisfaction with obstetric nursing ser-
vices and seek to measure factors that
A community health nurse might be in- the researcher identies as probable
terested in examining the impact of a causes of variability in parent satisfaction
birth control video on adolescent decision with nursing services. For example, are
parents satised with the current deliv- ating positive change. Unlike conventional
ery of nursing services in the maternity research, which mainly contributes to
setting? What factors inuence parent knowledge development that may not nec-
satisfaction? If parent satisfaction levels essarily be used in practice, the emphasis
have changed, what changed them? in contemporary applied research is to use
the research to inform the change. The
change may include emancipatory social
c. Qualitative Research Designs
and political aims, improvement in a given
Qualitative designs emphasize studying context, problem solving in an immediate
the world from the perspective of the par- situation, a planned intervention, transfor-
ticipants being studied, therefore em- mation, or empowerment. This new wave
phasizing the subjective dimension of trend serves as an important tool in devel-
human experience. A study examining oping nursing decisions, prescriptions,
the meaning of a mothers experience in and actions. It uses a multiplicity of de-
the loss of a newborn would be an exam- signs, including both experimental and
ple of a qualitative study (probably using nonexperimental designs, and both quali-
a phenomenological perspective). Such a tative and quantitative approaches.
study would require in-depth interviews Contemporary applied approaches are
with a few mothers about their experi- discussed in Chapter 8 and include femi-
ence with the loss of a newborn child. nist, action, health promotion, and evalua-
tion research. Box 1.2 illustrates an exam-
ple of action research by a group of nurses
d. Field Study Approaches who are engaged in a national study to
bring about positive change in the every-
Field studies include a number of quanti- day lives of girls and young women who
tative and qualitative designs in which are exposed to sexual harassment.
data are collected in naturally occurring
settings as opposed to contrived ones. A
nurse scientist may, for example, want to
4. Use of Computers in Nursing
examine how we have institutionalized the
Research
birthing process in our health-care system.
How do we organize the maternity ward? Researchers rely on computers for two
What are the rules of behavior and rituals major reasons. First, research may entail
surrounding the birth of a child? How are collecting information about many vari-
crises, such as the death of the child, man- ables from a large number of individuals.
aged? What are the unwritten tactics used Because computers work quickly, they
by nurses to support maternal clients, make it relatively easy to do calculations
newborns, their families, and visitors? This involving large numbers. Second, re-
researcher might decide to implement a searchers frequently build models involv-
participant observation eld study, taking a ing many variables; computer technology
position as a staff nurse in a city hospital to makes it possible to do the kind of multi-
study the factors involved. variate (many variable) analysis that is
necessary to test such models.
e. New Wave Research: Qualitative researchers are increas-
Contemporary Applied Approaches ingly relying on computers to assist in
searching, sorting, and classifying non-
The goal of contemporary applied re- numerical information. A number of com-
search is to produce knowledge that can puter programs have been developed to
be used for the specic purpose of gener- perform qualitative analyses. Programs
16 GETTING STARTED
BOX 1.2 Nurse Researchers at Work
SEXUAL HARASSMENT: EVERYDAY VIOLENCE IN THE LIVES

OF GIRLS AND WOMEN
Violence is a significant public health issue with and challenges of girlhood, and how they
consequences for individuals, families, and com- learned about what being a girl is supposed to
munities. The emergence of a rapidly growing be. Although no specific questions about vio-
international organization, the Nurses Net- lence or harassment were asked in the focus
work on Violence against Women Interna- group interviews, the theme of subtle and ex-
tional, speaks to the critical interest of nurses in plicit forms of violence emerged in all of the ses-
this complex problem. An insidious form of vi- sions. The girls described experiences that
olence that particularly affects girls and young evoked fear, intimidation, feelings of belittle-
women is sexual harassment. Growing evi- ment, and a diminished sense of self. When
dence suggests that sexual harassment is a com- speculating on factors that may have con-
mon occurrence in their everyday lives. This tributed to such experiences, the girls men-
project explored how violence becomes nor- tioned media, religious dogma, and educational
malized in the lives of girls and young women. factors. Notably absent from their discussion
The objectives of the project were twofold: was the role of gender, including male power
(1) to explicate the diverse ways in which girls and control within a patriarchal society.
and young women are socialized to expect vio- The study suggests that nurses have numer-
lence in their lives, and (2) to examine how so- ous opportunities to empower adolescent girls
cial policies, legislation, and institutions alleviate to deal with sexual harassment. Teaching girls
or perpetuate the problem in this population. a process of resistanceincluding behaviors
The study used a participatory action research such as adopting a range of healthful habits,
approach and was guided by the principles of speaking out against sexism, and maintaining
feminist research. Implicit in the assumptions of relationships with which they are able to speak
the research was the assertion that traditional freely and hear their own voicesmay help
notions of girls at risk may not be useful when girls minimize the impact of such violence on
addressing the topic of violence. Rather, given their lives.
the insidious and pervasive nature of many Nurses are in an ideal position to develop
forms of violence, including sexual harassment, health education programs for girls that in-
all girls must be considered at risk. corporate a gender-based analysis of the chal-
Focus group interviews were conducted in lenges girls face. In addition to empowerment
five national centers with ethnically and geo- program development, the study challenges
graphically diverse groups of girls. Questions nurses to simultaneously, actively, and loudly
explored the girls thoughts about a range of is- join in the larger political struggle against male
sues such as growing up female, likes and dis- dominance and one of its primary weapons,
likes about being a girl, the most significant joys sexual harassment.
SOURCE: Summarized from Berman, H., McKenna, K., Arnold, C., et al (2000). Sexual harassment: Everyday vio-
lence in the lives of girls and women. Advances in Nursing Science, 22(4), 3246.
such as Ethnograph, Superle, Frame- tured data indexing searching and theo-
work, and TEXTAN are popular programs rizing) is a software program developed
that perform content analysis. In addi- specically for qualitative data analysis.
tion, most word processing programs It is designed to aid researchers in han-
can perform sort operations and search a dling non-numerical and unstructured
text le for selected words or a string of data in qualitative analysis, by support-
words. NUD*IST (non-numerical unstruc- ing processes of coding data in an index
system, searching text, or searching pat- (general principles), and predictions. Early
terns of coding and theorizing about the nurse researchers modeled knowledge de-
data. Chapter 7 explains NUD*IST com- velopment on the physical sciences using
mands that are used in computer analysis the positivist perspective. Many labels,
of qualitative data. such as behaviorist, empiricist, and scien-
The SPSS (Statistical Package for the tist, are attached to its practitioners, along,
Social Sciences) is a collection of proce- of course, with some less-attering labels,
dures for processing social science and such as abstracted empiricist or number
nursing data. Versions of SPSS date back cruncher. Some critics of the positivist per-
to the 1960s. Most universities and many spective claim that positivist researchers
research organizations have SPSS. Al- have become overly impressed with the
though many statistical packages are numbers and the tools used to process
available, few can match the scope of them and have lost sight of the goal of try-
SPSS. Appendix A of this book explains ing to understand human behavior and re-
SPSS commands (SPSS for Windows ver- sponses. In any case, positivism is the pre-
sion 10.0) that are used in computer analy- dominant approach that nursing science
sis. Readers who have access to earlier has taken to develop knowledge. But, as
versions of SPSS will nd that, with a few we shall see, the past decade has wit-
modications, the commands presented nessed an interest in the interpretive and
in this book will work. In addition, a spe- the critical perspective in nursing.
cial SPSS supplement is available to in- Most positivists would agree with
structors using this text, which contains a Carlo Lastruccis denition of science as
series of computer assignments designed . . . an objective, logical, and systematic
to familiarize students with SPSS. method of analysis of phenomena, de-
vised to permit the accumulation of reli-
able knowledge (1967, p. 6). An objective
approach is designed to minimize bias, is
E. KEY METHODOLOGICAL
impersonal, and seeks its authority in
PERSPECTIVES IN
fact, not opinion. A logical approach uses
NURSING SCIENCE
deductive rules, and a systematic ap-
proach is consistently organized and
Among the research literature, we identify
makes use of such techniques as statisti-
three important perspectives in the devel-
cal analysis. Finally, reliable knowledge
opment of nursing science: the positivist
refers to knowledge one can count on,
perspective, the interpretive perspective,
knowledge that allows one to predict out-
and the critical perspective. Let us investi-
comes accurately: . . . reliable knowledge
gate the assumptions that these perspec-
is that which is both objectively and em-
tives make about science, human behav-
pirically veriable (Lastrucci, 1967). Pre-
ior, and values as well as their preferred
dictions are empirically veriable if they
methodologies and possible weaknesses.
can be tested for accuracy by making sys-
tematic observations.
1. The Positivist Perspective
Most people have some familiarity with a. Positivist Assumptions about

the positivist perspective because it is the Science and Human Behavior
one used in the physical sciences. It
emerged from a branch of philosophy There are a number of basic postulates,
known as logical positivism, which oper- or working assumptions, that positivist
ates on strict rules of logic, truth, axioms nurse scientists make in approaching
18 GETTING STARTED
their research. Again, following Lastrucci, Positivists are interested in understand-

we can summarize seven of the major ing the patterns of human response. An
postulates as follows (1967, p. 3746): emphasis is placed on identifying, mea-
suring, and expressing the relations
1. All behavior is naturally determined. To under- among variables with mathematical pre-
stand human behavioral responses, we cision. And because cause-and-effect re-
should look for causes in the natural lationships are of central concern to the
world. This postulate emphasizes a positivist, it is no surprise that emphasis
mechanistic view of the world: each is placed on prediction.
outcome is produced by one or more Positivists search out ways of testing
external causes. theories of human behavior. For posi-
2. Humans are part of the natural world. Human tivists, this implies using the traditional
responses can therefore be studied us- research process or quantitative re-
ing the methods used to study the be- search. Typically, this is accomplished
havior of other species, although the by establishing hypotheses (predictions)
study of human behavior is made far about relationships, measuring the vari-
more complex by the need to take lan- ables, and then analyzing the relation-
guage into account. ships to see if evidence can support or re-
3. Nature is orderly and regular. The natural fute the prediction. Efforts are then made
world is orderly, predictable, and there- to replicate (repeat) the study to see if
fore knowable. The patterns of nature similar ndings result when studying a
may be identied and observed. Events different population. According to this
that appear to be random may simply process, information gained from one
reect our inability to fully compre- study is never sufcient to include in the
hend the natural forces at work. body of science. A study must be repli-
4. All objective phenomena are eventually know- cated several times with similar results
able. There are no intellectual limits on produced each time before the informa-
what we may eventually know about tion can be considered a fact. Facts
nature or about human behavioral re- from studies are then related to each
sponses. other using abstract thought processes,
5. Nothing is self-evident. Our knowledge of in a way that seems to best explain the ab-
human responses should be demon- stract world.
strated objectively. And although we A key indicator of an adequate explana-
may wish to use folk wisdom or com- tion for positivists is the ability to predict
mon sense as a starting point, ulti- outcomes. Thus, positivists wish to iden-
mately we must test ideas systemati- tify the key causes in the variation of some
cally. variable. For example, positivists may be
6. Truth is relative. What is regarded as a sci- interested in identifying the factors that
entic truth today may be disproved contribute to health-promoting lifestyles
or modied tomorrow. There is a dy- in adolescent girls, or compliance with
namic element to what we know. Our medication schedule in asthmatic chil-
knowledge is always on the road to dren, or job satisfaction of staff nurses
some ultimate truth, but never quite working in nurse-run clinics, or the choice
reaches it. of nontraditional comfort measures for
7. Knowledge comes from experience. The belief women in labor. In each case, the re-
that we need to test our understanding searcher is interested in understanding
of the world systematically with knowl- which factors best predict the phenome-
edge gained through our senses is a fun- non under investigation. Thus, one mea-
damental principle of science. sure of the success of a study is when
most of the variation is accounted for, and maintain that science progresses when
when the causes can be linked to some researchers systematically test alterna-
general theory useful to nursing practice. tive explanations of behavior (attempting
Unfortunately, the reputation of nurs- to rule them out) without trying to sup-
ing science that deals with human report some pet theory or favored project.
sponses and natural systems is not im-
pressive when it comes to accurate
predictions. Generally, it is thought that c. Research Designs Associated with
only the physical sciences are sufciently Positivism
well developed to predict with mathe-
matical accuracy. Indeed, most of the The positivist perspective relies mainly
highly predictable outcomes in science (but by no means exclusively) on experi-
are based not on natural, unaltered sys- ments, surveys, and secondary data analy-
tems, but rather on engineered sys- sis (information collected by persons
temssystems designed to take into ac- other than the researcher). Typically,
count various inuencing factors. positivist perspectives rely on some form
For example, you can predict that no of numerical analysis rather than on ver-
matter what the weather, the bridge you bal descriptions.
drive over every day to work will be there
for you tomorrow. It will be there be-
cause it is designed to withstand weather d. Criticisms of Positivism
and trafc conditions well beyond its nor-
Many criticisms have been leveled against
mal loads. So unless something cata-
the positivist perspective to knowledge.
strophic happens, the bridge will be
Some of the major criticisms are:
there. In contrast, the leaf that blew off
the tree in your yard may or may not have Some critics argue that value-free re-
moved since yesterday. To estimate its search is an unattainable goal. Some
current location would be foolhardy be- studies have shown that despite the
cause so many factors could inuence best intentions, it is extremely difcult
where it has moved to since yesterday. for a researcher to prevent his or her
Similarly, a huge challenge confronts a biases or expectations from exerting
nurse who wishes to predict the amount some inuence on the results of a
and type of social support required by study (see Chapter 9).
adolescents to resist smoking. So many Some critics argue that the so-called
factors may inuence the adolescents value neutrality of positivism is itself a
decision to smoke or not to smoke in a value.
given context. Prediction for a positivist Other critics argue that some areas
nurse scientist is akin to estimating the within nursing do not lend themselves
whereabouts of the leaf. Natural systems to scientic measurement. For exam-
pose a tough challenge for researchers ple, it is difcult to quantify the experi-
keen on prediction. ence of meaning in life in the elderly,
the phenomenon of courage in the face
b. Positivist View of the Role of Values of terminal illness, or resilience in vic-
in Research tims of sexual abuse.
Some scholars argue that the posi-
Positivists argue that research should be tivistic view ignores a crucial aspect of
value free. Researchers should put their social realitynamely, that different
personal values aside to avoid inuenc- people may experience and perceive
ing the outcomes of studies. Positivists the same events differently. The sub-
20 GETTING STARTED
jective experience of a clients world is tionary changes within the eld of nurs-
an alternate social reality that may be ing as more researchers use it to develop
explored by a nurse scientist; it is the the elds knowledge base.
one emphasized particularly by those
who use the interpretive perspective.
Qualitative researchers believe that a. Interpretive Assumptions about
the whole is greater than the sum of Science and Human Behavior
the parts and that in their treatment of
data, positivists treat only the isolated Rather than conning itself to behavior
partstherefore, positivists seem to alone, the interpretive perspective ex-
be dening the whole as a sum of the amines how people make sense of their
parts. lives, how they dene their situation, and
how their sense of self develops in inter-
action with others. In this perspective,
2. The Interpretive Perspective humans act and interact on the basis of
symbols that have meaning for them. Hu-
The German scholar Max Weber (1864 mans are always in a process of becom-
1920) was particularly inuential in devel- ing; they are inuenced by how they see
oping methodological approaches that themselves, by how others see them, and
stressed the importance of the interpre- by what they want to become. Symbolic
tation individuals put on their actions interactionists, for example, stress the
and on the actions and reactions of oth- idea of role modelingthat is, the extent
ers. He emphasized Verstehen, the empa- to which an individual chooses to emu-
thetic understanding of behavior. The late someone elses attitudes and behav-
researcher should try to imagine how a iors. Note that in this view, the individual
particular individual perceives social ac- is actively choosing to be similar to some-
tions. How does the individual feel? What one else. In contrast, most theories that
are the individuals motivations? What use the positivist or critical perspectives
meaning does the individual attach to a emphasize the extent to which individu-
particular event? als are shaped, or molded, by the institu-
The interpretive perspective and tions in their society.
methodologies were further developed The interpretive perspective believes
by others outside of nursing, such as a single reality does not exist. Rather, re-
Mead (1934), Goffman (1962), Glaser and ality is based on perceptions, so it can be
Strauss (1967), Lincoln and Guba (1985), different for each person and can change
Denzin and Lincoln (1994), and Strauss with time. Interpretive scientists believe
and Corbin (1997). However, within the that what we know has meaning only
past two decades, this perspective has within a given context. Meaning is pro-
gained increasing popularity as an appro- duced by perceptually putting pieces to-
priate approach for exploring the depth, gether to make wholes. Because percep-
richness, and complexity of nursing phe- tion varies with the individual, many
nomena. The works of Chenitz and Swan- different meanings to a situation or event
son (1986), Leininger (1985, 1990), Morse can exist.
(1991, 1998), Parse (1987), Sandelowski The philosophical underpinnings of
(1995), Streubert and Carpenter (1999), the interpretive perspective are very
and Stern (1998) have brought this per- much in keeping with the nursing elds
spective to the forefront of knowledge de- own understanding of the world. Its focus
velopment in nursing. The interpretive on a context-related and holistic ap-
perspective will no doubt undergo evolu- proach to knowledge development that
underpins practice is consistent with described in detail because there is a de-

many of the elds central values (Mas- sire to provide in-depth understanding of
terson, 1996). This perspective encour- the social world from the participants
ages us to work with clients in their own view. A key question for these researchers
environment to enhance our understand- is, Does the explanation offered make
ing of the emotions and perceptions in- sense to the people whose behavior is be-
volved in a given situation. As an exam- ing explained? Communication of the re-
ple, Finfgeld (1998) explored courage sults of such studies usually emphasizes
among middle-aged adults with long-term rich textual descriptions rather than nu-
health concerns. This study described merical analyses.
the process of being courageous and the
role that nurses can play in bolstering
clients courageous desires and behav- d. Criticisms of the Interpretive
iors. It provided specic examples of Perspective
skills that nurses should use in their prac-
tice with chronically ill clients. There- As we have seen, the interpretive per-
fore, this study is an example of how spective differs markedly from the posi-
knowledge gained from the interpretive tivist perspective. Any proponent of the
perspective can be used to improve nurs- existence of general abstract laws would
ing practice. criticize the interpretive perspective for
its subjectivity. Moreover, some critics
reject the interpretive theorists assess-
b. Interpretive Views of the Role of ment that all values are equally valid and
Values in Research point out that this assessment is itself a
value. Other critics argue that, with its
Interpretive nurse scientists argue that
emphasis on eld studies, the interpre-
values are relativethat is, they would
tive perspective does not enable the re-
argue that denitions of what constitutes
searcher to make clear generalizations.
appropriate or inappropriate behavior
We cannot easily determine whether the
depend on the socialization that one has
ndings of a case study are particular to
received from ones society. These de-
it or more generally applicable. The dif-
nitions shift over time and across soci-
culty is even more acute when studies
eties. Researchers should try to under-
cannot be replicatedan ethnomethod-
stand (empathetically) and to explain the
ological study of individual understand-
values of the people being studied (i.e.,
ings of events would be a good example.
the actors).
The emphasis on the single case gets us
into an old dilemma of knowing more and
c. Research Designs Associated with more about less and less, but in fairness,
the Interpretive Perspective this criticism can be leveled in varying
degrees against all types of systematic in-
The interpretive perspective relies mainly quiry. Finally, if emphasis is placed on the
on eld studies and emphasizes using individual level of analysis, it seems to be
participant observation and in-depth in- difcult to understand social patterns: to
terviewing techniques. Some of the most what extent can we understand interrela-
frequently used qualitative methodolo- tions among client groups by studying in-
gies in nursing include phenomenology, teractions among individual clients? But
grounded theory, ethnography, and eth- the response to this criticism may be that
nomethodology. Interpretive eld studies no nursing or related theory can hope to
typically involve a few individuals who are explain all client behavior.
22 GETTING STARTED
3. The Critical Perspective of practice (Stevens and Hall, 1992). The

critical paradigm can help nurses to see
Numerous schools are included under the world through a different lens. It can
the critical perspective, but the theorists be viewed from a global perspective in
generally credited with developing this terms of the world of nursing and its de-
perspective are Karl Marx (18181883) velopment. Also, it can be seen from a
and George Simmel (18581918). Variants practical perspective in terms of multi-
of critical theory abound in all social sci- disciplinary relationships in which issues
ence disciplines. In brief, the essential such as gender, relationships with medi-
themes that a critical researcher might cine, and the perceived role in society
explore include the scientic study of so- have caused nurses to feel oppressed
cial institutions and the meanings of so- (Barragan, 1998). Research in this tradi-
cial life; the historical problems of domi- tion can lead to emancipation for nurses
nation, alienation, and social struggles; and a valuing of their contributions to
and a critique of society and the visioning health care.
of new possibilities (Creswell, 1998). The
philosophy of critical social theory guides
the researcher in seeking to understand a. Critical Assumptions about Science
how people communicate and how they and Human Behavior
develop symbolic meanings in a society.
According to the critical perspective,
Many of the meanings in society have
human behavior consists of different
been taken for granted rather than dis-
groups who are attempting to enhance
cussed and debated. Nurses need to be
their interests at the expense of less pow-
aware of this because a number of mean-
erful groups. Whether it is physicians ex-
ings impede free and equal participation
erting inuence over members of the
of nurses and clients in society. A nurse
health-care team or men exercising dom-
researcher using this perspective tries to
inance over women, this perspective
uncover constraints that limit full partic-
stresses that conicting interests charac-
ipation by all members in society. Under-
terize human relations. Proponents of the
standably, then, in many ways, feminist
critical perspective argue that scientists
theory ts well with the critical perspec-
have an obligation to act as advocates
tive and could be considered a subset
working for changes in society, changes
of critical social theory even though,
needed to bring social justice to all. Re-
methodologically, feminism draws much
searchers help bring about such changes
of its inspiration from the interpretive
by making us more sensitive to social
perspective. Proponents of the critical
problems. In turn, this knowledge em-
perspective share a common desire to
powers citizens, helping them to become
improve the condition of humanity.
agents of social transformation (Fay,
Participatory action research uses
1987, p. 27). The fundamental goal of the
critical social theory methods. It is re-
critical perspective is to bring about a
search for the purpose of taking action
truly egalitarian societyone with an
and creating change. The focus is on
equality of opportunity as well as an equal-
learning about how people actually expe-
ity of result.
rience a specic issue or problem. This
knowledge is key to knowing what ac-
tions will make a practical difference in b. Critical View of the Role of Values
peoples lives and why. Through action in Research
research, challenges can be made to cur-
rent nursing practice and, on some occa- Critical scientists forthrightly declare
sions, even result in the transformation that whereas certain values are correct,
others are not. In short, the critical per- ignoring all others. Critics maintain that
spective takes an absolutist view of val- critical researchers are unlikely to make
ues. Researchers working in the critical a rigorous attempt to exclude competing
tradition would argue that the relativist explanations of behavior or to nd sup-
position of mainstream social science port for competing value systems. In fair-
has had the consequence of supporting ness, however, this criticism could be lev-
the established order in society, and has eled at all perspectives.
had, therefore, little impact on reducing Table 1.2 summarizes the main ideas
social inequities. Scientists in the critical of the three major contemporary method-
tradition favor imposing moral absolutes ological perspectives. Each perspective
in order to deal with inequalities. can increase our understanding of human
responses to health and illness situations
c. Research Designs Associated with across the life span. Rather than rejecting
Critical Perspectives any one of them outright, we would do
well to regard each perspective as poten-
The research designs associated with the tially valuable in our quest for nursing
critical perspective span a broad range. knowledge.
But, given the interest in understanding the
relations between groups in society and in
understanding how social change occurs, it F. TYPES OF RESEARCH: SOME
is no surprise that critical social scientists IMPORTANT DISTINCTIONS
tend to work with historical materials and
thus pay particular attention to the analy- Before you can begin to conduct re-
ses of secondary data. This focus tends to search, you need to become familiar with
emphasize macrovariables (i.e., proper- a variety of choices of techniques, types,
ties of societies) rather than microvari- and populations. These distinctions are
ables (i.e., properties of individuals). described in this section and are summa-
Research and explanations are judged rized in Table 1.3.
to be valid if they lead to an improvement
in the social condition of humanity. The 1. Quantitative Versus Qualitative
critical perspective thus has a strong prac- Approaches
tical orientation. Social science analysis is
seen as a means of achieving greater social Research approaches can be classied as
justice for all. either quantitative or qualitative. The dis-
tinction is based on the degree to which
d. Criticisms of the Critical the analysis is done by converting obser-
Perspective vations to numbers or using narrative text
to describe human experiences. The dis-
Similar to the positivist perspective, the tinction also reects differences in the
critical perspective relies on general theo- types of questions asked, the kinds of evi-
ries. But positivists, as well as interpretive dence considered appropriate for answer-
researchers, would reject the imposition ing a question, and the methods used to
of absolute values. Positivist critics would process this evidence. It is probably best
argue that the difculty with adopting a to think of the quantitativequalitative dis-
value position in any scholarly discipline tinction as a continuum.
is that it enhances the likelihood of dis-
torting research to promote ones per- a. Quantitative Research
sonal values. Critical researchers may be
too selective, reporting only ndings that Quantitative research seeks to quan-
are compatible with their values and tify, or reect with numbers, observations
24 GETTING STARTED
Table 1.2 Summary of Three Key Methodological Perspectives in Nursing Research

Criterion Positivist Interpretive Critical
View of science A tool for uncovering A tool for understanding A tool that should be
general laws of cause and the reality experienced used to improve the
effect in social behavior by people condition of the
oppressed
View of human Caused by forces acting Determined by context Consists of groups
behavior on the individual and individual perception attempting to exploit
Characterized by of meaning others for their own
regularity and order advantage
Goals of research To predict behavior; to To provide an adequate To improve the social
test general theories of reection of peoples conditions of the
behavior by testing of experience of the oppressed
hypotheses social world To achieve a just
Testing grounded theory society
Advocacy
Role of values Research should be value Research should be value Absolutist; research
in research free; relativistic free; relativistic should impose moral
absolutes derived
from theory
Research designs Surveys, experiments, In-depth interviews, Historical, comparative,
associated with quasi-experiments, participant observations, interviews, advocacy
secondary data, historical eld studies, document research (uses both
analysis (tends toward analyses (tends toward qualitative and
quantitative orientation) qualitative orientation) quantitative
approaches)
about human behavior. It emphasizes pre- of the research problem. Theories pro-
cise measurement, the testing of hypothe- pose explanations of phenomenahow
ses based on a sample of observations, things work, how parts are intercon-
and a statistical analysis of the data. Quan- nected, and how things inuence each
titative researchers attempt to describe other. Theories provide explanations to
relationships among variables mathemati- account for social patterns or for rela-
cally and to apply some form of numerical tionships. Many theories are required to
analysis to the relationships being exam- account for the multitude of experiences
ined. Quantitative researchers, like physi- encountered by human beings. The more
cal scientists, treat their subject matter general the theory, the greater the num-
like an object. Quantitative research is ber of predictions that can be derived
characteristic of the positivist perspec- from it. Indeed, the number of predic-
tive. It is also used (together with qualita- tions that can be derived from a theory is
tive research) in the critical perspective. also a measure of the theorys power.
Box 1.3 provides an example of quantita- Theories may be viewed as lying on a con-
tive research. tinuum of explicitness. At one end of this
Quantitative research can be conceived continuum, we would place a theory in
of as having three levels: the theoretical, which there are detailed statements of
the conceptual, and the operational. the relationships between the concepts
of the theory and a specication of its
(i) Theoretical Level underlying assumptions (in Chapter 2 we
The theoretical level in a project is the refer to these as grand theories); at the
most abstract, general conceptualization other end, we would place the theories
Table 1.3 Summary of Key Distinctions in Types of Research

Unit of Analysis
Individual Aggregation
Dimension One or few Several or many One or few Several or many
Tendency in Qualitative Quantitative Qualitative Quantitative
orientation
Types of study Case study Survey Case study Comparative
Phenomenology Experiment Participant
Grounded theory Quasi-experiment observation
Ethnomethodology
Disciplines Psychiatry Psychology Anthropology Anthropology
Sociology Sociology Sociology Sociology
Nursing Political science Political science Political science
History Archaeology Archaeology
Education Nursing Nursing
Nursing
Example Describing the Variations in desired Describing changes Comparing birth
birthing family size by gender in birth rates over rates by
experience and age time in one community
community
Types of Thematic analysis; Statistical Content analysis of Statistical
analysis verbatim interview; verbatim
quotations quotations
that offer explanations of particular rela- weight, intelligence, peer pressure, social
tionships (these are identied as practice support, and anxiety. Derivations made
theories). at the theoretical level may be formed
Nurse scientists use theories to de- into conceptual hypotheses. A concep-
scribe, explain, or predict limited proper- tual hypothesis is a statement of the re-
ties of reality. An important role of lationship between two or more concep-
methodologists is to try to rene them to tual variables. Ordinarily, a hypothesis
see if they hold true under all conditions. will take the form of the greater X, the
It is through efforts to disconrm theories greater Y. For example, the higher ones
that we extend our general knowledge of socioeconomic status, the higher ones
human behavior. One way to test a theory perceived health status.
is to derive a prediction (or hypothesis) The conceptual denitions of vari-
from it and then test that hypothesis. ables serve two important purposes.
Chapter 2 discusses the nature of theo- First, they should provide a clear state-
ries and ways of testing them in greater ment of what is meant by the variables.
detail. Second, they should help us decide how
each variable should be measured. For
(ii) Conceptual Level example, if we dene the concept socio-
The conceptual level denes the vari- economic status as differences in access
ables that are to be used in the research. to scarce resources, then we would try
A conceptual variable is an idea that has to measure the variable with indicators
a dimension that can vary. Conceptual that reect this denition as precisely as
variables can be relatively simple or possible. This brings us to the opera-
quite complex. Examples include gender, tional level, which we will consider next.
26 GETTING STARTED
THE EFFECTS OF RELAXATION THERAPY ON PRETERM

LABOR OUTCOMES
The purpose of this experimental study was to telephone contact would contribute to the
determine the effect of relaxation therapy on treatment effect. After several weeks in the
preterm labor outcomes in a sample of 84 study, 23 subjects in the experimental group
pregnant women. Participants were recruited reported that they had ceased doing the exer-
from private physician offices and a hospital- cises. Data from this group of subjects were re-
based maternity triage unit. A total of 44 partic- tained but analyzed as a third group.
ipants were assigned to the experimental group Results suggest that women who remained
and 40 were assigned to the control group. loyal to the experimental protocol of daily relax-
The experimental treatment consisted of ination exercises had significantly longer gestation
home audiotaped instructions on strategies for periods, delivered larger newborns, and had
conscious release of tension. Subjects were in- higher rates of pregnancy prolongation than ei-
structed to do the exercises daily. Progress was ther the control group or the nonadherent
monitored weekly through telephone contact group. These findings suggest that teaching daily
by the researcher. Subjects in the control group relaxation strategies to pregnant mothers can
were also contacted by telephone weekly and reduce the incidence of premature labor and the
asked how they were progressing. This was complex problems that result from it. The use of
meant to protect against the possibility that the such therapy warrants further investigation.
SOURCE: Summarized from Janke, J. (1999). The effects of relaxation therapy on preterm labor outcomes. Journal
of Obstetric, Gynecologic and Neonatal Nursing, 28, 255263.
(iii) Operational Level (iv) Linkages between Levels:

Operationalization refers to the selec- Validity and Reliability
tion of indicators (measures) to reect The theoretical, conceptual, and opera-
conceptual variables and to the imple- tional levels of a research study do not ex-
mentation of a research project. If so- ist in isolation from one another. Important
cioeconomic status is dened as differ- linkages connect them: testing a theory
ences in access to scarce resources, any properly entails documenting explicitly
measurement should attempt to reect the connections between the theoretical
this denition. In this case, a measure of and the conceptual levels and between the
annual income might appropriately reconceptual and the operational levels.
ect the conceptual variable, as dened. Quantitative nurse researchers use
And, in the study of perceived health sta- two terms, validity and reliability, to refer
tus, the classication of health status into to the connection between the concep-
excellent, very good, fair, and poor would tual and the operational levels. Validity
constitute a measure of the concept. The refers to the extent to which a measure
operational level consists of the mea- reects a concept, reecting neither
surement of variables as well as the col- more nor less than what is implied by the
lection and analysis of data. Later chap- denition of the concept. It is not unusual
ters provide a number of suggestions for for researchers to use markedly different
the measurement of variables and a vari- indicators for similar conceptual vari-
ety of procedures for collecting and ana- ables. Although one might dene socio-
lyzing data. economic status as differential access
to scarce resources, another might de- that are presented in tables and graphs.
ne it as a hierarchy of respect and pres- Variables are sometimes cross-classied
tige. Measures are valid to the extent to show how one variable changes as an-
that the chosen indicators reect the other variable changes. A table might
concepts as dened. show, for example, how average levels of
Reliability refers to the extent to job satisfaction of nurses vary by degree
which, on repeated measures, an indica- of perceived autonomy in the workplace.
tor will yield similar readings. One can This same nding could also be presented
think of reliability as the extent to which in the form of a graph. A great variety of
a measurement will produce similar read- statistical procedures are available to
ings for similar phenomena. A tire gauge quantitative researchers. It is important
that indicates 26 lb of pressure now, but to learn which procedures are appropri-
29 lb a moment earlier, suggests an unre- ate for a given research problem.
liable gaugeeither that or a bad leak in
the tire. In survey research, we some- b. Qualitative Research
times repeat a question to test for relia-
bility. Both responses to the question Qualitative research emphasizes verbal
should be the same if the item is generat- descriptions and explanations of human
ing reliable responses. (The concepts of behavior. Rather than concerning itself
validity and reliability are discussed in primarily with representative samples,
greater detail in Chapter 13.) qualitative research emphasizes careful
and detailed descriptions of life experi-
(v) Types of Questions Asked ences in an attempt to understand how
The types of statements or questions the participants experience and explain
asked in quantitative research often de- their own world and give meaning to it.
scribe variables, examine relationships The tools for gaining information include
among variables, and determine cause- participant observation, in-depth inter-
and-effect interactions between vari- views, or an in-depth analysis of a single
ables. Typical questions would include: case. At the macro level, qualitative re-
What are Hispanic-American mothers searchers tend to look at whole institu-
beliefs when their children have phys- tions or organizations; at the micro level,
ical disabilities? qualitative researchers focus on individ-
What support factors in the work envi- ual behaviors and responses. Qualitative
ronment are most important in deter- research is characteristic of the interpre-
mining the level of job satisfaction ex- tive perspective. It is also used (together
perienced by nurses in critical care with quantitative research) in the critical
units? perspective. Box 1.4 provides an example
What is the relationship between use of a qualitative research project.
and need of nursing services in rural We should note that the stages of a
areas serviced by independent nurse qualitative research project include most
practitioners? of those described earlier, but the se-
Is a school-based, health-promotion quence is typically cyclical rather than
intervention program effective in in- linear. This is because qualitative meth-
creasing the rate of healthy lifestyle ods are concerned with gaining a deeper
choices by adolescents? understanding of phenomena from the
participants perspective, which is ever
(vi) Types of Analysis evolving. Thus, inductive approaches to
In quantitative research, ndings are typ- inquiry are used, in which the researcher
ically expressed in terms of relationships moves back and forth between the data
28 GETTING STARTED
A NURSE PRACTITIONER MODEL OF PRACTICE

The purpose of this qualitative investigation was nursing role functions and emphasizes holism,
to describe the nurse practitioner role in a caring, and a health perspective for critically ill
neonatal intensive care unit (NICU) so that an neonates and their families. They function in a
advanced practice nursing model unique to the highly technological environment with a high
role of the nurse practitioner in the NICU level of responsibility and accountability. As one
could be developed. An ethnographic case NP stated:
study approach was used. The practices of
I think that advanced practice allows you to
seven NICU nurses were studied through par-
practice with a combination of the nursing and
ticipant observation and interview procedures.
medical model. I think that not only do I make
The researcher, as ethnographer, operated
medical diagnoses and perform diagnostic pro-
in the participant-as-observer role and there-
cedures and order the therapeutic regimens or
fore had access to a wide range of information.
prescriptions but I think my role is different from,
Shadowing the NPs, the researcher took exten-
say, a physician in that I also look at the nursing
sive notes and used a tape recorder to record
piece of it and that is the effect that illness has
observations. Unstructured, open-ended inter-
on the infant and the family.
views were conducted with each NP at the con-
clusion of the observation period. This study clearly delineates the role of the
Data collection occurred over a 15-month NP in NICU as one that blends nursing and
interval. Data were analyzed for the presence of medicine within a holistic approach to promot-
themes and patterns. The findings were re- ing health in critically ill neonates. Future re-
viewed with each participant for confirmation. search should address the outcomes of the
Results suggest that NPs practice within a NPs holistic focus on the health of neonates
model of care that incorporates medical and and their families in NICUs.
SOURCE: Summarized from Beal, J. (2000). A nurse practitioner model of practice. The American Journal of Maternal
Child Nursing, 25(1), 1824.
collection and analysis stages to derive a How do nurses handle patients who
dynamic truth that reects human inter- refuse to follow instructions?
action in social settings. What is it like to be diagnosed with a
(i) Types of Questions Asked terminal illness?
The types of questions asked in qualita-
(ii) Types of Analysis
tive research often concern how social
In qualitative research, ndings are typi-
systems operate, how individuals relate
cally expressed by quoting interviews or
to one another, how individuals perceive
relating experiences the researcher has
one another, and how they interpret their
had in the eld. The ndings are com-
own and others behavior. Questions
pared with the literature to see if similar
such as these are typical:
results have occurred in other studies.
What is the meaning of the experience Most nal reports have few, if any, tables
of participating in public health educa- or graphs. Because much qualitative re-
tion sessions for mental health clients? search is based on a small number of par-
What is the experience of men who are ticipants or on an in-depth examination
primary caregivers for spouses with of one group, it is often inappropriate to
Alzheimers disease? even attempt to quantify the results.
Various sections of Chapters 6 and 7 and achieve goals. The eld of nursings
elaborate on the procedures that may be emergence alongside medicine has
used to conduct different types of quali- crystallized this view. If the researcher
tative studies and present an alternative reects beliefs about person, health,
view to that of quantitative researchers nursing, and the environment from the
on such issues as the meaning of validity simultaneity paradigm (alternate world-
and the goals of research. view), he or she will tend to pose prob-
lems that are best suited to a qualitative
approach. The simultaneity paradigm
c. Choice of Strategy views person as a unitary being in con-
tinuous mutual interrelationship with
Although considerable debate exists
the environment, co-creating health
over the relative merits of quantitative
through mutual interchange with the
and qualitative research strategies, the
environment (Parse, 1987). The emer-
issue is a false one, to a large degree. It is
gence of different philosophies and
false because both approaches have
nursing paradigms provides grounding
their legitimate place in nursing research.
for research and is necessary to ad-
The choice of strategy is inuenced by
vance the boundaries of the discipline.
the following factors:
Sometimes the author of a research re-
The nature of the question asked. If
port will use a qualitative reporting style
the question has to do with the nature
in order to communicate effectively.
of a human experience, a small number
Qualitative research uses language and
of observations, or a single case, then
presentations familiar to all educated
the use of quantitative techniques
members of a society.
would be inappropriate. If the ques-
It is also important to note that there
tion has to do with patterns of human
are qualitative dimensions in quantita-
responses or descriptions of whole
tive research. In the process of turning
populations or a test of relationships
observations into numbers, there are typ-
among variables or differences, the ap-
ically a number of judgment calls, a num-
proach will be quantitative.
ber of qualitative decisions that must be
The predisposition of the researcher.
made. So, whether the researcher wishes
If the researcher has a preference for
to decide how to design a question to nd
qualitative research strategies, he or
out how much self-esteem someone has,
she will tend to ask questions that are
or how to classify educational programs
best answered by using such methods.
as promoting health, or what determines a
The philosophy and nursing paradigm
healthy lifestyle choice, a number of sub-
(worldview) of the researcher. If the
jective judgments must be made. And al-
researcher reects beliefs about per-
though these may be dened numeri-
son, health, and environment (the phe-
cally, there is nonetheless a qualitative
nomena of concern to nursing), from
dimension to the research decision. Vir-
the totality paradigm, the researcher
tually all studies that claim to be quanti-
will tend to pose questions and study
tative have important qualitative dimen-
problems best answered by quantita-
sions to them.
tive methods. The totality paradigm
has been the primary operating belief
system in nursing. It posits the person d. Methodological Triangulation
as a total organism with bio-psycho-so-
cial-spiritual features that interact with Methodological triangulation is the ap-
the environment to maintain balance plication of diverse methods to generate
30 GETTING STARTED
and collect data about one phenomenon both quantitative and qualitative ap-
(Kimchi, Polivka, and Stevenson, 1991; proaches have value for specic situa-
Creswell, 1994). The methods may be tions. They use both methods but usually
drawn from within-methods approaches, in a parallel manner in a study with little in-
such as different types of quantitative data tegration of the procedures or the ndings.
collection strategies in the same study The situation dictates the methods a sit-
(e.g., a survey and experiment), or it may uationalist will use. Pragmatists attempt
involve between-methods approaches to integrate quantitative and qualitative
drawing on quantitative and qualitative methods in a single study. They argue that
data collection procedures (e.g., ethnogra- a false dichotomy exists between the two
phy and survey research) in the same approaches and that researchers should
study. Triangulation is a term borrowed make the best available use of techniques
from the navigational tradition and refers to answer questions of substantive impor-
to the use of multiple referents to draw tance. This perspective views triangula-
conclusions about what constitutes truth. tion as a means to richer and more in-
It is based on the assumption that any bias sightful analysis of complex phenomena
inherent in particular data sources, inves- rather than an end in itself. The increasing
tigators, and methods would be neutral- number of triangulated studies in the liter-
ized when used in conjunction with other ature suggests that nursing is moving in
data sources, investigators, and methods. the direction of the situationalist and prag-
Nurse researchers are using method- matist perspectives.
ological triangulation with increasing fre-
(ii) Models of Combined Designs
quency as a means of achieving a more
Creswell (1994) has advanced three mod-
complete understanding of nursing phe-
els of combined design as a means of
nomena. The complex nature of nursing
adding scope and breadth to a study.
lends itself well to study using multiple
These include:
methods.
The two-phase design. Here quantita-
(i) Perspectives on Methodological tive and qualitative studies are pre-
Triangulation sented and discussed in two distinct
Some researchers believe triangulation is phases. This enables the researcher to
a method for linking quantitative and qual- clearly separate the two paradigms
itative methods in a single study. In 1959, and respective assumptions behind
Campbell and Fisk sought to use more each phase. The disadvantage is that
than one method to study a psychological it becomes difcult for the reader to
trait to ensure that the variance was re- see the connection between the two
ected in the trait and not in the method. phases.
Hence, the triangulation method is not The dominant/less-dominant design.
new to social scientists. It is, however, a Here one paradigm, either quantitative
new method for nursing science. or qualitative, dominates the study,
Rossman and Wilson (1985) identify and a small component of the study
three perspectives on triangulation: purist, (typically, the data collection stage) is
situationalist, and pragmatist. Purists be- drawn from the alternate paradigm.
lieve that quantitative and qualitative A classic example of this approach is
approaches are derived from different, mu- a quantitative study with a small quali-
tually exclusive understandings and be- tative interview component in the
liefs about the nature of research, and data collection phase. Gillis (1993) con-
hence, the paradigms and methods should ducted a triangulated study following
not be mixed. Situationalists believe that this classic approach. Determinants of
health-promoting lifestyles were ex- mixed methods may contribute to new di-
plored in 184 adolescent girls and their mensions, contradictions, and fresh per-
mothers and fathers using self-adminis- spectives that may lead to new research
tered questionnaires to measure the questions and signicant areas of inquiry.
study variables. Triangulation was ac-
(iv) Challenges of Triangulation
complished by analysis of data from
Methodological triangulation presents
semi-structured interviews with a sub-
unique challenges to researchers. Con-
set of eight adolescent participants.
nelly, Bott, Goffart, and Taunton (1997)
Another example is qualitative observa-
outline the major problems created by this
tions with a limited number of infor-
approach. Combining quantitative and
mants followed by a quantitative survey
qualitative methods is resource intensive
of a sample from a population. This
in terms of investigator knowledge, skill,
approach presents a consistent para-
time, and nances. On the other hand, if
digm picture in the study and still gath-
two or more researchers with skill and ex-
ers limited information to probe in
pertise in each paradigm work together,
detail one aspect of the study. The dis-
the best of both worlds can emerge. The
advantage is that purists from both the
interpretation of inconsistent and contra-
quantitative and qualitative worldviews
dictory numerical and textual ndings
claim that central assumptions of each
must be considered. When this happens,
method do not match the data collec-
the researcher needs to explore possible
tion procedures.
explanations and reconcile differences in
The mixed-methodology design. Here
meaning. Furthermore, replication of com-
the quantitative and qualitative para-
bined design studies is difcult and time
digms are combined at many method-
consuming. This may limit the number of
ological steps in the design such as the
investigations that would enable valida-
purpose statement, literature review,
tion of known ndings. Finally, the investi-
and research questions. This approach
gator must decide how to weigh data
enjoys the advantages of both the
sources and determine whether each
quantitative and qualitative paradigms
method is equally sensitive.
but adds considerable complexity to
the design. Inductive and deductive
reasoning are required in the same 2. Descriptive versus Explanatory
study. A sophisticated knowledge of versus Predictive Research
both paradigms is required by the re-
searcher to link the diverse methods In addition to determining whether your
together appropriately in the mixed- study should be purely quantitative,
method approach. purely qualitative, or triangulated, you
must also consider the goal, or the main
(iii) Benets of Triangulation concern, of your work: specically, do
The main benets of methodological trian- you wish to describe, explain, or predict?
gulation are convergence on truth and a Descriptive research (also called ex-
deeper understanding of the phenomena ploratory research) emphasizes the accu-
under investigation. Additional benets in- rate portrayal of nursing phenomena. A
clude the emerging of overlapping and dif- study that is primarily descriptive is
ferent facets of a phenomenon, a develop- mainly concerned with the accurate
mental informing of the second method by description of some aspect of society. A
the rst method used sequentially, and in- researcher may wish to assess specic di-
creased credibility and validity of results mensions or characteristics of individu-
owing to corroboration of data. Finally, als, groups, situations, or events by sum-
32 GETTING STARTED
marizing the commonalties found in dis- cioeconomic levels more likely to initiate
crete observations. Descriptive studies smoking during adolescence?
state what is. For example, a nurse sci- By contrast, the primary goal of ex-
entist may wish to assess the support planatory research is to understand or to
among citizens for implementation of a na- explain relationships. They engage corre-
tional no-smoking policy in public areas. lational designs to study relationships
With the goal of gauging the general senti- between dimensions or characteristics of
ments in a society toward the policy, the individuals, groups, situations, or events.
researcher tries to describe as precisely They explain how the parts of a phenom-
as possible what proportion of the popu- enon are related to one another (Fawcett
lation supports the policy. and Downs, 1986). For example, why is it
Fundamentally, the descriptive study that young women who initiate smoking
is about what and how many of what. It behavior during their teen years are
is directed toward answering questions more likely to be from higher socioeco-
such as, What is this? Descriptive research nomic backgrounds than those who do
using the empirical method involves ob- not start smoking? Here the issue is
servation of a phenomenon in its natural twofold. First, what is the relationship be-
setting. It uses case studies, surveys, tween socioeconomic background and
grounded theory, ethnographies, and phe- initiation of smoking behavior? Second, if
nomenological approaches, all of which there is a relationship, why does it exist?
are explained in detail later in this text. Explanatory studies ask why questions.
Because the goal of descriptive re- Predictive research moves beyond ex-
search is to describe, researchers fre- planation to the prediction of precise rela-
quently draw a sample to make estima- tionships between dimensions or charac-
tions about a particular population. As teristics of a phenomenon or differences
used by researchers, the term population between groups. Predictive research typi-
refers to the collection of individuals, cally engages the empirical method of ex-
communities, or nations about which one perimentation. This involves the manipu-
wishes to make a general statement. To lation of some phenomenon to determine
save money and time, the researcher its effect on some aspect of another phe-
draws from the population a sample that nomenon. The area of health promotion
will be representative of the population as provides a rich source for predictive stud-
a whole. Although including the whole ies on the impact of such interventions on
population would prove to be more accu- health outcomes and longevity in various
rate (as in a census), the costs may be pro- populations. Through prediction, one
hibitive. Public opinion pollsters, market can estimate the likelihood of a particular
researchers, and census takers typically outcome in a given context (Chinn and
emphasize descriptive accuracy in their Kramer, 1991), although it does not neces-
research. All explanatory studies have de- sarily enable one to control the outcome.
scriptive dimensions, and some descrip- A key objective of nursing research is to
tive studies have explanatory dimensions. design studies that yield the necessary
To explore the differences between fe- knowledge to enable nurses to control or
male students who initiate smoking in manipulate the practice situation to pro-
their teen years and female students who duce the desired outcomes for clients.
do not smoke, a researcher would want
to describe the characteristics of the two 3. Pure versus Applied Research
sets of students. Are rural students more
likely to be nonsmokers in their teen Intent is yet another distinction to con-
years? Are young women from higher so- sider when choosing a type of research.
Nurse scientists who accumulate infor- Peers of high SES students have healthy
mation in order to further our under- nutritional practices and the students
standing of relationships or to formulate inuence one another in selecting food
or rene theory are engaging in pure re- choices.
search; those who are interested in g- Parents of high SES students have
uring out how to bring about specic greater expectations concerning the
changes in nursing practice, education, nutritional diet choices of their chil-
or administration are engaging in ap- dren.
plied research. Simply stated, whereas High SES students know that they have
pure researchers value knowledge for its the nancial resources to select nutri-
own sake, applied researchers wish to tional food choices and therefore plan
have an impact on some specic nursing on it.
problem. High SES students have been more ex-
posed to role models who select
healthy food choices and are more
a. Pure Research likely to model themselves on such in-
dividuals.
A nurse scientist engaged in pure research
tries to explain the patterns of human re- We can identify three important goals
sponses to health and illness states across of pure research:
the life cycle. However, it is usually possi-
ble to devise many different explanations To test existing theories of human be-
to account for any particular response. A havior
challenge of nursing science is to deter- To explain observed patterns of be-
mine which, if any, of the possible expla- havior
nations accounts for any given pattern. To document our knowledge of the
Patterns may be understood through a va- emergence, modication, and persis-
riety of qualitative and quantitative tech- tence of patterned human behavior
niques. But whatever the approach, the ul-
timate goal is to offer better descriptions
and better explanations of human re- b. Applied Research
sponses to health and illness states.
Frequently nurse scientists are con- Again, applied research focuses on vari-
fronted with interesting ndings that ables that can be changed by interven-
cry out for an explanation. Suppose, for tion to achieve desired goals. For exam-
example, that a project is being done that ple, an applied researcher may want to
measures high school students socio- increase attendance at weekly prenatal
economic status (SES) and their intentions classes for adolescent mothers, and thus
to follow a healthy diet. And suppose, dur- would test the effectiveness of an inter-
ing data analysis, that a relatively strong vention such as phoning class partici-
relationship emerges indicating that the pants the night before the class meets to
higher an individuals SES, the greater the remind them to attend. Or, the researcher
likelihood that the individual will consume may want to increase peoples use of seat
a healthy diet. At this stage, a pure re- belts and therefore studies the effect of
searcher may wonder what explains the watching a crash simulation with unre-
pattern that has emerged. In this example, strained drivers. Applied research in-
pure research means doing research to cludes feminist, action, evaluation, and
identify which of the following explains health promotion research. These are dis-
variation in healthy diets: cussed in Chapter 8.
34 GETTING STARTED
4. Units of Analysis: Individuals is easy to compute average scores, but it

versus Aggregations is unlikely that we will have sufcient
cases to produce meaningful averages for
Nurse scientists study communities, all of the communities. For example, if
groups, institutions, and individuals. you studied 300 people from 25 different
Moreover, an institution such as the fam- communities and then wanted to move
ily may be studied either across cultures from the individual to the community
or within a culture. In doing research, it is level, you would have only 12 people
difcult to deal simultaneously with more from each community on which to base
than one level of analysis. At the outset of the average for the community.
a project, we should ask ourselves, Am I It is important, therefore, when you
studying individuals or aggregations? start designing a study to be absolutely
clear about the units of analysis that you
plan to use. Again, Table 1.3 summarizes
a. The Individual as the Unit of the relationships between the many fac-
Analysis tors discussed in this section.
If we are studying individuals, then we
should pose questions that concern indi-
vidual properties only. All of the data
G. RESEARCH AS A
collected should measure variations be-
PROFESSIONAL MANDATE
tween individuals on a variety of subjects.
Because knowledge is the foundation of
And any analyses of the data will have in-
professional practice, all nurses can and
dividuals as the basic unit. Most surveys
should develop and apply research skills
and experiments use the individual as the
in practice. This means that, although not
unit of analysis, although it is possible to
all nurses conduct research, every nurse
have individuals report on data for other
should become an intelligent consumer
levels of analysis such as communities,
of research, at the very least. This re-
companies, or groups. Furthermore, indi-
quires an understanding of the research
viduals may be counted and used to pro-
process and the ability to critically read
duce a measure for some aggregation: the
and evaluate research studies, determine
proportion of college graduates in a com-
their credibility, and assess the appropri-
munity is an example of such a variable.
ateness of research ndings for use in
practice. Prudent evaluation and use of
b. An Aggregation as the Unit ndings are critical to the development of
of Analysis a scientic knowledge base for practice.
Now that you understand the impor-
Alternatively, if we are studying aggrega- tance of nursing research to the provi-
tions, then we should pose questions that sion of quality care and to the develop-
concern properties of groups, institu- ment of nursing as a scientic discipline,
tions, communities, or nations. Individ- what role will you play in the research
ual characteristics, such as income, edu- process? This is an important question
cational level, or ethnic background have for you to reect on as you achieve full
to be expressed as averages or propor- professional status as a registered nurse.
tions. If we begin a study using an indi- As a professional nurse, the role you will
vidual level of analysis, then it is difcult, play in research will be inuenced by
without careful advance planning, to your level of educational preparation.
switch to a different level, such as the The trend is toward an expanded re-
community level. By using computers, it searcher role with advanced education
and expertise. However, as we have tice and usually an in-depth perspec-

stated, nurses at all levels of educational tive on nursing problems specic to
preparation have roles and responsibili- various practice settings. This enables
ties to assume when it comes to nursing these nurses to identify clinically rele-
research. The following list delineates vant problems and provide expert
the research mandate appropriate to var- clinical advise on ways that services
ious educational backgrounds. should be delivered and projects con-
ducted. Graduate education enables
Associate degree education. Gradu-
these nurses to assist others in the use
ates of associate degree programs
of research ndings. According to the
have several important roles to play in
ANA (1989), nurses with masters de-
the world of nursing research. They
grees are clinical experts who collabo-
need to develop an appreciation of the
rate with experienced researchers in
signicance of research to nursing
proposal development, data collec-
practice. They can assist with the iden-
tion, analyses, and interpretation.
tication of clinical practice research
problems, data collection, and appro-
Doctoral education. The main goal of
nurses at the doctoral and postdoc-
priate use of research ndings in prac-
toral level is the development of re-
tice in consultation with professional
search and scholarly ability. Doctoral
nurses (ANA, 1989).
programs focus on advanced research
Baccalaureate education. Graduates
training, with emphasis on the process
of baccalaureate programs have an im-
of building theory and advancing nurs-
portant role to play as consumers of
ing knowledge. Graduates of doctoral
research. This assumes they have an
programs are best prepared to ap-
intimate understanding of the re-
praise, design, and conduct research
search process; an ability to critically
(LoBiondo-Wood and Haber, 1998). Ac-
read, analyze, and critique research
cording to the ANA (1989), they de-
studies; and the judgment to interpret
velop nursing knowledge through orig-
and evaluate research ndings and de-
inal research and theory development.
termine their appropriateness for use
They also conduct funded indepen-
in the practice setting. Baccalaureate
dent research projects.
graduates may be active members of
research teams, participating in vari- The purpose of nursing research is to
ous phases of a study under the direc- generate a scientic knowledge base for
tion of experienced researchers. In practice. If the eld of nursing is to be
this capacity, these nurses may help accountable to the clients it serves, it must
generate clinical research questions to develop an empirically grounded body of
be investigated, participate in the col- scientic knowledge on which to base clin-
lection and recording of data, adminis- ical decisions and actions. The scientic
ter research protocols to clients, and approach as reected in research provides
always act ethically to protect clients a means to do this. In todays world of
rights. health-care restructuring and reform,
Masters education. Graduates of mas- nurses are increasingly required to be ac-
ters programs also play important countable for the quality of care they pro-
roles in nursing research. They have vide and demonstrate that they are engag-
a rich understanding of clinical prac- ing in evidence-based practice.
36 GETTING STARTED
E X E R C I S E S
1. Examine one issue of a nursing re- (e) Classify each by whether it falls
search journal. For each article do the most into a positivist, interpretive,
following: or critical perspective.
(a) Copy the title of each article.
2. Identify ve examples of research
(b) Classify each article by whether it that fall predominantly into each of
is predominantly quantitative or the positivist, interpretive, and critical
qualitative in orientation. perspectives. Did you encounter dif-
culties in coming up with examples for
each perspective? Do you think there
(c) Classify each by whether it is pre-
is a trend in nursing toward one of
dominantly descriptive, explana-
these perspectives? From your exami-
tory, or predictive.
nation of various journals and books,
estimate the percentage of research in
(d) Classify each by whether its intent nursing that would fall into each of the
seems to be pure or applied. perspectives.
RECOMMENDED READINGS
Chow, J.D. (1999). Interruption to research de- hensive introduction into the major meth-
sign: Substance driven research. Advances ods used in nursing research.
in Nursing Science, 22(2), 3947. The feminist Hilton, B., Thompson, R., and Moore-Dempsey,
and critical theory approaches guided this L. (2000). Evaluation of the AIDS Prevention
investigation of adolescents perceptions of Street Nurse Program: One step at a time.
health messages in magazines geared to- Canadian Journal of Nursing Research, 32(1),
ward teens. The article demonstrates how 1738. This is a clear example of new wave
interpretive inquiry (hermeneutics) trans- research using both qualitative and quanti-
formed the study to one that became sub- tative methods in the area of participatory
stantively driven and contributes to the research.
qualitative approaches in the human sci- Mitchell, M., and Jolley J. (1997). Research Design
ences. Explained (3rd ed). Orlando: Harcourt Brace
Crookes, P., and Davies, S. (1998). Research Jovanovitch College Publishers. This is one of
into Practice. Edinburgh, Scotland: Balliere the most approachable methods texts in psy-
Tindall. This is a comprehensive text with chology. It contains numerous examples and
an excellent discussion of essential skills helpful suggestions for students.
for reading and applying research in nurs- Streubert, H.J., and Carpenter, D. R. (1999).
ing and health care. Qualitative Research in Nursing (2nd ed).
Fain, J.A. (1999). Reading, Understanding, and Philadelphia: J.B. Lippincott. This clearly
Applying Nursing Research. Philadelphia: written text presents the essentials of qual-
F.A. Davis. This book provides a compre- itative research as it relates to nursing.
Chapter 2
Understanding Theory
CHAPTER OUTLINE
A. Explanations D. Theoretical Frameworks

1. Types of Explanations 1. Use of Nursing and Non-Nursing
2. The Nature of Nursing Science Frameworks
Explanations 2. Conceptual Maps
3. Flaws in Explanation 3. Examples of Frameworks Guiding
B. Theory Nursing Research
1. Purpose of Theory E. Testing and Using Theory in Research

2. Relationship of Theory, Research, and 1. Testing Middle-Range Theory
Practice 2. Testing Practice Theory
3. Classication of Theory 3. Structural Equation Modeling
4. Levels of Theory 4. Testing Alternative Explanations
C. Concepts and Statements: Building 5. Theoretical Triangulation
Blocks of Theory F. Points of Agreement on Research and
1. Concepts Theory
2. Propositions and Hypotheses
KEY TERMS
Alternative False dilemma Selected evidence

explanations Grand theories Situation-specific theory
Anecdotal evidence Hypotheses Source of spuriousness
Axiomatic derivation Illegitimate appeal to variable
Borrowed theories authority Structural equation
Conceptual framework Independent variable modeling
Conceptual map Intervening variable Theory
Concepts Middle-range theories Theoretical framework
Conditional variable Missing evidence Theoretical triangulation
Confounding variable Objective Treatment variables
Constructs Practice theories Unwarranted conclusion
Controlled observations Precise communication Variables
Control variable Propositions Verifiable
Dependent variable Provincialism
Empathetic explanations Replacement of terms
37
38 GETTING STARTED
A
s nurses, we want to further our un- wisdom, and truisms. Indeed, it is some-
derstanding of human responses to times difcult to nd an outcome that is not
health and illness in a constantly covered by some cultural maxim.
changing environment. But we lack well- There are problems, however, with
conrmed and widely accepted theoreti- nonscientic explanations. First, numer-
cal explanations for client behavior. For ous pairs of contradictory explanations
this reason, the work of nurse researchers can be found: the rst explains one re-
is vital. In their quest for understanding, sult, and a parallel one explains the oppo-
they may challenge commonly accepted site outcome. No matter what happens, a
explanations for human behavior, gather ready explanation is available. This poses
new evidence to support existing theories, a tremendous threat to the advancement
or generate new theories to guide nursing of nursing knowledge because scientic in-
practice. quiry into client responses is snuffed out
This chapter introduces beginning re- with pat answers: Of course he wont
searchers to the role of theory in under- practice his breathing exercises! You cant
standing the world of nursing. The basic teach an old dog new tricks! Box 2.1 lists
elements of theory are described and a few contradictory proverbs to illustrate
the purpose of theoretical frameworks in the prevalence of conicting explanations
research is explored. The chapter also in our world. Can you think of other con-
summarizes three theoretical frameworks icting ones?
commonly used to guide nursing re- Second, it is often true that the most
search. Finally, we present some simple compelling predictions that the nurse
theory-testing techniques that may be scientist can make will be those that
applied by beginning researchers to help run counter to common sense. If you
them understand the elusive patterns are able to make a prediction that goes
found in the nursing world. We begin by startlingly against common understand-
examining the role of explanation in nursing and it turns out to be true, you will
ing knowledge and nursing research. have a powerfully convincing study.
A. EXPLANATION 1. Types of Explanations
We can ask whether there are different

We probably could not survive in nursing
kinds of explanations that are more or
without being able to anticipate, at least to
less useful in understanding observations
some extent, the behavior of our col-
of the nursing world. In order to explain
leagues and clients. For example, in greet-
the occurrence and meaning of nursing
ing your nursing colleague, Nancy, you an-
phenomena, nurse researchers may use
ticipate that you will get an upbeat,
the following categories of explanation.
positive response rather than a hostile
one. You have generalized from your
previous experiences with her when you a. Deductive Explanation
conclude that you will be greeted warmly.
We rely on such generalizations to guide When nurse scientists use deductive ex-
much of our interaction. In addition, all planations, they try to show that the phe-
cultures have phrases that attempt to nomenon to be explained is a logically
offer general explanations of human necessary consequence of the explana-
behavior. These are variously referred tory premises. For example: if A B and
to as adages, old wives tales, maxims, B C, then A C. As a variant on Emile
proverbs, sayings, aphorisms, clichs, folk Durkheims (1897) analysis of suicide, we
UNDERSTANDING THEORY 39
BOX 2.1 A Few Contradictory Sayings
Absence makes the heart grow fonder./Out of sight, out of mind.

Beauty is as good as ready money./ Beauty buys no beef.
Nothing ventured nothing gained./Better safe than sorry.
A farthing saved is twice earned./A penny saved is a penny earned.
Birds of a feather flock together./Opposites attract.
Judge not a book by its cover./Fine feathers make fine birds.
He will shoot higher who shoots at the sun than he who aims at a tree./Hew not too high lest the
chips fall in thine eye.
Two heads are better than one./Too many cooks spoil the broth.
You cant teach an old dog new tricks./Its never too late to learn.
Haste makes waste./A stitch in time saves nine.
Better to have loved and lost than never to have loved at all./Better safe than sorry
might propose the following set of inter- reasoning, we get a testable hypothesis
related propositions: (iii). If, indeed, the data are consistent
(i) In any social grouping, the suicide with the prediction, this would constitute
rate (SR) varies inversely with the degree one piece of evidence that is supportive
of social support (SS). of the theory of suicidenamely, that
social support explains the connection
<SS 0 SR (the less the social support, between ruralness and suicide rates.
the higher the suicide rate) We could not claim to have proven the
theory to be correct because alternative
(ii) The degree of social support (SS) theories may also make the prediction
varies with the degree of ruralness that suicide rates will be comparatively
(RURAL) low in rural communities. In this illustra-
tion, note that social support is a com-
>RURAL 0 >S (the more rural the mon term that links size of community
community, the greater the social support) (RURAL) to suicide rate. The deduc-
tively derived hypothesis is easily tested
(iii) If the previous two statements are by examining the relationship between
correct, we can deduce that: suicide rate and community size.
>RURAL 0 <SR (the more rural the
community, the lower the suicide rate) b. Probabilistic Explanation
In the below example, we have a deduc- Probabilistic explanation rests on linking

tive model of the form: a particular case to its general category.
When asked to explain why Mary, a nurs-
>RURAL 0 >Social support ing student, spends excessive hours in
the library, if you respond by saying,
<Social support0 >Suicide rate
There are many nursing students in the
<RURAL 0 >Suicide rate library these days doing reserve read-
ings, you have given a probabilistic ex-
In this illustration, by combining a the- planation. Mary spends a lot of time in the
oretical proposition (i) with an untested library because she is a member of a cat-
assumption (ii) and then using deductive egory of individuals who are required
40 GETTING STARTED
to do reserve readings. The explanation nurses and clients share intimate experi-
suggests that an individual tends to be ences such as birth, death, diagnosis of
like others in the same category. terminal illness, and so on, empathetic
Another example is the explanation explanations take on increasing impor-
commonly offered in response to com- tance. Qualitative nurse researchers pay
plaints that, Ive been buying lottery tick- much attention to empathetic explana-
ets for years, but I never win the big prize. tions because, to nurse scientists work-
The probabilistic explanation is that the ing in this tradition, they are the key to
chances of winning the jackpot are about adequate explanations. An explanation or
one in 14 million. In other words, the odds a description of a behavior that rings true
are against any one individual. to both the researcher and the client
whose behavior is being revealed is key
to judging the adequacy of an empathetic
c. Functional Explanation
explanation.
Functional explanations explain the pres-
ence of some phenomenon in terms of the
role it plays in maintaining some system. 2. The Nature of Nursing Science
For example, if the presence of the family Explanations
unit in all cultures is explained in terms of
its role in producing new members for Nursing science explanations serve the
each society, this would be a functional same function as proverbs, adages, and
explanation: the family exists to propa- anecdotes: in all cases, they attempt to
gate to maintain the social system. account for human behavior or to explain
human responses to various health and
illness states. Does a generalization made
d. Causal Explanation by a nurse scientist substantially differ
from one made by a nonscientist? To be-
Causal explanations dominate the think- gin to come to grips with this question,
ing of many researchers. In a causal ex- lets examine quantitative and qualitative
planation, an event is explained by making evidence and argument.
reference to preceding inuencing events.
The explanation traces the sequence of
steps, each inuencing the next, that has a. Quantitative Evidence
led to some event. Think of a line of domi-
noes falling in sequence after one has For the quantitative nurse scientist, an
been knocked over, or consider all the important goal is to arrive at general state-
factors that may inuence choosing a ments, statements that can be applied to
healthful lifestyle. a variety of situations. A generalization
depicts the typical. The following state-
ments constitute generalizations:
e. Empathetic Explanation
Among high school students, females
Empathetic explanations are those that experience more mental health prob-
stress the experience of coming to see or lems than males.
to understand an imagined possibility. Males are more likely to be involved in
The individual feels, Ah-ha, Ive got itI motor vehicle accidents than females.
can see that I would do the same thing in Students who sit in the front or middle
similar circumstances. Everyday life is of the classroom get higher grades than
full of empathetic explanations and un- those who sit in the back or to the
derstandings. In the nursing world, where sides of the classroom.
In contrast to nonscientists, the evi- programs of study); and, nally, the re-
dence a quantitative nurse scientist would searcher should indicate the procedures
look for would be based on objective, ver- used to measure the variables and to
iable, controlled observations and would process the resulting data.
be communicated in a precise manner
(Browne and Keeley, 1990). By objective,
we mean observations that are free from b. Qualitative Evidence
bias; by veriable, we mean information
For qualitative nurse scientists, more em-
that could be conrmed by tests con-
phasis is placed on the extent to which
ducted by others; by controlled observa-
explanations and descriptions ring true
tions, we mean those in which other con-
to both the researcher and to the people
founding factors are minimized or taken
who are being described. The evidence it-
into account; and by precise communi-
self may be based on a painstaking analy-
cation, we mean that the information is
sis of documents, on in-depth or focus
unambiguous. In nursing science, many
group interviews, or on lengthy partici-
of our generalizations are probabilistic;
pation in a group (see Chapter 7). But in
this means that in most cases (but not
all cases, the evidence itself and the in-
all), the statement will be true. Certainly
terpretations placed on it are judged in
not all high school girls have more men-
terms of how well they are perceived to
tal health problems than their male coun-
deal with the matter at issue.
terparts; and certainly not all students in
the back row get lower grades than those
sitting in the middle of the room. Because
nurse scientists arguments must be based 3. Flaws in Explanations
on evidence, information must be col-
Nurse scientists design studies, make ob-
lected and then appropriately analyzed.
servations and analyze them, and then
The resulting conclusion is then an
prepare a report on the results. During
informed opinionthat is, an opinion
each stage, one must be careful not to fall
based on evidence that has been col-
prey to errors in thinking or in argument.
lected under controlled circumstances.
The following section highlights common
In making probabilistic generalizations,
errors of particular relevance to nurse
the nurse scientist has to indicate pre-
scientists.
cisely how the information was collected,
what kind of sampling procedures were
used, and how the data were analyzed. a. Illegitimate Appeal to Authority
A probabilistic generalization (e.g., stu-
dents who choose nursing as a career are As researchers, sometimes we may be
more likely to have altruistic vs. material- tempted to argue that something is bad
istic values) should indicate the popula- (e.g., euthanasia) by using an illegitimate
tion about whom the generalization is be- appeal to an authority. For example, in
ing made (e.g., all nursing students in the this case, we could state that euthanasia
world; students in a particular country, is bad because Nobel Prize winner in
region, community, or in the nursing pro- physics, Dr. X., claims that it is. In this
gram at the University of Texas). The re- case, the status of the person expressing
searcher should also indicate the size the view should not inuence our judg-
and the method of selecting the sample ment. Legitimate authority, in contrast, is
on which the generalization is based (e.g., to be found in the peer-reviewed and
a convenience sample of 125 students in scholarly literature of nursing and related
nursing and 125 students in non-nursing disciplines. Does the preponderance of
42 GETTING STARTED
evidence in the research literature point d. Missing Evidence

to one conclusion? Are the exceptional
ndings to be explained by quirks in When the reports of researchers are ex-
methodology or in the samples studied? amined, read carefully to make a judgment
In any case, nursing science proceeds by as to the adequacy of the evidence for
testing and retesting old ideas under new each point made. Although it would be un-
conditions. realistic to expect all assumptions to be
identied and all statements to be demon-
strated, watch out for signs of important
b. Provincialism missing evidence and be certain that evi-
dence for key arguments is provided.
As researchers, we carry with us the bag- Frequently a nurse scientist is con-
gage of our culture. This makes us vul- fronted with interesting ndings for
nerable to the danger of provincialism, of which an explanation should be offered.
tending to see things as our culture sees Suppose, for example, that a project is
them. Thus, if it is politically correct to being done that measures high school
view men and women as equally skilled at students scores in a health promotion
child rearing, then this may blind us from course focusing on health behavior deci-
looking at areas in child rearing in which sion making and relates this variable
this may not be true. Kahane argues pro- to the age of their rst experiencing sex-
vincialism tends to make us concentrate ual intercourse. And suppose, during
on our own societyto the exclusion of data analysis, that a strong relationship
other culturesand via loyalty, to inu- emerges indicating that the higher ones
ence our acceptance or rejection of al- score in the course, the older the average
leged facts or theories, whatever the na- age of rst sexual intercourse. At this
ture of the evidence (1988, p. 47). stage, the researcher may wonder what
explains the pattern that has emerged.
c. False Dilemma Possibilities such as the following might
come to mind:
A false dilemma is set up when a re-
searcher argues that something is caused
Peers of students with high course
scores avoid early initiation into sex
by either A or by B. Then, having provided
and inuence their friends to do like-
some evidence that B is not responsible,
wise.
the researcher falsely concludes that A
must be the cause. The problem here is
Among students with high course
scores, parental supervision is such
that there may be several other possibili-
that the opportunities to engage in pre-
ties that have not been considered. For
marital sex are more limited.
example, one may argue that either diet
or a sedentary lifestyle explains the inci-
Students with high course scores have
higher educational aspirations and
dence of osteoporosis in women. Now
avoid early initiation into sex because
suppose that the data from a small sam-
they think that to do so would impede
ple of women indicated that sedentary
their chances of going to college.
lifestyle was not related to osteoporosis
in these women. Because other possible Unfortunately, the study may not have
explanations have been omitted, this nd- been designed to test, and possibly to
ing paves the way for an inappropriate rule out, competing explanations. There-
conclusion that diet explains osteoporo- fore, there may be no empirical evidence
sis in women. for the conclusion presented by the re-
searcher. The problem here is that al- of convenience (clients who were present
though much data may be presented doc- at the oncology clinic the day when the
umenting a variety of relationships, the questionnaire was administered)is not
reader of such a report may be convinced, only small but because the clients were
inappropriately, that the researchers nu- not selected using a strict sampling pro-
merous data, tables, and gures support cedure, it cannot be taken to represent
the particular explanation being offered. even the oncology clients at that clinic.
Such explanations may sound good and Avoid conclusions that state generaliza-
appear reasonable, but they lack evidence. tions that go beyond those justied by
the sampling procedure. (Details on sam-
pling procedures are discussed in Chap-
e. Anecdotal Evidence ter 15.)
Suppose an incident of a person on a day

pass from a mental health facility who f. Selected or Suppressed Evidence
commits an assault is cited in an attempt
to justify the conclusion that the mental In citing evidence, researchers should at-
health care system is under-resourced. tempt to fairly represent the ndings in
What do you conclude? Although this the research literature and their own data.
type of anecdotal evidence is commonly We are all familiar with public debates
used, it is hardly sufcient to warrant the that have gone on in recent years on such
conclusion drawn. What are the overall issues as environmental illnesses, the dan-
rates of assault being committed by such gers of smoking, abortion, and mercy
persons? What are the social and human killing. In each case, expert researchers
costs of more restrictive day passes? Are who claim to have special knowledge on
denials of day passes to mental health these matters and claim also to represent
clients more likely to lead to rehabilita- the consensus in their discipline have
tion, lead to decreases in assault rates, or been called on to testify. The trouble is
provide greater protection, in the long that each side seems to be able to pro-
run, to society? Be wary of anecdotal evi- duce an expert who supports its point of
dence. view. These experts are probably pre-
Anecdotal evidence is the laypersons senting selected evidence, choosing to
equivalent of arguments based on unrep- report only studies that support a partic-
resentative samples. When nurse scien- ular point of view, and ignoring the evi-
tists attempt to describe something (e.g., dence that runs counter to what they are
the effectiveness of a specic nursing in- attempting to demonstrate. Advocacy is
tervention with breast cancer clients), not science: both are legitimate enter-
they typically do so by working with sam- prises, but they are not the same thing.
ples. The conclusions should not go be- A similar situation exists with respect
yond the adequacy of the sample. A very to researchers who report their own data.
small sample (e.g., an anecdote involving Given the need to keep reports short,
just one case) should not be used to gen- some information is not reported; there-
eralize about all of humankind. You can fore, it is possible that researchers do
imagine how inappropriate it would be to not represent their own data fairly. Some
generalize about oncology clients atti- information may be considered unre-
tudes toward an alternative pain control portable or insufciently interesting to
procedure based on a sample of 10 breast merit inclusion. Readers cannot tell how
cancer clients. Such a samplelikely one much information has been omitted.
44 GETTING STARTED
g. Unwarranted Conclusions BOX 2.2 Definitions of Theory

Sometimes in analyzing data, a researcher Theory is a statement that accounts for or
will come to an unwarranted conclusion. characterizes some phenomena (Stevens,
Confusing correlation and cause is an ex- 1984).
ample of this type of error in reasoning. Theory is a general statement that ex-
Researchers take pains to measure vari- plains the interrelationships among ob-
ables and to investigate how they are reserved facts and propositions (Seaman,
lated to one another. However, to show 1987).
that as A increases, so does B does not Theory consists of a set of concepts and
demonstrate that A causes B. Although propositions stating the relations between
them. The set of propositions must be in-
this issue is explored in greater detail
terrelated so that one can derive new
in Chapter 17, let us look at one example
propositions by combining them deduc-
here. Some theorists argue that there tively, and furthermore, the propositions
is a connection between engagement in must be amenable to empirical testing
health behaviors and health value. Let us (Homans, 1964).
assume that such a correlation, in fact, Theory is defined as a set of interrelated
exists (some would disagree): are we jus- concepts, definitions, and propositions that
tied in concluding that the valuing of present a systematic view of phenomena
health led to the engagement in health be- for the purpose of explaining and predict-
haviors? What are the possibilities when ing phenomena (Chinn & Kramer, 1995;
we have a correlation between any two LoBionda-Wood, 1998).
variables?
Health Value has inuenced

engaging in Health Behaviors HVHB
Engaging in Health Behaviors

inuences Health Value HBHV lar denitions of theory in the research
literature. The term theory has different
They inuenced each other HBHV uses in the world of health care. Nurses
Other factor(s) inuenced both of them ? use the term in everyday conversation to
refer to a hunch or idea about some phe-

HV HB
nomena or event that is related. Scien-
In drawing conclusions, it is important tists use the term in a precise manner .
to sufciently explore alternate possibili- This text views theory as a systematic vi-
ties. Be careful not to confuse association sion of reality that describes, explains, or
with cause. predicts something. It is derived from the
Greek word theoria, which means vision.
Theories consist of words or symbols
that represent something in the real
B. THEORY world. Theories are not facts but simply
ways of perceiving aspects of reality.
The term theory, similar to the term re- Theories are dynamic and change with
search, has many different meanings in subsequent testing as scientists examine
nursing literature. The rst step in un- theory through research. Similar to the-
derstanding theory is to be clear on its ory in all sciences, nursing theory
meaning. This can prevent much confu- changes and is modied as new knowl-
sion and frustration on the part of begin- edge becomes available through re-
ning researchers. Box 2.2 provides popu- search.
1. Purpose of Theory placed within the context of theory, the

ndings have greater potential for con-
In order to survive in society, all of us tributing to knowledge development in
need a shared understanding of human the discipline than isolated research
behavior. We learn what a frown means, that is not linked to a particular theory.
what a jaunty wave of the hand means, This is because theories are abstract
and we come to a common understand- constructions of the mind and hence
ing of what a vast array of words and sym- transcend the limits of specic circum-
bols mean in our culture. We learn about stances such as time and place that ap-
appropriate and inappropriate behavior. ply in studies of a particular problem.
As children, we might learn, for example, A guide in selecting the appropriate
that it is okay to swear in front of our method to answer the research ques-
same-gender peers but that it is not action and in interpreting the results
ceptable to use the same words in front of A source of professional autonomy.
our parents or in mixed company. We Theory guides the nurses thinking and
speculate on why others behave as they actions and provides a basis to de-
do and we think about our own behavior. scribe, explain, or predict factors that
Theory provides a structure for the inter- inuence nursing care. It enables nurses
pretation of individuals behavior and of to explain nursing actions and their
events. Many theories are required to ac- outcomes.
count for the multitude of experiences
encountered by human beings. They state What do we gain by connecting re-
what something is, how something hap- search to new or existing theoretical per-
pens, or why it happens. Their function, spectives? The answer has to do with
therefore, is to describe, explain, or pre- power. Research that limits itself to the
dict specic aspects of reality. particular or to the unique will not con-
Theory serves a multitude of functions tribute to our general understanding of
in nursing, but its primary goal is to pro- the human condition. General explana-
vide a framework that links research and tions allow research to have implications
practice and contributes to making sci- for our overall understanding of human
entic ndings meaningful and generaliz- responses, as well as increase our knowl-
able. Theory also functions to provide: edge of the particular variables involved
in a nursing research project.
A structure for interpreting observed Unfortunately, in many journal articles,
behaviors, situations, and events the research problem is not linked to a
A means of imposing order on un- theory or conceptual framework. Silva
ordered experiences (1986) suggests this may be caused in
An efcient means for summarizing and part by a lack of commitment to theory
explaining observations from a num- testing by the discipline. Silva conducted
ber of isolated research studies. A the- a study to identify the nursing theories
ory enables a researcher to contribute that were tested in nursing research. She
the ndings of even a small study to found 62 articles that reported using a
the larger theoretical perspective that nursing theory. Of the 62 articles, only
uses the same theory. nine actually tested a theoretical propo-
A source to generate hypotheses that sition. Deets (1990) reviewed 300 articles
predict what will be found when data published in a 12-month period to iden-
are collected and analyzed tify studies that tested one or more of the
A framework for guiding research. eld of nursings metaparadigm con-
When a problem or research question is cepts (nursing, person, health, and envi-
46 GETTING STARTED
ronment). Only six articles addressed you to collect data about relationships
one or more of the central concepts. Betz (with propositions or hypotheses) de-
and Beal (1996) conducted a systematic rived from the theory. The ndings of the
analysis of 302 published pediatric nurs- research can be woven into a program of
ing studies to evaluate use of nursing the- research to build theory that will direct
ories. Their review found that only 17 in- subsequent theory development. When
vestigations used nursing theories as the theory has been tested and validated
frameworks. If one accepts the assump- through research and subsequently used
tion that one of the purposes of research to direct nursing practice, the cyclical
is to test and generate theory and thus relationship of theoryresearchpractice
advance the science of nursing, then one is complete. Research that generates
must question doing nursing research theory requires the researcher to iden-
that does not adequately address the the- tify and describe nursing phenomena
oretical elements in a study. observed in practice and articulate the
relationships between the phenomena.
The existing knowledge about a nursing
2. Relationship of Research, phenomenon guides the researcher in
Theory, and Practice asking the appropriate question for a
given study and guides him or her in se-
Theory, research, and nursing practice lecting the design, method, and analysis
are intimately related processes. As pre- to answer the research question.
viously stated, the function of theory is to
describe, explain, or predict the occur-
rence of nursing phenomenon. The func- 3. Classification of Theory
tion of research is to generate or test a
theory (Fawcett, 1999; Chinn and Kramer, The authors of this text believe there is
1995) that will inuence practice either some advantage to viewing theory along
directly or indirectly. Hence, we see the a continuum of abstractness that reects
cyclical nature of the relationship. Re- the breadth of the concepts and relation-
search designed to test theory requires ship statements in the theory. Figure 2.1
Figure 2.1 Theory continuum.

illustrates the theory continuum that haviors than adolescents with fewer -
runs from the most abstract grand theo- nancial resources. Use of grand theories
ries to middle-range theories to practice- in this way to derive statements to guide
level theories. The major theories re- research is important in moving discus-
ported in the nursing literature can be sion of grand theory (conceptual models)
classied along this continuum. down the continuum of abstractness to a
more specic level. The deductive method
a. Grand Theories is particularly useful in developing middle-
range theory in substantive areas that are
Grand theories (or conceptual models) are consistent with a given grand theory.
the most abstract type of theories. They
embody the beliefs, traditions, goals, and
values of the discipline. They take into ac- b. Middle-Range Theories
count the phenomena of central concern
Middle-range theories are located on the
to nursing: person, health, nursing, and
continuum midway between the most
environment. They broadly dene these
abstract ideas and the most concrete.
concepts and link them together through
They deal with limited aspects of nursing
relationship statements reective of the
phenomena and are more testable and
theorists view of the discipline. They ex-
generalizable than grand theories. They
plain universal relationships that describe
contain well-dened concepts and propo-
what the discipline of nursing is all about.
sitions. Common examples include the
They remain broad, abstract representa-
health promotion theory (Pender), stress
tions of reality. Although grand theories
and coping theory, and theories of social
are not directly tested through research,
support. Lenz and associates (1995, 1997)
they serve as frameworks to guide the se-
suggest that nursing as a discipline must
lection of concepts and phenomena im-
move from focusing on the current grand
portant to nursing concerns. In this in-
theories to middle-range theories if con-
stance, grand theories highlight the
tinued development is to occur. The lack
importance of the phenomena for investi-
of theory-testing studies in the nursing
gation (Fitzpatrick and Whall, 1996). Exam-
literature, discussed earlier, can be used
ples include Roys adaptation model and
as evidence to support the position of
Rogers science of unitary man model.
Lenz and associates.
Researchers may examine relation-
ships that are derived from grand theo-
ries. For example, most people are famil- c. Practice Theories
iar with the simple deductive statement
that if AB and BC, then AC. In a sim- Practice theories are narrower in scope
ilar manner, one may derive a relational and more specic than middle-range the-
statement from a grand theory to be ories. Concepts are specically and nar-
tested in a research study. For example, rowly dened and readily measured. Prac-
in her Self-Care Model, Orem (1995) pos- tice theory propositions produce clear
tulates that basic conditioning factors directives for application in practice.
(i.e., age, health state, available resources, Much of what nurses follow today in hos-
patterns of living) inuence ones self-care pital procedure manuals is actually prac-
agency (i.e., ability to engage in behaviors tice theory that has been developed in-
to keep ones self healthy). A researcher ductively by expert nurses who have
may deduce from this relationship that observed and described ways of practic-
adolescents with adequate nancial re- ing based on trial and error. Common ex-
sources will engage in more health be- amples include theory on mouth care, tak-
48 GETTING STARTED
ing patients oral temperature, decubitus what the theory does. This work has been
ulcer therapy, and use of restraints. Prac- instrumental in guiding the development
tice theory is designed for immediate ap- of theoretical research in nursing. The
plication to practice when appropriate. levels include:
A new level of theory known as situa-
tion-specic theory is emerging in nursing
Level 1, referred to as factor-isolating
theory. It is descriptive in nature, used
to incorporate diversities and complexi-
when little is known about the phe-
ties in nursing phenomena. It incorpo-
nomenon. It asks the question, What
rates sociopolitical, cultural, and historic
is this? Given the level of develop-
contexts of nursing encounters (Im and
ment of nursing as a scientic disci-
Meleis, 1999). According to these au-
pline, it is appropriate that many of
thors, situation-specic theory answers
our investigations are at this level.
a set of coherent questions about situa-
tions that are limited in scope and focus.
Level 2 is factor-relating theory. It re-
quires correlating factors in such a
They characterize this level of theory as
way that they create a meaningful situ-
one that focuses on specic clinical prac-
ation. It concerns the relationship
tice phenomena, reects a specic eld
among concepts and asks the ques-
of practice, is conned to developing and
tion, What is happening here?
understanding of a specic population,
and provides a blueprint for action. An
Level 3 is situation-relating theory. It ex-
plains and predicts how situations are
example of an emerging situation-specic
related. It focuses on the question,
theory is the work of Im and Meleis
What will happen if?
(in press) on menopausal transition of
Korean immigrant women. The situation-
Level 4 is situation-producing theory. It
is prescriptive theory and it answers
specic theory is developed from transi-
the question, How can I make X hap-
tions theory, a middle-range theory. Al-
pen? Level 4 is the most powerful in
though transition theory explains human
terms of being able to prescribe nurs-
beings during any kind of transition, the
ing activity. It moves beyond descrip-
situation-specic theory is narrowly fo-
tion, explanation, and prediction to
cused to explain the menopausal experi-
control.
ence of a specic Korean immigrant pop-
ulation. As such, the concepts in the Diverse views exist on what consti-
theory are grounded in Korean culture tutes the appropriate level of theory in
and the menopausal experience of immi- nursing. Given its complexity, we believe
grant women. all levels are needed: theories of varying
A criticism of some practice theory is scope and specicity are required to ac-
that it may be so narrow in scope that it count for the multitude of human re-
is not useful for the development of nurs- sponses to health and illness states.
ing theory beyond providing ideas for
further theory development. If the prac-
tice theory deals with only one situation
at one point in time, it is too narrow to be C. CONCEPTS AND
useful in many nursing situations. STATEMENTS: BUILDING
BLOCKS OF THEORY
4. Levels of Theory
Nursing students often struggle with intro-
Dickoff and associates (1968) produced a ductory chapters on research and theory
seminal piece of work that identied four because they do not understand the lan-
levels of theory classied according to guage used by scientists and researchers,
who share a unique vocabulary. It is im- the real world. Making a concept poten-
portant that students and nurses inter- tially observable facilitates testing of the
ested in reading and applying research idea. For love to be considered a con-
ndings in their practice learn this new lan- struct, the researcher would have to list
guage so they can approach the research all of the aspects of love that could be ob-
literature intelligently. The following terms served in the real world. Some authors
are used frequently by members of the re- distinguish construct from concept by
search community. stating that constructs are constructed
or invented by the researcher.
1. Concepts
Concepts are the building blocks of the- b. Variables

ory. They are abstract entities that repre-
Variables are concepts that are observ-
sent broad general ideas that are not
able, measurable, and have a dimension
directly observable in the real world.
that can vary. For example, temperature
However, they are formed from real-
is a variable that is observable, measur-
world observations of objects or events.
able, and varies from high to low.
For example, love is a concept. If asked to
Variables usually are so specic that
observe love in a population, most peo-
they capture only a portion of the general
ple would experience some confusion be-
meaning conveyed at the conceptual level.
cause the word can be dened in a num-
Concepts, constructs, and variables can
ber of ways. For research purposes, love
be viewed along a continuum of speci-
would need to be dened in observable
city, similar to theory. At one end, we
terms. Figure 2.2 illustrates the possible
have concepts that are general, vague,
meanings of the concept of love.
and abstract, and at the other end we
have variables that are narrow in mean-
a. Constructs ing, observable, and measurable. For ex-
ample, if the researcher takes the con-
Constructs are concepts specied in
cept of love and denes it as a variable,
such a way that they are observable in
the denition would be so explicit that it
would clearly identify how the variable
love is to be measured. Although the
reader may not agree with this denition
of the variable, the meaning and method
of measurement or observation would be
clearly understood. Figure 2.3 illustrates
how the concept of love might be de-
scribed, viewed as a construct, and de-
ned as a variable.
c. Types of Variables
In designing, implementing, and evaluat-

ing studies to test theory, researchers
distinguish various types of variables.
The following discussion relates to vari-
ables as used in research studies. It is in-
Figure 2.2 Specific meaning of concept.
cluded in this chapter on theory to help
50 GETTING STARTED
Figure 2.3 Conceptconstructvariable continuum.
students see the link between concepts variation depends on other variables. For
as the basic building blocks of theory and example, in an investigation of the factors
variables as the basic unit of scientic re- that inuence womens decisions to
search. choose home birthing rather than hospi-
tal birthing programs, birthing program
(i) Dependent Variables choice would be treated as the depen-
A dependent variable is a variable dent variable. However, this would not
thought to be inuenced by other vari- preclude exploring the possibility that, in
ables. It is the effect in a cause-and-effect turn, variations in birthing choice have
relationship. As its name suggests, its an impact on a variable such as recovery
rate. In such a case, recovery rate would several treatments used simultaneously,
be treated as the dependent variable. and their individual and joint effects
would be assessed.
(ii) Independent Variables
In general, we should be careful to en-
An independent variable is a cause in a
sure that the variables we treat as inde-
cause-and-effect relationship. It is a vari-
pendent are indeed different variables.
able that has been selected as a possible
For instance, if we measured the weight
inuence on variations in a dependent
of premature infants in pounds and then
variable. Typically, one nds a number of
repeated the measurements, but this
independent variables in a study. Once
time using kilograms, our two sets of
again, it is how the variable is treated
measures would lack independence. In
that is, how it is thought ofthat deter-
short, they would simply represent dif-
mines whether it is an independent or a
ferent measures of the same variable.
dependent variable, not the nature of the
The researcher must be careful not to fall
variable itself.
into the trap of thinking that there is a
In our example of birthing program
powerful causal connection between two
choices, many different factors may be
variables when the measures lack inde-
treated as independent variables. We
pendence and thus represent two differ-
might well include such factors as rural
ent measures of the same thing.
or urban home community, number of
previous deliveries, types of family sup-
(iii) Control Variables
port available, socioeconomic status,
A control variable is a variable that is
mothers participation in the labor force,
taken into account in exploring the rela-
and the presence of role models who
tion between an independent variable
have opted for home birthing programs.
and a dependent variable. There are three
In experimental designs, treatment
basic types of control variables: the in-
variables are those whose effect on some
tervening variable, the conditional vari-
dependent variable is being assessed. For
able, and the source of spuriousness (or
example, in an experiment to test the ef-
confounding) variable.
fectiveness of a structured exercise pro-
gram on the perceived level of well being Intervening variable. An intervening
in middle-aged women, two groups of variable (I) is a variable that links an
subjects would be assigned to either the independent variable (X) to a depen-
exercise program group or to the group dent one (Y). An intervening variable
that did their standard routine of activi- represents an explanation of how the
ties. In this study, exercise would be the independent variable inuences the
treatment variable and well being would dependent variable. It may be dia-
be the dependent variable. There may be grammed as:
X I Y
Suppose we are investigating the rela- with a high SES are more likely to be ex-
tion between socioeconomic status posed to other mothers in home birthing
(SES) and preference for home birthing programs. In other words, exposure to
programs. A possible explanation of how people in home birthing programs inter-
SES influences the type of birthing pro- venes to account for the program pre-
gram preferred would be that mothers ferred.
52 GETTING STARTED
Conditional variable. A conditional charge from the hospital: we might

variable (C) is a variable that accounts want to nd out whether that relation-
for a change in the relationship be- ship is fundamentally altered (or is en-
tween an independent variable (X) and tirely different) for each gender. Ac-
a dependent variable (Y) when the gen- cordingly, we might test men and
eral conditions change. Suppose we are women separately for a relationship be-
investigating the relationship between tween SES and attitudes. Here gender
SES and attitudes toward early dis- would be the conditional variable, as in:
Males
X Y
Females
X Y
Source of spuriousness variable. A tween socioeconomic background and

source of spuriousness variable (S/S) choice of a birthing program by pregnant
is a variable that is viewed as a possible women, we might consider the possibility
inuence on both the independent vari- that rural or urban background is a
able (X) and the dependent variable source of spuriousness. Here the idea is
(Y), in such a way that it accounts for that it may be the type of community that
the relationship between them. In other the women come from that inuences
words, the relationship between X and family socioeconomic achievement as
Y is spurious because it is produced by well as inuencing birthing program pref-
the inuence of S/S on both X and Y. erences. The relationship between SES
and birthing program choice might there-
If we were exploring the relation before be spurious. As in:
X Y
S/S
In experimental research, a source of rural the place of residence, the lower the
spuriousness variable is typically referred suicide rate. Hypotheses advance nurs-
to as a confounding variable. A con- ing knowledge by conrming or refuting
founding variable is one that may be inu- the theoretical propositions from which
encing the outcome of an experiment sys- they are derived.
tematically and that, after it is recognized,
is treated in the design and controlled.
D. THEORETICAL
2. Propositions and Hypotheses FRAMEWORKS
Propositions and hypotheses are ele- Theoretical frameworks are brief de-
ments of theories, and they represent dif- scriptions of a theory, or portions of a
ferent levels of abstraction. theory, to be tested in a research project.
A theoretical framework describes the
basic structure of ideas (i.e., theories,
a. Propositions concepts, propositions) within which the
study is to be conducted and the results
Propositions are statements of proposed interpreted (Diers, 1979). It is a verbal
relationships between two or more con- description of existing theory relevant
cepts in a theory. Propositions link con- to the phenomena under investigation,
cepts of a theory together so something based on a review of the literature.
can be described, explained, or pre- Nursing research studies that are
dicted. The concepts in a proposition are guided by a theoretical framework usu-
not always measurable, rendering the ally make this fact known by explicitly de-
proposition untestable. Propositions are scribing the framework early in the re-
also referred to as axioms or theorems. search report. Typically a section titled
An example of a proposition would be the Theoretical Framework briey describes
statement that in any social grouping, the major features of the theory or frame-
the greater the social support, the lower work so readers will be able to under-
the suicide rate. stand the theoretical basis for the study.
b. Hypotheses 1. Use of Nursing and Non-

Nursing Frameworks
Hypotheses are statements of predicted
relationship between two or more vari- A well-developed study should have an ex-
ables. Hypotheses are generally used in plicitly stated frameworkthat is, the hy-
the context of a research study. They potheses should emerge from a theoretical
form a bridge between the abstract world framework that is explicitly stated in the
of theory and the concrete reality of nurs- article. Unfortunately, the frameworks in
ings practice world. In this sense, they nursing studies are often implicitleaving
unify the two domains. Hypotheses may it to the reader to ponder the literature re-
be directional (specifying a direction for view to determine the relationships among
the relationship, e.g., the greater X, the variables from previous investigations.
greater Y), or nondirectional (simply stat- This may present a challenge for the be-
ing that there will be a difference between ginning research student.
groups or variables, e.g., as X varies, Y The theoretical framework for a nurs-
varies). In the suicide example, the hy- ing investigation depends, to a large ex-
pothesis derived might be the more tent, on the researchers creativity in
54 GETTING STARTED
formulating what the particular prob- witnessed for some time in the nursing
lem, propositions, or hypotheses are an literature.
example of in the nursing world. For ex- As we have seen, many nurse re-
ample, a pediatric problem related to ap- searchers who use a theoretical frame-
propriate preparation of toddlers for hos- work to guide their studies use one that
pital admission may be conceptualized has been previously developed and pub-
within the context of child development lished elsewhere. But occasionally, nurse
theory, social cognitive theory, or oper- researchers may link their research proj-
ant conditioning theory. However, it may ect to theory by synthesizing the ndings
also be thought about within the context from previous studies to create a new the-
of preparation for any new experience. ory. Unfortunately, some nursing studies
When viewed in this manner, theories on claim a theoretical linkage after the fact.
preparation and information giving, theo- Even though this may enhance the mean-
ries on teaching or learning, and stress ingfulness of the ndings, articially link-
and adaptation theories may become ing a problem to a theory is not a desirable
part of the theoretical framework for the way to build knowledge in a discipline.
study of hospital preparation of toddlers. When a theory is retrospectively linked
Nursing research studies are often ca- to a research problem, the researcher of-
pable of being conceptualized in a num- ten fails to consider nuances of the the-
ber of ways because of the complex na- ory in the study design. Ideally, a theo-
ture of nursing practice. Which theory to retical perspective is used to direct a
select as a framework is a choice made study; then, the design of the study, the
by the researcher. It is not a matter of measurement of key constructs, the
choosing a correct theory or an incor- analysis of data, and the interpretation of
rect theory to guide the investigation. ndings will be inuenced by that theory.
Rather, it is important to be able to In reading research articles published
clearly see the logical links between the in the nursing literature, you may en-
theoretical framework described and counter the term conceptual framework.
the research proposition or hypotheses Some authors make the distinction that
to be tested. the framework of a research study is
As a young science, nursing often deals called a conceptual framework when it
with phenomena that are not unique to does not contain a specic theory that ex-
nursing. As a consequence, a signicant plains the expected relationship between
number of theories used as frameworks in variables but rather synthesizes relevant
nursing research are based on theoretical literature about the proposed hypothe-
work from other disciplines, such as soci- ses. In most cases, the terms theoretical
ology (e.g., the sick role theory), psychol- and conceptual framework are used inter-
ogy (e.g., socialcognitive theory) or changeably.
physiology (e.g., theory of pain percep-
tion), to name a few. These are borrowed
theories that is, theories taken from 2. Conceptual Maps
other disciplines and applied to nursing
questions and research problems. Bor- A conceptual map is simply a diagram
rowing is generally viewed as acceptable of the concepts and relationships ex-
as long as the relevance of the theory pressed in a theoretical framework. A con-
to nursing is explained. Given the stage ceptual map is an efcient way to commu-
of development of nursing science, this nicate what is known about a phenomenon
trend toward multidisciplinary and multi- more clearly than a verbal description.
theoretical perspectives is likely to be Conceptual maps contain all the concepts
of a theory. Arrows indicate the proposed increasing popularity as a framework to

direction of the relationships among the guide research. Orems theory describes
concepts. Conceptual maps should be the practice of nursing from the perspec-
supported with references from the litera- tive of self-care. It focuses on the ability of
ture. Figures 2.4 through 2.6 are concep- an individual to take care of him- or herself
tual maps that illustrate the three theories (self-care agency). Self-care is the ability to
described next. initiate and perform activities on ones
own behalf in order to maintain life,
3. Examples of Frameworks health, and well-being. A special type of
Guiding Nursing Research self-care called therapeutic self-care is re-
quired when a person is ill or requires spe-
A number of nursing theories appear in the cial skills or knowledge to perform self-
literature today. Box 2.3 highlights some of care. If one is dependent in some way (e.g.,
the best-known theories used by nurse re- physically challenged, frail, or on com-
searchers to guide research investiga- plete bedrest), then family members or
tions. For a detailed description of these support caregivers may assume the care
theories, see the Recommended Readings role. This is referred to as dependent care.
at the end of this chapter. We encourage The goal of nursing in Orems theory is
you to read these theories and speculate to help people meet their own therapeu-
about them as possible sources of testable tic self-care demands. Nursing care is
hypotheses or research questions. provided only when there is a decit in
This section briey describes two pop- the self-care or dependent care ability of
ular nursing theories and one borrowed the individual or the caregiver. When this
theory that have been used by nurse re- occurs, the nurse operates within one of
searchers to derive predictions that are three nursing systems to meet the self-
tested through research. Also provided care needs of the client. The three nurs-
are summaries of published research ing systems are wholly compensatory,
studies that were conceptualized within partly compensatory, and supportiveed-
the respective theoretical frameworks. ucative. Figure 2.4 illustrates a model of
Orems theory, and Box 2.4 describes a
a. Orems Theory of Self-Care nursing research study whose theoreti-
cal framework included Orems theory as
Orems theory of self-care (1995) is an ex- well as self-efcacy theory. As you read
ample of a nursing theory that has gained the summary of this nursing study, note
BOX 2.3 Examples of Frameworks to Guide Nursing Research
Grand Theories (Conceptual Models) Practice Theories

Roys adaptation model Theory of interpersonal relations
Rogers model: science of unitary man Theory of clinical reasoning
Neumans systems model Theory of non-nutritional infant sucking
Middle-Range Theories Non-Nursing Theories (Borrowed Theories)
Penders theory of health promotion PRECEDE model of health behavior
Jessors problem behavior theory Theory of reasoned action
Andersons stress and nursing support systems The health belief model
theory Social learning theory
56 GETTING STARTED
teraction between the nurse in his or her

professional role and the client in his or
her state of anxiety manifested in illness.
Through investigative interviewing, the
nurse works with the client to uncover un-
used competencies and dormant powers.
Through interpersonal interaction with
the client, the nurse facilitates the clients
Image/Text rights unavailable ability to transform symptom-bound en-
ergy into problem-solving energy. Latent
capacities become realized competencies
by the client. The resultant transformation
of anxiety moves the client toward health
and outside the boundary of the nurse
client relationship. The theory has been
used extensively to direct practice, but
only limited empirical testing has been
done through research.
Forchuk (1994) conducted a nursing in-
vestigation to test Peplaus theory by de-
termining if variables she identied within
her theory as signicant were related to
the development of the nurseclient rela-
tionship during the orientation phase of
the process. Box 2.5 presents the high-
how the theoretical framework guided
lights of this investigation.
development of the study hypotheses
and interpretation of results. For a com-
plete description of the study, refer to the c. Theory of Planned Behavior
original source.
The theory of planned behavior (Ajzen
1985, 1991) is a theory borrowed from the
b. Peplaus Interpersonal discipline of psychology. The theory can
Relations Theory be used to explain virtually any behavior
over which an individual has control. It is
Peplaus interpersonal relations theory designed to predict behavior and to en-
(1998) is a nursing theory that has been hance understanding of the psychological
widely cited by nurses, particularly psy- determinants of that behavior (Fig. 2.6).
chiatric nurses, as a framework that According to the theory, intention
guides practice. This theory describes the to perform or not perform a behavior is
nurseclient relationship as the crux of the immediate determinant of that behav-
nursing. The theory states that this rela- ior. Intention, in turn, is a direct function
tionship evolves through four intercon- of three independent variables: attitude,
nected phases: orientation, identication, subjective norms, and perceived behav-
exploitation, and resolution. Figure 2.5 il- ioral control. The theory of planned be-
lustrates a model of Peplaus theory. The havior has been used by nurse investiga-
central components of the theory include tors to study a multitude of health
interpersonal process, nurse, client, and behaviors. Examples include sexual deci-
anxiety. The interpersonal process is the sion making, birth planning intentions,
primary component and includes the in- Lamaze childbirth intentions, contracep-
EFFECTS OF OREM-BASED NURSING INTERVENTION

ON NUTRITIONAL SELF-CARE
The purpose of this study was to test the effec- The control group did not receive the home
tiveness of an intervention based on Orems visit nor were diet issues dealt with in the three
theory of nursing on the nutritional self-care of follow-up phone calls unless the patient intro-
104 patients who suffered from myocardial induced the subject. Data were collected from
farctions (MIs). Orem posited eight universal both groups 7 weeks after discharge.
self-care requisites for maintenance of health. Results indicated significant differences be-
One of these, food, was the focus of this study. tween groups on their intake of total fat and
The theory of self-efficacy was also included in saturated fat, and food habits questionnaire
the theoretical framework. Self-efficacy is a per- scores, with the treatment group scoring bet-
sons belief in his or her own ability to perform ter on these measures at time 2. Significant
a certain behavior and be motivated to persist changes over time on SCA and self-efficacy for
in that behavior. It was expected that if the healthful eating were noted for the treatment
nursing interventions were effective in assisting group between time 1 and 2. The control group
patients to eat a healthful diet, then patients increased only their self-efficacy during this pe-
would be more confident in this ability and riod. At time 2, there were no significant differ-
more motivated to eat a healthful diet because ences between the group means on measures
of their increased self-efficacy. of either SCA or self-efficacy.
Nutritional self-care was measured with a The nature of the relationship between SCA
24-hour diet recall for 3 days and completion of and nutritional self-care was expected to be
the food habits questionnaire. Self-care agency positive and moderate because nutrition is only
(SCA) was measured with the SCA scale. Self- one of the eight aspects of self-care addressed
efficacy for healthful eating was measured by in the instrument. Study findings were consis-
the eating habits confidence scale. Baseline data tent with this expectation. The finding that self-
were collected from both groups in the hospi- care increased in the treatment group but not
tal. Treatment group subjects were then visited the control group suggested that the effect
in their homes 1 week after discharge to begin of the nursing action was related to SCA.
the treatment intervention, which included ob- Concepts from the psychological theory of self-
servation and questioning on dietary behavior, efficacy were less useful in this study. In con-
meal preparation, and encouragement in nutri- clusion, Orems theory proved useful in guiding
tional self-care. During the next 6 weeks, all strategies for effective care and in providing an
subjects received three follow-up phone calls. explanation of how the intervention worked.
SOURCE: Summarized from Aish, A.E., and Isenberg, M. (1996). Effects of Orem-based nursing intervention on nu-
tritional self-care of myocardial infarction patients. International Journal of Nursing Studies, 33, 259270.
tive decision making, intention to engage mary of this study, pay particular atten-
in leisure activities, problem drinking, tion to the hypotheses statement and the
losing weight, and exercising. Box 2.6 pre- results of the study, which lend some
sents the summary of a study by Hanson support to the theory of planned behav-
(1997), which applied the theory of ior. The study is an example of how re-
planned behavior to cigarette smoking in search can lead to theory renement and
teenage girls. The theory was tested us- how theory can direct future research.
ing the path analysis method discussed For a complete description of the study,
later in this text. As you read the sum- refer to the referenced source.
58 GETTING STARTED
Image/Text rights unavailable
E. TESTING AND USING As you know, a theory may be thought

THEORY IN RESEARCH of as a cluster of tentative ideas put for-
ward to explain something; research is
As we have seen, the function of theory then conducted to test the adequacy of
is to describe, explain, and predictthat these ideas. Considerable debate exists
is, to offer satisfactory, testable explana- in nursing science about the proper rela-
tions for relationships. A testable expla- tionship between theory and methods.
nation is one that can be disconrmed Most seem to agree that they are best
(i.e., one that may turn out to be true or viewed as partners in our attempts to ex-
false). A satisfactory explanation is one tend our understanding of human behav-
that follows tough rules of evidence and ior. In that way, both theory and methods
is free from the aws in explanation dis- may be regarded as tools. Renements in
cussed at the beginning of this chapter. our theories are neither more nor less im-
THE ORIENTATION PHASE OF THE NURSECLIENT RELATIONSHIP:

TESTING PEPLAUS THEORY
This investigation used a correlational design to viii. Taken together, the nurses preconcep-
test Peplaus theory regarding influences during tions of the client, level of anxiety, and in-
the orientation phase of the nurseclient rela- terpersonal relationships will be a better
tionship. The orientation phase is a clinically sig- predictor of progress in the development
nificant yet difficult phase with clients with a of therapeutic relationships than any one
chronic mental illness. The study tested eight nurse variable alone.
hypotheses that focused on the independent Purposive sampling was used to select
variables of preconceptions, interpersonal rela- subjects from a long-term care program serv-
tionships, and anxiety, measured at the begin- ing mentally ill clients. The sample included
ning of the nurseclient relationship. These 124 newly formed nurseclient dyads. Two in-
were expected to predict later results for the struments measured the dependent variable,
dependent variable, the development of the development of the therapeutic relationship:
therapeutic relationship as proposed by the The Relationship Form and the Working Al-
theory. The hypotheses tested were: liance Inventory (WAI). The Personal Resource
i. Clients more positive preconceptions of Questionnaire, the Beck Anxiety Inventory, and
the nurse will be related to greater pro- the Preconceptions Semantic Differential Scale
gress in the development of therapeutic re- measured the independent variables of inter-
lationships. personal relations, anxiety, and preconcep-
ii. Nurses more positive preconceptions of tions, respectively.
the client will be related to greater pro- Results indicate that the preconceptions of
gress in the development of therapeutic re- both the nurse and client were most strongly
lationships. related to the development of the therapeutic
iii. Clients more positive interpersonal rela- relationship (hypotheses i and ii supported).
tionships will be related to greater progress Clients other relationships were significant on
in the development of therapeutic relation- the WAI only (hypothesis iii partially sup-
ships. ported), and nurses other interpersonal rela-
iv. Nurses more positive interpersonal relationships were not significant (hypothesis iv not
tionships will be related to greater progress supported). Anxiety, of nurses and clients, was
in the development of therapeutic relation- not significantly related to progress in the rela-
ships. tionship (hypotheses v and vi not supported).
v. Higher levels of anxiety in the client will be Hypothesis vii was partially supported, and hy-
related to less progress in the development pothesis viii was not supported.
of therapeutic relationships. In conclusion, this study supports some of the
vi. Higher levels of anxiety in the nurse will be tenets of Peplaus theory, but not others. It gives
related to less progress in the development direction for further theory refinement and for
of therapeutic relationships. future research. Specifically, anxiety and other
vii. Taken together, the clients preconceptions relationships may be less important than the the-
of the nurse, level of anxiety, and interper- ory currently suggests. Continued testing of this
sonal relationships will be a better predic- theory will guide future theory refinement and
tor of progress in the development of ther- the application of theory in practice.
apeutic relationships than any one client
variable alone.
SOURCE: Summarized from Forchuk, C. (1994). The orientation phase of the nurse-client relationship: Testing Pe-
plaus theory. Journal of Advanced Nursing, 20, 532537.
60 GETTING STARTED
portant than renements in our methods. range theories. It is likely that, as middle-
They are equal partners in the knowledge range theory becomes the norm, new
enterprise. nursing research methods will be devel-
Hypothesis testing is a part of theory oped. Middle-range theories involve a
testing but, by itself, does not necessarily number of formally stated, interconnected
constitute a test of theory. The funda- propositions; contain clearly dened vari-
mental issue in theoretical research is to ables; and indicate the nature (direct or in-
demonstrate the link between a set of the- verse) and the direction (what is causing
oretical propositions and the conceptual what) of the relationships among the vari-
hypothesis that has been selected for ex- ables. The propositions normally take the
amination. Only to the extent that this form of the greater the A, the greater the B.
connection has been articulated success- Not all the concepts in the propositions
fully can we claim our research has bear- need to be directly measurable; however,
ing on theory. This section considers at some point, a connection to measurable
methods of testing theories. variables must be made. It is this link that
must be specied for ones research to
1. Testing Middle-Range Theory have bearing on a theoretical formulation.
We will examine two methods used to
As the discipline of nursing matures, the demonstrate such connections: these are
emphasis on knowledge development is axiomatic derivations and replacement of
moving from grand theories to middle- terms.
THE THEORY OF PLANNED BEHAVIOR APPLIED

TO CIGARETTE SMOKING
The purpose of this study was to evaluate the The Fisbein/Ajzen-Hanson Questionnaire
adequacy of Ajzens Theory of Planned Behav- (FAHQ) was constructed in accordance with
ior to predict cigarette smoking intention in the Theory of Planned Behavior to measure
three groups of teenage girls. Participants were the study variables. This instrument was ad-
141 African-Americans, 146 Puerto Ricans, and ministered to subjects attending family plan-
143 non-Hispanic whites, age 13 to 19 years. ning clinics.
The hypothesis tested was derived from the Separate path analyses were used to test the
theory and tested all the predictor variables in study hypothesis for each group. To summarize,
the theory. It stated: The belief-based mea- all six paths of the model were statistically signif-
sures of attitude, subjective norm, and per- icant for African-Americans. Hence, the hypoth-
ceived behavioral control have an indirect effect esis was supported for African-American teenage
on cigarette-smoking intention and a direct ef- girls. For Puerto Rican and non-Hispanic whites,
fect on attitude, subjective norm, and perceived only the relationships among attitude, perceived
behavioral control, respectively. These, in turn, behavioral control, and smoking intention were
have direct effects on cigarette smoking inten- supported. Hence, the hypothesis was not sup-
tion in African-American, Puerto Rican, and ported for these two groups. This study suggests
non-Hispanic white teenage females. These that the Theory of Planned Behavior provides an
represent the six essential paths diagrammed in adequate explanation of cigarette smoking among
Figure 2.6, the conceptual map of the theory. African-American teenage girls.
SOURCE: Summarized from Hanson, M.J. (1997). The theory of planned behavior applied to cigarette smoking in
African-American, Puerto Rican, and non-Hispanic white teenage females. Nursing Research, 46(3), 155162.
a. Axiomatic Derivations are accurate. Derivations are made to lo-

cate testable hypotheses that then con-
Axiomatic derivations are new state- stitute a test of the theory. The reason
ments of relationship logically derived one makes such derivations is to provide
from a given set of assumptions and many different tests. The reason one
propositions. An example is provided in wants to do different tests is that ideally
Box 2.7 using a set of propositions found one wishes to identify a theoretically pre-
in work by Teevan (2000). dicted relationship but one that is not
Instead of using low premarital sexual obvious.
permissiveness as the example of a Why would one want a counterintu-
conservative value, one could instead itive relationship? Here the answer is psy-
use any number of alternative replace- chological rather than scientic. If an un-
ments such as: expected relationship is predicted, and if,
indeed, the results of the study conrm it,
Low likelihood of using drugs
the evidence is much more convincing.
Deferring marriage
On the other hand, if the predicted rela-
Choosing traditional careers for ones
tionship is commonsensical, even if it is
gender
conrmed, the critic will claim that the
The new derived propositions should results only show what everyone already
be true if the assumptions, derivations, knew and certainly do not demonstrate
and original theoretical propositions any theory. The verication of Albert Ein-
62 GETTING STARTED
permit one to rene and specify condi-

tions under which the theory does, or
does not, hold.
Using a combination of axiomatic de-
rivations and replacement of terms pro-
vides powerful, yet simple, methods of
deriving interesting testable hypotheses.
The student would be well advised, how-
ever, to directly test the propositions of a
Image/Text rights unavailable theory and, if that is not possible, to re-
strict axiomatic steps to a minimum in lo-
cating testable hypotheses. Because hu-
man relations are so complex and because
few relations are extremely powerful, it is
problematic to make a large number of
axiomatic steps and still have an ironclad
connection to the original set of proposi-
tions. So, keep it simple!
2. Testing Practice Theory
Routine hospital practices are often the

subject of nursing research. Because of
the desire to base practice on evidence,
steins simple but counterintuitive pre- graduate students and nurse researchers
diction that a rapidly moving clock should frequently evaluate current nursing prac-
run slower than a clock that is moving tices. Indeed, many ne masters theses
slowly provided powerfully convincing ev- are done evaluating a routine practice. So
idence for his theory of relativity. whether it is assessing the comfort lev-
els of patients controlling their pain
medication or assessing the impact of the
b. Replacement of Terms timing of the bathing of newborns on the
temperature decreases experienced by
A further way of extending the number of
newborns, the research can provide evi-
predictions is to use a technique called
dence for continuing current practice or
replacement of terms. Replacement of
altering practice. Indeed, one could argue
terms refers to replacing general theoret-
that the key to evidence-based practice is
ical concepts by specic examples of
the continuing questioning and evalua-
these concepts. For example, if the gen-
tion of routines found in nursing practice.
eral concept is health behavior, such a
Indeed, it is conceivable that future di-
concept could be replaced by a specic
rections in nursing research will include
instance of health behavior such as seat-
increased efforts at evaluating situation-
belt use, exercise, or meditation. To the
specic practice theories.
extent that one is able to derive new pre-
dictions through such replacements, one
can provide a virtually unlimited number 3. Structural Equation Modeling
of testable relationships. One can then
select the better ones, choosing those Structural equation modeling is an ad-
that are counterintuitive and those that vanced statistical method designed to
test theories. In a theory, all the concepts tion/regression family of statistical tech-
are expected to be related. Conceptual niques; a preliminary discussion of these
maps discussed earlier in the chapter ex- techniques is given later in the text.) For
press this web of relationships. Testing a more detailed discussion, interested
the structure of the relationships in a the- students are advised to consult an ad-
ory as a whole provides more informa- vanced statistics or methods text.
tion about the validity of the theory than
testing only specic relationships or 4. Testing Alternative
propositions. This can be achieved using Explanations
structural equation modeling with path
analysis. The researcher anticipates cer- The testing of alternative explanations is
tain outcomes from the structural equa- a primitive form of theory testing. This is
tions if the model is correct. This consis- the case because theory and explanation
tency does not prove the accuracy of the do the same thing: answer a why ques-
theory but it does support it. Figure 2.7 il- tion. The reason we identify the testing of
lustrates a causal model used by Porter alternative explanations as a primitive
and associates (1996) in their study of inform of theory testing is that such formu-
uences on the sexual behaviors of ele- lations typically fail to fully identify the
mentary and middle school youths. Note linkages to the framework that is being
the arrangement of the variables and how used.
they are ultimately linked to the depen- The rst thing we need in order to test
dent variable, self-reported initiation of a theory using alternative explanations is
sexual intercourse. Models such as the a relationship that, we believe, needs to
Porter et al. one are known as path analy- be understood. Then we propose a series
sis models. The arrows and lines repre- of alternative explanations for the rela-
sent the causal connections between the tionship. After that, all we have to do is to
variables in the model. (Such models are measure the appropriate variables and
analyzed using techniques in the correla- do the analysis.

64 GETTING STARTED
For example, two possible explanations possess more conservative attitudes

might be proposed for the connection be- and this leads to a deferral of initiation
tween the age of initiation of sexual inter- into sexual intercourse; or
course and socioeconomic background: Whether students with higher SESes
are more likely to experience less peer
Whether students with higher SESes pressure to engage in sexual activity.
> SOCIO-ECONOMIC < Age of Initiation into

STATUS BACKGROUND Sexual Intercourse
> Inculcated with

Conservative Sexual
Attitudes
> Associated with Peers

who do not Pressure to
Engage in Sexual
Intercourse
Appropriate research would allow us the theory. Without such explicitness, it

to evaluate the adequacy of each of the is difcult to construct truly general, and
two possible explanations. We would therefore powerful, theories of human
need measures of the students SES back- behavior. When we seek to explain par-
ground, age of initiation into sexual activ- ticular relationships with highly particu-
ity (if at all), a measure of how conserva- larized explanations, we do not advance
tive their attitudes are, and a measure of the general state of knowledge much.
the amount of peer pressure they have Hence, more formal approaches to the-
experienced to engage in sexual activity. ory construction are to be encouraged.
The rst explanation uses a socialization Nonetheless, especially for beginning re-
of conservative values explanation (but searchers, testing alternative explana-
connected to what general theory of hu- tions for relationships may make impor-
man behavior?). The second relies on a tant contributions to the discipline.
peer pressure explanation but, again, to The basic requirements for the testing
what general theory is this connected? of alternative explanations are that:
Nevertheless, because both explanations
are answers to a why question, each may One has a relationship between two
be viewed as constituting a theory of hu- variables that one suspects will be
man behavior (albeit an implicit theory). strong (either positive or negative);
The advantage of the more formal con- One has a series of alternative expla-
structions of nursing theory (middle- nations for the relationship;
range theory) is that we are forced to That it is possible to get measures for
make explicit our core assumptions of the appropriate variables; and
That after data are collected, suitable in nursing research. There is a strong
procedures are used for the analysis of tendency for some researchers to be
the data so that an evaluation can be identied with the work of one particular
made of the adequacy of the compet- nurse theorist or one methodological ap-
ing explanations. proach. And even if people span a variety
of approaches, we tend to label them as
supporters of one particular theory and
5. Theoretical Triangulation paradigm. There are squabbles between
the various schools of thought or para-
Recently nurse researchers have pro- digms over the kinds of evidence that
posed theoretical triangulation as a should be used and, indeed, serious ar-
strategy for theory testing. This involves guments over which questions we should
the testing of two competing or alterna- be asking as nurse scientists.
tive theories within one study. As previ- On occasion, such debate is carried on
ously stated, almost all phenomena can with a distinct lack of tolerance for differ-
be explained using alternative explana- ences. Perhaps it is a measure of the im-
tions. Hence, a researcher who directly maturity of our discipline that we are oc-
tests alternative explanations using a sin- casionally divided into warring camps.
gle sample of subjects is in a position to By the end of their university nursing pro-
make powerful comparisons about the gram, most students will witness at least
utility of competing explanations. Theory one professor make an unrelenting attack
triangulation can rule out rival hypothe- on some theoretical or methodological
ses, avoid premature acceptance of plau- perspective not shared by that individ-
sible concepts, prevent ignoring con- ual. Although debates in universities can
tradictory propositions, and help in re- be healthy and invigorating, so too can
ning theoretical systems (Banik, 1993). they be destructive and counterproduc-
Bennett (1997) provides a concise exam- tive if they fail to give the student a bal-
ple of theory triangulation in her study anced view of the various approaches.
of nurses attitudes about AIDS. In this But are there also points of agree-
study, Orems (1995) theory of nursing ment? Most nurse scientists would agree
systems and Travelbees (1971) theory that we should:
of interpersonal nursing are used to ex-
plore whether patient characteristics Study the full range of human behav-
and lifestyle are more likely to inuence ior in health and illness states. This is
nurses attitudes than nurses own caring to acknowledge that it is equally legiti-
competencies. Students are encouraged mate to study both social construc-
to read this source if they are interested tions of reality (individual subjective
in learning more about this method of perceptions of reality) and objective,
theory testing. This example demon- measurable structural patterns found
strates how the exploration of related in any society. In short, both qualitative
questions not addressed by a single the- and quantitative methodologies have
ory may be aided by searching for links contributions to make in understand-
among middle-range theories. ing human responses and behavior.
Use the methodology appropriate to
the questions asked. Most people
F. POINTS OF AGREEMENT ON agree that the methodology we apply
RESEARCH AND THEORY should be appropriate to the question
asked. And although we could never
Nurse scientists have a variety of per- agree that either qualitative or quanti-
spectives available to them on theory tative methodologies are better, we
66 GETTING STARTED
could probably agree that both have scrutiny. Whatever our efforts, an im-
their place in nursing research. portant norm is that our scholarly
Use systematic explanation. Expla- work should be scrutinized by our
nations offered by nurse scientists peers. It is through circulating our
should be based on a careful, painstak- analyses that others may criticize
ing analysis of the evidence. them, suggest alternatives, and, ulti-
Subject our scholarly work to peer mately, rene the product further.
E X E R C I S E S
1. Choose a major theoretical work iden- perspective. You may use a nursing or
tied in the nursing literature. Then do non-nursing perspective. Outline a
a propositional inventory of a section study that would allow you to reject
of the work, selecting the major propo- the various explanations.
sitions involved. Label the concepts
with letters, using the same letter for
those you presume are intended to re-
fer to the same variable. Attempt to de- 3. Starting with the proposed explana-
rive new theoretical statements using tions you identied in question 2
axiomatic derivations and replace- above, use Fitzpatrick and Whall
ment of terms. Be certain to specify as- (1996) Conceptual Models of Nursing:
sumptions you are making in doing the Analysis and Application (3rd ed.) to lo-
derivations. Identify testable proposi- cate a theory that could be linked to
tions you think are not obvious on a each of your three explanations.
commonsense basis.
4. Examine the articles in one issue of a

2. Choose a relationship that you think nursing research journal. Search for an
would hold true and propose three illustration of each type of awed rea-
alternative explanations for the rela- soning discussed in the text. Describe
tionship. Connect your proposed ex- why do you think each example is an il-
planations to an existing theoretical lustration of awed reasoning.
Chinn, P., and Kramer, M. (1995). Theory and relationship between theory and research.
Nursing: A Systematic Approach. St Louis: It addresses the analysis and evaluation of
Mosby. This text discusses nursing theory research reports and proposals.
development and emphasizes the relation- George, J.B. (1995). Nursing Theories: The Base
ship between theory, research, and prac- for Professional Nursing Practice (4th ed.).
tice. It provides a solid foundation for un- Norwalk, CT: Appleton & Lange. This text
derstanding theory development in relates the work of 21 nurse theorists to
nursing. nursing practice. It provides a good dis-
Fawcett, J. (1999). The Relationship of Theory cussion of the place and value of nursing
and Research. Philadelphia: F.A. Davis. This concepts and theories to practice.
text presents an excellent discussion of the Im, E., and Meleis, A. (1999). Situation-specic
theories: Philosophical routes, properties, vances in Nursing Science, 22(3), 5875. This
and approach. Advances in Nursing Science, is a good example of middle-range theory
22(2) 1124. This article discusses situa- developed from grounded theory research.
tion-specic theory as a future direction of Meleis, A. (1997). Theoretical Nursing: Devel-
theory development in nursing. opment and Progress (3rd ed.). Philadel-
Leenerts, M., and Magilvy, J. (2000). Investing phia: J.B. Lippincott. This text provides a
in self-care: A midrange theory of self-care comprehensive and creative discussion of
grounded in the lived experience of low- nursing theory and promotes application
income HIV-positive white women. Ad- in practice of emerging theories.
Chapter 3
Deciding What to Study
CHAPTER OUTLINE
A. Choosing a Topic and Stating the 2. How to Develop Theoretical or Causal

Problem Models
1. How to Choose a Research Topic D. Specifying Hypotheses and Procedures
2. How to Identify Problems Not of Analysis
Answerable through Research 1. How to Specify Hypotheses
3. How to Focus the Research Problem and 2. How to Specify Methods of Analysis
Purpose
E. Understanding Research Design
B. Reviewing the Literature
1. How to Describe a Studys Design
1. Importance of the Literature Review 2. What Factors Guide Selection of a Study
2. How to Conduct the Literature Design?
Review 3. How to Choose a Research Design
3. How to Write the Literature Review
F. Predictable Setbacks and Challenges
C. Specifying the Model G. Determining Your Readiness to Start
1. How to Identify the Variables the Project
KEY TERMS
Antecedent variable model Longitudinal Research problem

Candidate variable model Multivariate models Research question
Causal model Path model Research topic
Cross-sectional Problem statement Retrospective
Discrepancy Prospective Source of spuriousness
Intervening variable Retrospective model
Literature review Research design
68
DECIDING WHAT TO STUDY 69
disciplines. To provide some sense of the
D
eciding what to study is sometimes
the most difcult part of a project. range of topics that can be explored, a
But care taken during the beginning listing of some projects carried out by
stages of a project will pay off hand- nurse researchers is found in Box 3.1.
somely. A carefully designed study that
enables the researcher to come to conclu-
1. How to Choose a Research
sions about a nursing phenomenon will
Topic
provide the basis for a solid research proj-
ect. If you can also include theory-testing Given the considerable effort required to
dimensions and connect the project to the complete a projectand even more so to
existing literature, you will have the be- build a research careerone of the single
ginnings of an excellent project. most important considerations in choos-
We will begin with the elements of ing an area of investigation is genuine in-
choosing a research topic and problem, re- terest. Do not underestimate this point!
viewing the literature, selecting an estab- Research programs require signicant
lished or developing a new theoretical time, energy, intellectual input, and com-
framework or causal model to guide the mitment from the investigator; therefore,
study, and formulating a research hypoth- it is important to select topics that you
esis or question. After the question or hy- are passionate about and that will hold
pothesis has been determined, we will pro- your interest over the life of the projects
ceed to select a design appropriate to the that comprise your program of research.
question. This includes making decisions Many nurse authors have identied
about the sample, setting, measurement sources of researchable topics in nursing
methods, data collection, and data analy- (Brown, 1999; LoBiondo-Wood, 1998; Fain,
ses procedures. This chapter deals pri- 1999; Norwood, 2000; Parahoo, 1997). In
marily with the rst four elements. Subse- the United States, the National Institute of
quent chapters discuss data analysis and Nursing Research (NINR), in planning
the interpretation of the results of a study. future research for the years 2000 to 2004,
This chapter focuses on the beginning has selected seven key areas that will
steps of a typical quantitative study. For dominate the institutes research portfolio
students planning on answering a quali- and be funding priorities. These include
tative research question, the steps for chronic illness, health promotion and dis-
such projects are outlined in Chapters 6 ease prevention, quality and cost effective-
and 7. When feasible, however, distinc- ness of care, management of symptoms,
tions are made in this chapter between adaptation to new technologies, health
qualitative and quantitative approaches. disparities, and palliative care. Many re-
search problems can emerge from these
seven areas.
Some suggestions of how to go about
A. CHOOSING A TOPIC AND
choosing a topic based on our experience
STATING THE PROBLEM
as researchers and the views of other au-
thors are included here.
A research topic is dened as the broad
general area you wish to investigate. The
exciting thing about nursing research is a. Improving Clinical Practice
that most of the research topics have not
been studied sufciently. This is because Curiosity is one of the greatest assets a
nursing research is in its infancy com- nurse investigator has in identifying re-
pared with many other research-based search problems. As an inquiring student
70 GETTING STARTED
NURSE RESEARCHER PROJECTS
These projects were designed by nurse re- journals and include a variety of research de-
searchers or interdisciplinary research teams signs. Full references are provided in the refer-
with nurse researchers as part of the member- ence list at the conclusion of the text.
ship. They were selected from refereed nursing
Research Project Title Author

Curriculum trends in nurse practitioner programs: Bellack et al., 1999
Current and ideal
Infertility education in baccalaureate schools of nursing Sherrod, 1998
Competencies of liberal education in post RN Gillis, MacLellan, & Perry, 1998

baccalaureate students: A longitudinal study
Determinants of a healthy lifestyle in new fathers Walker, Flescher, and Heaman,

1998
Maternal smoking or nonsmoking and feeding choice Edwards, Sims-Jones, &

Briethaupt, 1998
The male experience of caregiving for a family member Parsons, 1997

with Alzheimers disease
Nerves as status and nerves as stigma: Idioms of distress: Davis & Joakimsen, 1997
And social action in Newfoundland and Norway
Stressors in families with a child with a chronic condition Ogden-Burke, 1998
Views on assignment of publication credit for scholarly Butler & Ginn, 1998
and scientic work
Courage in middle-aged adults with long-term Finfgeld, 1998

health concerns
Entry into community-based nursing practice: Hahn, 1998

Perceptions of prospective employees
The changing self-concept of pregnant and parenting Alpers, 1998

teens
Care: A value expressed in philosophies of nursing Tuck et al., 1998

services
Hospital nurses and health promotion Berland, Whyte, & Maxwell, 1995
Womens anger: Relationship of suppression to blood Thomas, 1997

pressure
continued on next page
BOX 3.1 Nurse Researchers at Work (Continued )
NURSE RESEARCHER PROJECTS
Research Project Title Author
Follow-up study of children with growth deciency Reifsnider, 1998
Therapeutic touch as a nursing intervention Meehan, 1998
What a difference a nurse makesthen and now MacKinnon, 1997
Youngsters caring for adults with cancer Gates & Lackey, 1998
of research, you simply need to reect on to seeing discrepancies in practice and

any practice situation and ponder about not accept pat answers to problems that
discrepancies you see in practice. Simply are frequently inappropriately accepted in
stated, a discrepancy is the difference be- practice.
tween the ways things are in practice and
the way they ought to be, or between what b. Exploring a Pattern of Incidents
one knows and what one needs to know to
eliminate a problem (Diers, 1979). The previous example provides an impor-
Many nursing investigations are born tant characteristic of research problems
out of a nurses desire to improve practice that is, a pattern of incidents in clinical sit-
situations and eliminate discrepancies so uations that bears a relationship to more
that clients can cope better, feel better, or general situations. If only one elderly pa-
live life more fully. Discrepancies become tient was confused and agitated after hip
relevant sources of nursing problems. For surgery, you would not design a study to
example, you may notice that elderly pa- address this isolated problem. However, if
tients who undergo surgery for a fractured a problem can be conceptualized at a
hip experience confusion and delusions higher level that will be generalizable to
after surgery. This is a serious problem be- others, it should be investigated. Astute
cause they are often disoriented and at- nurses reect on their practice to identify
tempt to get out of bed on their own. As a common patterns of phenomena across
nurse, you spend a great deal of time try- client situations that require investigation
ing to prevent falls, orient them, and reas- to be fully understood. Examples of such
sure them. You mention this to more ex- patterns may include separation anxiety in
perienced nurses on the unit, and they hospitalized children; effective coping in
may comment that everyone is like that af- social support groups; burnout in long-
ter surgery, that it is an expected effect of term caregivers of clients with Alzheimers
the anesthetic. You wonder whether this disease; hopelessness in terminally ill
rationale is based on evidence and, if so, clients; or intuitive decision-making pro-
what, if anything, could be done to pre- cesses in expert nurses. Rather than auto-
vent it. You have the beginning of a highly matically accepting a pattern as normal,
relevant and clinically signicant nursing you need to question its existence and ask
problem with potential to improve prac- what can be done to better understand the
tice. The point here is you need to be open phenomenon and enhance practice.
72 GETTING STARTED
c. Testing Folk Wisdom a new perspective, namely from the per-

spective of those experiencing it. That is,
Interesting projects can often be de-
much may be known about a particular
signed to test systematically the accu-
experience, procedure, or nursing inter-
racy of some taken-for-granted wisdom.
vention from the viewpoint of the nurse;
You can have a lot of fun testing knowl-
however, there may be little or no docu-
edge passed on from one generation of
mentation in the literature about the
nurses to another. If you try it, you might
clients view of the phenomenon. For ex-
be amazed at some of the results. Con-
ample, the literature may contain many
sider the following examples:
studies on cardiac stress testing and the
Do whirlpool baths consume fewer best ways to prepare a client for such
physiological resources than bedbaths? testing. However, you note that no one
Traditional wisdom has supported the has considered this topic from the per-
use of bed baths. Such folk wisdom war- spective of the client who is undergoing
rants testing through research. the stress testing. You may be interested
Among psychiatric nurses, are female in understanding preparation for stress
nurses more empathetic to clients testing from the clients perspective. At-
needs than male nurses? tention to the insiders view on a broad
Is daily vital sign monitoring as ef- range of phenomena can be a rich and ex-
fective as monitoring every 4 hours citing source of topics for nursing inves-
in predicting occurrence of complica- tigations. Those interested in exploring
tions in institutionalized elderly recov- qualitative research topics will nd this
ering from abdominal surgery? approach particularly useful in generat-
Does engagement in regular exercise ing researchable questions of relevance.
reduce the risk of falls in hospitalized
elderly clients?
Are audio taped end-of-shift nursing e. Tackling Current Issues
reports effective in promoting continu-
ity of care and informed decision mak- One method of selecting a project is to
ing by staff nurses? choose one that is the subject of profes-
Does it take more energy for cardiac sional debate. Such topics include nurses
patients to take showers than to brush as entry point to the health-care delivery
their teeth? system, cost of insured nursing services,
Spend some time thinking about these expanded roles for nurses, baccalaureate
and other commonly accepted views of versus diploma or associate degree edu-
how things work in nursing and then try cation for professional practice, and use
to think of some way in which these ideas of alternative therapies by nurses. A
could be tested systematically. Ask your- word of caution is warranted when se-
self how experimental, eld, survey, or lecting such topics for investigation be-
applied research designs might be used cause investigations of these types of
to study the problem. Which design topics may be politically motivated. For
would be ideal and which one would be example, one group may wish to show
most practical for each of these topics? that nurses can provide services cur-
rently delivered by physicians in a more
d. Understanding Phenomena from cost-effective and efcient manner than
the Insiders Perspective the latter group. Therefore, nurse re-
searchers need to be cautious not to bias
Many times, a researcher may be inter- the study so that results simply support
ested in exploring a concept or idea from the desired conclusions (see Chap. 9).
f. Inconsistencies in the Literature routines, skill mix of staff, and so on. The
most convincing theory-testing projects
Awareness of the research literature on a are those that make a counter-intuitive
particular topic may lead one to note that (against common sense) prediction that, if
there are inconsistencies in the reported it turns out to be true, will be a convincing
ndings. Such awareness is often a stim- demonstration of the theory.
ulus to design a study to explore the rea-
sons behind such inconsistencies. These
projects often are popular because they h. Testing Practice Theories
relate to debates current in the profes-
A review of the literature may reveal a
sion and highlight the need for additional
consistent relationship between two vari-
research.
ables, but alternative explanations for
this relationship may not have been care-
g. Testing a Theory fully tested. For example, a relationship is
found consistently in the literature be-
Researchers who have theoretical inclina- tween perceived self-efcacy and engage-
tions may wish to test a current theory of ment in health behaviors. In such cases,
nursing, a theory that appears to be emerg- it is reasonable to propose alternative ex-
ing in the nursing research literature but planations for the relationship; design a
that has not yet been tested, or one of the study; and then test which, if any, of the
many theories of human behavior that re- proposed explanations best accounts for
late to health or illness situations. Chapter the relationship. (See Chapter 2 for other
2 outlines methods for generating testable examples.)
theoretical hypotheses. The challenge is to
examine a relationship between variables
that is predicted by a theory but that, at i. Exploring Variations in a
the same time, is not obvious to common Dependent Variable
sense. For example, building on the previ-
ous scenario, you may believe that both Another approach is to try to under-
the nursing literature and evidence in stand variations in a dependent variable.
nursing practice suggest that a relation- Which factors influence engagement in a
ship exists between confusion in elderly health-promoting lifestyle by various age
clients and the experience of hip surgery. groups? What inuences engagement in
You decide that this warrants investiga- risk-taking behaviors by adolescents?
tion so means may be found to prevent Which variables inuence the choice of
confusion or assist the elderly to cope cancer treatments among female clients
more effectively. You might be interested with breast tumors? How do children
in testing a theory that predicts when with life-threatening illness nonverbally
and which clients experience confusion communicate their desire to terminate a
and agitation and what factors might nursing interaction? Here the goal is to
prevent this in the elderly. Consequently, understand the factors inuencing the
you might design a study of the variables dependent variable.
associated with confusion and agitation in
the elderly after hip surgery. Variables j. Providing an Evaluation
might include the patients themselves
(e.g., their age, general health state, demo- Nurse scientists are frequently called on
graphics), the surgical procedure, the ane- to provide an evaluation of a nursing in-
sthetic agents used, post-operative med- tervention or health-care policy. Evalua-
ications, postoperative nursing care, unit tions are often required for social, com-
74 GETTING STARTED
munity health, health promotion and pre- the indirect replication could take the
vention programs, as well as many hospi- form of answering a question that was left
tal-based programs. For example, a nurse open by the previous project.
researcher may be asked to evaluate the
clinical outcomes of premature infants
receiving conventional care and those re- 2. How to Identify Problems Not
ceiving developmental care during their Answerable through Research
stay in the neonatal intensive care unit.
Given the enormous nancial and tech- Finally, not all problems of interest to
nological resources allotted to neonatal nurses are answerable through research.
care, this problem is highly relevant to Some problems are philosophical in na-
optimize positive developmental out- ture rather than researchable. Problems
comes for premature infant survivors. A that involve choosing a moral course of
solid understanding of experimental de- action or that are value questions or pol-
sign and a good deal of imagination and icy questions are not answerable through
exibility are qualities well suited to do- research. For example, should children
ing evaluations. There is a demand for with terminal illness have extraordinary
nurse researchers who know how to do means used to prolong the dying process?
evaluation work. Or, is it ethical to deny a client informa-
tion about their diagnosis if the family re-
quest it? Or, is abortion on demand good
k. Implementing and Studying a practice? These questions cannot be an-
Nursing Action Simultaneously swered through empirical investigation;
rather, they are questions for philosophi-
In todays world of rapidly changing
cal inquiry and for public policy debates.
health-care practices, it is important to
It should be understood that although
be able to take action to solve problems
policy questions are not directly answer-
and systematically study the effects of
able through research, research can in-
the action at the same time. As we dis-
form or shape policy development. Nu-
cuss in Chapter 8, action research en-
merous professional journals, such as the
ables researchers to identify whether or
Journal of Nursing Scholarship, regularly
not the implemented solution is resolving
publish articles with policy implications.
the problem. Action research leads to the
generation of practical knowledge that
relates to practice in a particular setting.
3. How to Focus the Research
Action research topics are different from
Problem and Purpose
clinical problems in which one expects
the results to be generalizable across a
Having selected the research topic the
range of settings.
next step is to articulate precisely what
you wish to investigate. To do this, the re-
l. Replicating a Study searcher must decide on a specic aspect
of the research topic that warrants inves-
One can choose to replicate, or repeat, tigation and then proceed to describe the
some earlier study (following exactly the precise problem and develop a statement
same procedures as the original study but of purpose. Before we proceed to explain
based on a different sample). Alterna- the process of limiting the problem and
tively, one may do an indirect replication formulating a statement of purpose, it is
by adding a new dimension to the study. In helpful to clarify several terms important
addition to asking the original questions, to the conceptual phase of a research proj-
ect: research problem, problem state- to health-promoting lifestyles in adoles-

ment, research question, and statement of cent girls? Or, What is the meaning of
purpose. caring for nurses who provide services to
disenfranchised patient populations?
a. Terminology The purpose of a study is the specic
aim or goal you hope the study to accom-
The research problem is a situation or plish. For example, The purpose of this
circumstance that requires a solution to research is to examine the relationship of
be described, explained, or predicted. It maternal and paternal health-promoting
is an unsatisfactory situation that we lifestyles to health-promoting lifestyle in
want to confront (Norwood, 2000). There adolescent girls. Or, The purpose of this
is a knowledge gap in an area that needs research is to understand the meaning of
to be investigated. The research problem caring for nurses who provide services to
identies the knowledge gap that needs disenfranchised patient populations.
to be lled. Whereas the research topic is
simply a broad area of interest, the re-
search problem identies what is prob- b. Limiting the Problem
lematic about the topic.
The process of limiting the problem in-
The problem statement is a narrative
volves reecting on the topic, reviewing
that elaborates on the research problem
the literature to determine the current
and identies the specic area of con-
state of knowledge about the problem,
cern. It provides direction for the entire
and talking to those in the eld who have
study and guides the study toward a
had experience with the problem. In lim-
quantitative or a qualitative design. Writ-
iting the problem, the researcher is at-
ing the problem statement helps the re-
tempting to pare away all the unrelated
searcher identify uncertainties that need
information and narrow the problem for
clarication before the study can pro-
investigation to the point where a clear,
ceed (Thomas, 2000). The problem state-
unambiguous, researchable question will
ment includes six elements. These are:
emerge. To illustrate this process, sup-
1. Information about the research topic pose you wished to study the problem of
that provoked the study accidental falls in elderly patients. As you
2. The scope of the problem (e.g., how start to think about this investigation, de-
many people are affected by it) termine whether you wish to examine:
3. Why it is important to study the problem
4. How nursing science would be inu-
Risk factors for falls in the elderly
enced by the study
The types of elderly clients who sus-
tain falls versus those who do not
5. General characteristics of the popula-
tion of interest
The relationship between admission
assessment (e.g., age, medical diagno-
6. The overall goal or aim of the study or
sis, level of orientation, functional sta-
the question to be answered
tus) and the incidence of falls
The research problem stated in the in- Whether there are differences in physi-
terrogative form is referred to as the re- cal environmental factors in hospitals,
search question. It is stated in the pres- nursing homes, and private residences
ent tense. The advantage to stating the that contribute to the incidence of acci-
problem in the interrogative form is that dental falls in the elderly
a question invites an answer. For exam- Whether women are less inclined to
ple, What is the relationship of maternal sustain accidental falls than men, and
and paternal health-promoting lifestyle if so, what accounts for this tendency
76 GETTING STARTED
The list of issues you might wish to The aim of this research was to expli-
consider could be extended. After a prob- cate common life patterns . . .
lem has been selected, the work of speci- This study sought to elicit views of
fying the precise problem statement or nurses on assignment of publication
question begins. Suppose that you de- credit . . .
cided to focus on the rst question (iden- The objective of this research was to
tifying risk factors in accidental falls in the identify factors associated with mater-
elderly); then a thorough review of the lit- nalinfant bonding . . .
erature should be undertaken to nd out The purpose of this study was to an-
what other researchers have discovered. swer the following questions . . .
The review will also help you determine This qualitative study explored the
more precisely what is to be investigated. meaning of respect from the perspec-
At this point, it is helpful to specify, as tive of . . .
much as possible, the precise question to
be studied. This is because the question As research projects are developed, it
determines the research design, and one is normal for the rst ideas to be fuzzy
does not want to make unnecessary and to require an investigation of a rather
changes to the design later because of a grand scale. As the researcher reviews
poorly focused research question. the literature, becomes more familiar
with it, and becomes sensitive to practi-
cal considerations, the project should be-
c. Statements of Purpose come more focused and specic and
move to something that can be accom-
The statement of purpose of a study plished with the available resources.
ows from the problem statement and is
included in it. One research problem
statement may be the basis for several re- B. REVIEWING THE
search purposes. The research purpose LITERATURE
is usually a single declarative statement
that focuses the study and clearly identi- A literature review is a critical step in fo-
es what the researcher intends to do. cusing the research problem and state-
Box 3.2 provides examples of research ment of purpose. It is a process of re-
problem statements worded in the de- viewing the current knowledge about the
clarative and interrogative forms. In nurs- research problem, describing the charac-
ing research journals, however, you will teristics of previous studies in the area,
note that the majority of research prob- noting the similarities and differences in
lems tend to be expressed as statements research results, evaluating the strengths
of purpose. Some authors select to in- and limitations of previous studies, and
clude both a statement of purpose at the identifying gaps in knowledge relevant to
beginning of their article and specic re- the research problem. It should culmi-
search questions at the end of the intro- nate in a nal product that forms a chap-
ductory section of the article. See Box 3.3 ter in a research thesis (usually Chapter
for an example of this. 2). In a journal article, it may merit its
A variety of phrases is used by re- own heading or be included in the intro-
searchers to indicate the research pur- ductory section of a report along with the
pose. Some common ones include: research problem, purpose statement,
and question. (Some qualitative research
The purpose of this study was to de- designs do not include the review of the
scribe three classication systems . . . literature at this stage of the research
process. These differences in approach places the existing study in the context of
are discussed in Chapter 6). prior research studies and current knowl-
edge. The goals of a literature review
include:
1. Importance of the Literature
Review Etablishing the signicance of the re-
search problem
The literature review should tell us what Identifying sample characteristics to
is known about the research topic and help identify relevant demographic
problem and what we need to nd out. It variables
78 GETTING STARTED
Identifying gaps in knowledge about Identifying variables that others have

the problem found to be relevant to the problem at
Identifying limitations of previous stud- hand
ies that may lead to a new study design Identifying areas that, if explored,
Identifying areas where there seems to could lead to important new under-
be consensus among researchers standings of the phenomenon under
Noting where there are inconsisten- examination
cies in research ndings Identifying theoretical frameworks that
others have used to study the phenom- begin by checking through the most
enon recent issue. After you have found an
Identifying how other researchers have article that is close to your topic in ei-
made connections to theory ther the particular variables treated or
Identifying what other researchers con- the general concepts presented, check
sider to be important to study it for references to other articles.
Identifying how other researchers have These references are usually helpful
measured variables and analyzed their sources to include in the literature re-
data view. By starting with the most recent
Exploring the methods other investiga- issues, you will identify the latest re-
tors have used to study the research search. If an older article is referenced
question frequently, you should read it because
it may be considered a classic.
In reviewing the literature, you are try-
ing to get a sense of the state of scientic
You may not be able to nd published
results on the specic relationship or
knowledge about the topic and determine
category of individuals you wish to
how best to study the research problem.
study. If this is the case, focus on vari-
ables that researchers have found to
be related to the major dependent
2. How to Conduct the variable that you propose to examine.
Literature Review In many cases, it is appropriate to re-
view research that uses similar method-
Many graduate students panic about the
ologies or theories.
idea of creating a state-of-the-art litera-
ture review on their research topic. This
In summarizing the research studies,
try to answer the questions Who?,
is because they try to do it all in one sit-
What?, Where?, and When? for each
ting. We suggest breaking the review into
study you include in your review.
three stages: identifying the literature
sources, evaluating the research studies,
Next make a statement about whether
the study was sound or not (Chapter
and writing an integrative review of the
20 provides guidelines for critiquing
ndings. Usually the rst question posed
studies). Use words that reect an
is, Where do I start? Here are some use-
evaluative component in your descrip-
ful steps to get you started.
tion of the study. For example, awed
If you are part of a university or col- design, robust ndings, seminal piece
lege, ask professors who work in the of work, and so on.
area for any ideas they may have on The research question may indicate
where to get information. that it would be appropriate to check
Check with a reference librarian for various nonprint sources for inclusion
sources that may lead you to research in the literature review. These sources
done on your topic. include digitized data, photographs,
Check textbooks for leads. audio tapes, and video tapes.
Check the appropriate disciplines ab-
stracts because they provide brief de-
scriptions of published papers dealing 3. How to Write the
with a variety of topics. Box 3.4 lists Literature Review
some of the relevant abstracts, indexes,
and Internet resources available. Generally, when the literature is analyzed
Check journals that are likely to pub- in a report, the authors try to briey sum-
lish work in the area. It is advisable to marize the areas of agreement and dis-
80 GETTING STARTED
BOX 3.4 Abstracts, Indexes of Periodicals, and Internet Resources of

Relevance to Nursing*
Database/Index Web Site

Nursing, Health Care, and Related Disciplines
Cumulative Index to Nursing and Allied Health http://www.cinahl.com

Literature
Index Medicus
MEDLINE CD-ROM (Medicine) http://igm.nlm.nih.gov
HealthSTAR (Health Administration coverage) http://igm.nlm.nih.gov
BIOETHICSLINE (Biomedical Ethics coverage) http://www.nlm.nih.gov
EMBASE (European-based medicine coverage) http://erlsevier.com
CANCERLIT (Cancer coverage) http://cancernet.nci.nih.gov
AIDSLINE (Aids and HIV coverage) http://igm.nlm.nih.gov
Sigma Theta Tau http://www.stti.iupui.edu
National Information Center on Health Services http://www.nlm.nih.gov/nichsr.html

Research and Health Care Technology
(NICHSR)
Psychology
Child Development Abstracts and

Bibliography
Psyclit CD-ROM
Psychological Abstracts
PsychINFO http://www.apa.org/psycinfo
Education
ERIC CD-ROM http://ericir.syr.edu/eric
Anthropology
Abstracts in Anthropology


Relevance to Nursing* (Continued )
Anthropological Literature: Index to Periodical Articles
International Bibliography of Social and Cultural

Anthropology
Sociology
Human Resources Abstracts
International Bibliography of Sociology
Social Work Research and Abstracts
SciSearch and Social SciSearch http://www.isinet.com
Sociological Abstracts
Sociole CD-ROM
General Listing (Relevance to Nursing and Other

Disciplines)
Canadian Periodical Index (Canadian content)
Hospital Literature Index
Infotrac (Academic Index) CD-ROM
Social Sciences Citation Index
Social Sciences Index
Public Affairs Information Science Bulletin (PAIS)
Newspaper Indexes
Canadian News Index
Canadian Press Newsles
Index de lactualite
The National Newspaper Index

82 GETTING STARTED

Relevance to Nursing* (Continued )
New York Times Index
The Times Index
Health-Related Internet Directories and Search

Engines
Health Web http://www.healthweb.org
Medscape http://www.medscape.com
National Library of Medicine http://www.nim.nih.gov
Health Finder http://www.healthnder.gov
Evidence-based Nursing http://www.bmjpg.com/data

ebnpp.html
Health Web http://www.healthweb.org
National Center for Health Statistics http://www.cdc.gov.nchswww
Yahoo-Health http://www.yahoo.com/health
AltaVista http://www.altavista.com
Infoseek http://www.infoseek.com
*With special acknowledgement to Barbara Phillips of the St. Francis Xavier Library staff who assisted with the
preparation of this listing.
agreement in the literature. A critical as- shows one method of summarizing arti-
pect of reviewing the literature is the au- cles: by using such a grid system, you can
thors interpretations of the various nd- make additions to the list, provide a
ings and their relationships to the quick summary of areas of agreement and
present study. Article summaries may be disagreement among researchers, inter-
useful to researchers but are generally pret earlier literature in the eld, and re-
not appropriate in a nal report. What is late earlier ndings to ones own work.
required is a description of the current In preparing a discussion of the litera-
state of knowledge on the topic under in- ture reviewed, it generally seems best to
vestigation. The researchers need to dis- prepare an outline of the major concepts
cuss the most signicant studies and or variables and the relationships of sig-
those that form the theoretical back- nicance to the current study. This al-
ground for the current study. Table 3.1 lows the researcher to insert material on
Table 3.1 Background Characteristics Associated with Smoking
DeFronzo & Willms & Grossarth- Aloise-Young Covey Trush Pugh Kaplan & Glendinning Sun & Rogers Oygard
Pawlak Stebbins Matieecek & Hennigan et al. et al. et al. Weiler et al. Shun et al. et al.
Variables (1993) (1991) et al (1988) (1996) (1992) (1997) (1991) (1997) (1997) (1995) (1995) (1995)
Family
members
smoke
Health
problems
reported
Age
Education
Urban
community
Violence
in family
Gender
(male)
DECIDING WHAT TO STUDY

Year of 1993 1991 1988 1996 1992 1997 1991 1997 1997 1995 1995 1995
report
Number of 849 NA 5977 1971 8042 2086 NA NA 627 1320 827 104
subjects
Place of University McMaster London St. Johns New York University London Illinois Scotland Florida Norway Colorado
study of Conn. University of Surrey
indicates variable was positively related to incidence of smoking.

indicates variable was negatively related to incidence of smoking.
Blank spaces indicate the variable was not reported in the study.
Adapted from Gillis, A. (1998). The trends and variations in smoking. Course paper for Nursing 300, Antigonish, Nova Scotia: St. Francis Xavier University.
83
84 GETTING STARTED
each report or study at an appropriate sources should be cited in the reference

point in the review. It is best to report the list. If you have not recorded this infor-
ndings on one variable at a time. mation in full, you will waste a lot of time
One suggestion for organizing the re- retracing your steps to recover the infor-
view is to divide the variables into clus- mation.
ters as they relate to the major de-
pendent variable in your study. For ex-
ample, you might have (1) background
C. SPECIFYING THE MODEL
characteristics of respondents, (2) situa-
tional variables, and (3) other factors.
Clarity and precision are at the core of
Thus, if you were studying factors associ-
successful research. A theoretical or
ated with success on national nursing
causal model is a graphic representation
registration examinations, you might be-
of proposed causal interconnections be-
gin by discussing background character-
tween variables. In quantitative research,
istics (e.g., socioeconomic status, size of
one way to enhance accuracy is to draw a
home community, gender). These might
diagram of the causal connections be-
be followed by situational factors such as
tween variables. The advantage of a dia-
family responsibilities or life crises that
gram is that it encourages clear thinking
other researchers have related to suc-
about what you are doing. Thus, drawing
cess on nursing examinations. Then you
a diagram with causal arrows requires
might consider general academic factors
you to indicate which variable appears
such as nursing course grades and read-
causally before other variables. You are
ing and language skills. In each case, you
forced to be specic. Such diagrams can
should summarize the consensus (or lack
also replace stating formal hypotheses;
of it) in the research literature. And if you
each hypothesis may be reected in a
have provided a summary grid for the lit-
properly drawn diagram. You may recall
erature, readers will be able to review the
from Chapter 1 that hypotheses are tenta-
ndings quickly, and the plus and minus
tive predictions about the relationships
symbols will quickly reveal the degree of
between two or more variables.
consensus on the relationship of any par-
ticular variable to the dependent vari-
able. Box 3.5 illustrates a literature re- 1. How to Identify the Variables
view statement.
The literature review should have solid As you will recall from Chapter 2, variables
informative (who?, what?, when?, and may be dened as constructs that we in-
how?), evaluative (e.g., weak, robust, tend to measure. Obviously, then, a re-
awed), and integrative components (e.g., searcher must identify the specic vari-
ndings are consistent, supportive of, con- ables that will be studied in the research
trary to others) that present a balanced project. Variables may be identied by
perspective of the current state of knowl- many methods. Most researchers have
edge on the research topic. It is not neces- some idea of the variables that might in-
sary to include every article published on uence the dependent variable, so it is
the research topic; in many cases, this easiest to begin with these. A review of the
would make the review too lengthy. How- research literature also helps you identify
ever, do make sure that seminal pieces of the variables that other researchers have
work and those that focus most closely on used in doing similar research. Studying
the research problem are included. relevant theoretical frameworks and mod-
Be certain to record full bibliographic els and identifying the variables that are
details of each article or book used. When implied by the theories are also useful in
the nal report is being prepared, all the identifying research variables. Examining
FACTORS ASSOCIATED WITH NCLEX-RN SUCCESS:

A LITERATURE REVIEW
There is a need for more extensive empirical her student population. Johnson (1989) con-
data describing variables that contribute to suc- ducted a study with an ethnically and linguisti-
cess on the NCLEX-RN. Until recently, the cally diverse population and found that language
only reliable factors identified as predictors of was a major variable predictive of success on
success were previous academic achievements, the NCLEX-RN. Poorman and Martin (1991)
particularly earned grade point average (GPA). found that test anxiety was inversely related to
However, other factors may be involved. passing the NCLEX-RN. . . .
Academic Nursing Factors Nonacademic Factors
Glick, McClelland, and Yang (1986) found sig- Few studies have considered the influence of
nificant simple correlations between grades in nonacademic factors on NCLEX-RN perfor-
clinical nursing courses, cumulative nursing mance. Given the increasing number of minor-
GPA, and NCLEX-RN performance. Froman ity students, second-career students, adult
and Owen (1989) found that GPA in theoreti- learners, and students from diverse cultural
cal nursing courses was predictive of perfor- backgrounds attending nursing programs, it
mance on NCLEX-RN. Krupa et al. (1988) seems appropriate to identify the nonacademic
looked at grades received in the Introductory factors that are important in determining stu-
Nursing course and the Medical-Surgical Nurs- dents at risk of failing NCLEX-RN. Dell and Va-
ing course. Their conclusion was that grades line (1990) examined self-esteem influence on
from both courses were substantially and di- NCLEX-RN scores. Self-esteem and age did
rectly related to NCLEX-RN performance. contribute to the variance in NCLEX-RN
Grades in the other practicum courses were scores. Self-efficacy recently has been found to
not good indicators in their study. Jenks et al. be an important predictor of NCLEX-RN per-
(1989) found that grades in nursing theory formance (Owen & Froman, 1990). Tolland
courses correlated strongly with NCLEX-RN (1989) suggested that highly complex interrela-
performance. . . . tionships between intellectual ability and
nonacademic variables may determine whether
General Academic Factors students will succeed. . . .
Some authors have investigated the influence of In summary, it appears that both academic
general academic factors such as reading com- and nonacademic factors are influential in pre-
petency and language skills on NCLEX-RN suc- dicting performance on the NCLEX-RN. . . .
cess. Wolahan (1992) found that reading com- The current study will try to identify both aca-
petency at the 12th grade level and math skills demic as well as nonacademic variables associ-
were indicators of NCLEX-RN success among ated with NCLEX-RN success.
SOURCE: Summarized from Arathuzik, D., and Aber, C. (1998). Factors associated with national council licensure
examination: Registered nurse success. Journal of Professional Nursing, 14(2), 119126.
questionnaires for ideas concerning the 2. How to Develop Theoretical

variables that may be measured if a survey or Causal Models
is to be conducted may yield interesting
results. Finally, developing your own the- We will begin with the simplest models
oretical models or frameworks and gur- and gradually move toward the more
ing out which sources of spuriousness, in- complex ones. What all models have in
tervening, or control variables may be common is that a diagram can represent
relevant for the study is an exciting means each of them; each such diagram shows
of identifying variables for investigation. causal direction and the implied hypothe-
86 GETTING STARTED
ses. The hypothesis or set of hypotheses a. Two-Variable Models

translates the problem statement into a
precise statement that predicts the ex- In Chapter 1, the distinction between a
pected outcomes of the study. In this dependent and an independent variable
sense, the research hypothesis is a tenta- was made. The dependent variable is the
tive answer to the research problem or effect in a cause-and-effect relationship
question. The hypothesis is then empiri- or is the result of the inuence of an in-
cally tested through a process of data col- dependent variable. Conventionally, we
lection and analysis. We begin with the refer to the dependent variable as the Y
two-variable model. Chapter 17 and Ap- variable and the independent variable as
pendix A present procedures for analyz- the X variable. The relationship can be
ing these models. described simply with a diagram:
> X > Y
What does the diagram tell us? First, it represent the relationship is that it en-
describes a relationship between two ables the researcher to:
variables known as X and Y. Note that an
arrow points from X to Y; this tells us that
Indicate causal direction (the arrow
points at the dependent variable); and,
X is the independent variable and that Y
is the dependent variable. The hypothe-
Indicate if the relationship is positive
(note use of symbol) or negative
sis reected by this diagram argues that
( symbol).
X inuences Y. Next, note the symbol
before the X box. This symbol means Drawing a picture increases precision.
the greater; the opposite symbol, , There may be times when you cannot for-
means the less. Putting all these ele- mulate a problem in such a way that you
ments together, the hypothesis can be are able to say whether the relation is pos-
stated as: The greater X, leads to the itive or negative. In these cases, one can-
greater Y. If we wished to express a neg- not speak in greater than or less than
ative relationship, the rst symbol could terms. When you are unable to specify the
be reversed. In this case, the hypothesis nature of the relationship, use a ? to in-
would be stated as: The less X, leads to dicate that no prediction is being made.
the greater Y. Occasionally, you will not be in a posi-
If the X and Y were replaced with vari- tion to set out a causal order; in these
able names, we might, for example, be talk- cases, you may be faced with a situation
ing about the relationship between par- in which variables exist together and in-
ticipation in health-promoting lifestyles uence one another simultaneously. In
and self-efcacy, suggesting that greater this event, you can indicate reciprocal
perceived self-efcacy leads to greater par- causation by placing arrows at each end
ticipation in health-promoting lifestyles of the line linking the two variables.
(Gillis, 1994). The wording of the hypothe- Ideally, hypotheses are deduced from
sis indicates that perceived self-efcacy is theories, previous research, or logical rea-
the independent variable and participa- soning. In new areas of investigation, it
tion in health-promoting behavior is the may be difcult to develop a causal model
dependent variable. because there simply is insufcient evi-
The advantage of drawing a diagram to dence to support the development of ex-
planatory hypotheses. Many descriptive ( i ) An Intervening Variable Model

and qualitative investigations (see Chap- An intervening variable (I) is a variable
ters 6 and 7) proceed without a hypothesis that links an independent variable (X) to
statement or a causal model. Such investi- a dependent variable (Y). An intervening
gations may result in theory building or are variable represents an explanation of
precursors to theory generation. how the independent variable influ-
ences the dependent variable. The in-
b. Three-Variable Models terest here is in understanding the rela-
tionship between X and Ythat is,
Now let us turn our attention to various understanding the mechanism by which
three-variable models. Some terms need X is connected to Y. Frequently, a re-
to be reviewed. Besides independent and searcher is testing a number of alterna-
dependent variables, there are three ad- tive explanations of how X influences Y.
ditional types of variables that will need In the case of one intervening variable,
to be understood: intervening, source of the relationship could be diagrammed
spuriousness, and conditional variables. as follows:
> X > I > Y

In this diagram, I is the intervening inuences Y. Typically, one would pro-
variable, or the linking variable between pose a number of possible intervening
X and Y. The hypothesis is that variations variables, so the following diagram would
in X cause variations in I, which, in turn, be more appropriate:
X Y
I1
I2
I3
88 GETTING STARTED
In this diagram, three alternative ex- nection between X and Y. These matters
planations are suggested for the connec- are examined further in Chapter 17.
tion between X and Y. If the X, Y, and I
were replaced with variable names, we ( ii ) A Source of Spuriousness Model
might, for example, test the relationship A source of spuriousness model is one in
between previous health-care experi- which a variable is identied as a possible
ences (X ), utilization of health-care ser- inuence on both the independent vari-
vices (Y ), and cognitive appraisal (I1), in- able (X) and the dependent variable (Y) in
trinsic motivation (I2 ), and perceived such a way that it accounts for the rela-
health status (I3 ), suggesting that the tionship between them. In other words,
more positive ones previous health-care the relationship between X and Y may be
experiences, the more positive ones cog- spurious because it is produced by the
nitive assessment of the experiences, inuence of S/S on each of them. Here the
leading to greater utilization of health ser- researcher proposes that, although there
vices. Similarly, a second alternative exis a statistically signicant relationship
planation would be that previous positive between the variables X and Y, this rela-
health-care experiences inuenced the in- tionship may be noncausalthat is, only
tervening variable of internal motivation, existing because some third variable is in-
leading respondents to seek health-care uencing both X and Y. Having observed
services. The researcher would collect a statistically signicant relationship, the
data that measure each of the variables in- researcher will want to ensure that the re-
volved and conduct the appropriate sta- lationship is not spurious and, therefore,
tistical tests to determine, which, if any, of should run a number of spuriousness
the proposed alternative explanations or checks. The source of spuriousness
intervening variables explains the con- model may be diagrammed as follows:
X Y
S/S
In the source of spuriousness model, tween socioeconomic status (X) and

the researcher is suggesting that the re- choice of alternative therapies (Y) in
lationship between X and Y may be spu- women with breast cancer. We might
rious. If the variables X and Y were reconsider the possibility that rural or ur-
placed with variable names, we might, ban background is a source of spurious-
for example explore the relationship be- ness (S/S). Possibly the type of commu-
nity the women come from inuences causes variation in an independent vari-
their socioeconomic status as well as able that, in turn, inuences the depen-
inuencing their choice of alternative dent variable in the model. Thus, the an-
therapies. The techniques for examining tecedent variable is one that precedes
such relationships are examined in Chap- the main independent and dependent
ter 17. variables. This variable may be having an
impact on the independent variable that,
( iii ) An Antecedent Variable Model in turn, may be inuencing the dependent
An antecedent variable model is a causal variable. An antecedent variable may be
model that proposes a variable that diagrammed as follows:
> A > X > Y

If the A, X, and Y were replaced with pendent variable. Here the researcher is
variable names, we might, for example, proposing a number of independent vari-
test the relationships among perceived ables that may be inuencing the depen-
low job autonomy (A), emotional stress dent variable. This type of model is illus-
(X), and smoking behavior (Y), suggesting trated in Figure 3.1.
that the greater ones perception of low The variables on the left side of the dia-
job autonomy, the greater ones level of gram are the independent variables and
emotional stress, leading to an increase in are viewed as potential causes of varia-
the frequency of smoking behavior. tions in the dependent variable, frequency
In a sense, an antecedent variable com- of smoking behavior (Gillis, 1998). Note
bines one idea from the source of spuri- that the model uses the symbols and
ousness model and one from an interven- to indicate whether the independent vari-
ing variable model: ables are positively or negatively associ-
ated with the dependent variable. The in-
It is causally prior to both the indepen-
dependent variables are related to the
dent and the dependent variable (as in
dependent variable either one at a time or
a source of spuriousness model).
simultaneously, through procedures that
It converts the independent variable
are outlined in Chapter 18.
into one that intervenes between the
antecedent variable and the depen- ( ii ) Path Models
dent variable. A path model is a graphic representation
of a complex set of proposed interrela-
tionships among variables. A model of
c. Multivariate Models this type is shown in Figure 3.2 (Israel and
Shurman, 1990).
Models that use numerous variables are The following hypotheses represented
known as multivariate models. The rst in Figure 3.2. may be tested empirically:
we will consider is the candidate variable
model.
Objective conditions conducive to
stress have a direct relationship to
( i ) Candidate Variable Model health outcomes (arrow C).
A candidate variable model is one that Variations in stressors cause variations
proposes several independent variables in perceived stress, which inuences
as possible causes of variation in a de- health outcomes (arrows A and B).
90 GETTING STARTED
Perceived stress directly affects health D. SPECIFYING HYPOTHESES

outcomes (arrow B). AND PROCEDURES OF
Supportive relationships and the exer- ANALYSIS
cise of control can directly reduce ob-
We state hypotheses (or diagram them)
jective stressors (arrow 1) and the
before data analysis to ensure that the re-
perception of stressors as stressful
searcher is not tempted to invent ones af-
(arrow 2) and directly affect health
ter the data have been analyzed that will
outcomes (arrow 3).
conform to expectations. The specied
Supportive relationships and the exer-
hypotheses represented in the model in
cise of control can inuence the effects
Figure 3.2. are examples of hypotheses
of stressors (arrow 1a) and perceived
that can be tested empirically. After the
stress (arrow 2a) on health outcomes.
hypotheses are specied and diagrams of
All the models discussed reect the the hypotheses illustrated, it is appropri-
causal thinking of the researcher in such ate to specify the analysis procedures
a way that this reasoning can then be that will be used to test whether or not the
tested in later stages of the research proj- data collected support the hypotheses.
ect. (See also Box 2.7, which features a Hypothesis testing is central to scien-
path model of inuence on the initiation tic research. After you have stated the
of sexual intercourse.) hypotheses and selected your design,
you can proceed to identify your sample, potheses than to present a written version
select your data collection instruments, of each one. Similarly, when many alterna-
gather your data, and analyze the results. tive explanations are being tested for a par-
Specically, your hypotheses are tested ticular relationship, a good diagram clearly
during the data analysis stage through shows the causal model and the implied re-
statistical procedures. search hypotheses.
In theory-testing projects, it is impor-
1. How to Specify Hypotheses tant not only to state the derived hy-
potheses but also to indicate the steps
Good diagrams can replace formal hy- that were taken in making the derivation
potheses statements. By using and or derivations. You may recall from Chap-
symbols, it is possible to indicate greater ter 2 that hypotheses in a theory-testing
than and less than relationships; arrows study are derived from the propositional
may be used to indicate causal direction. statements in the theory. Only to the ex-
As noted in the hypotheses statements tent that such derivations can be traced
associated with Figure 3.2, a variety of can one claim to have tested a theory. For
terms are used to express the relationship example, from the theoretical proposition
between the dependent and independent that social support is related to health
variables in a hypothesis statement. Com- outcomes (see Fig. 3.2), the researcher
mon expressions include varies with, can derive the hypothesis that adoles-
positively inuences, negatively inu- cents with positive peer support will per-
ences, direct relationship to, inversely ceive their health status more positively
related, and reciprocal association than adolescents with little peer support.
with. Particularly in candidate variable Hypotheses are statements specic to an
models, which include a large number of empirical situation. In this example, the
variables, it is easier to diagram the hy- hypothesis relating peer support to per-
92 GETTING STARTED
ceived health status could be tested em- or blueprints for woodworking. These
pirically to establish the validity of the patterns are directions or guidelines that
claim. As evidence accumulates to sup- help you create beautiful masterpieces.
port the proposition that social support is This section introduces you to a pattern
related to health, the worthiness of the for designing a research project. A re-
theory is established. Remember that one search design is like a blueprint for a
does not prove the hypotheses or the the- study. It guides the investigator in plan-
ory through hypotheses testing. Conclu- ning and implementing the study. It pro-
sions are always considered tentative; vides a detailed plan for data collection
however, with increasing research evi- and analysis and is the critical element
dence to support conclusions, hypothe- linking the theoretical framework and
ses come to be accepted or believed. questions with the resultant data. The
In stating hypotheses, one should note term research design is frequently used in-
that although there are two types of terchangeably with the term methodology.
hypothesesthe research and the null If you think about the research design like
hypothesesthe researcher is primarily a blueprint for a house, it quickly will be
concerned with the research hypothesis, apparent how critical the design is to the
which states the expected outcome of the successful implementation of the study.
research. The distinction between the The blueprint indicates to the contractor,
two is discussed in Chapter 12. the carpenters, and other tradesmen
what parts of the house should be con-
2. How to Specify Methods structed rst, and how and what pieces of
of Analysis materials go together to construct the
shell and then the walls, and so on. Simi-
Specifying methods of analysis in advance larly, research designs indicate the steps
forces the researcher to be committed to and actions the researcher should take
particular procedures. Selecting these pro- and in what order they should be done to
cedures also has implications for the way carry out the very best investigation of a
in which variables will be measured. Thus, phenomenon and arrive at an answer to
if cross-tabulation tables are to be used the research question. There are no rigid
exclusively, it will not be necessary to get rules for selecting a research design;
ratio level measures on the variables. It is rather, the researcher is guided by exam-
probably unreasonable to require a re- ining the research question, the purpose
searcher to specify cut points for cross- of the research, and his or her own philo-
tabulation tables because of the large num- sophical approach to inquiry.
ber of variables that may be involved (see You may recall from Chapter 1 that dif-
Chap. 16). However, it is reasonable to in- ferent types of research (e.g., descriptive,
dicate the number of categories that will explanatory, predictive) exist. Given differ-
be used and the principle that was used in ent types of research and different re-
making the cut points (perhaps by splitting search questions, it is necessary to have a
the sample into thirds or at the midpoint). variety of research designs. If you decide
to explore a phenomenon that is not well
understood and has had little empirical in-
E. UNDERSTANDING vestigation, you will most likely use a de-
RESEARCH DESIGN scriptive research design. For example,
the question, What is the experience of
As a creative research student, you have caring for a family member with Alz-
no doubt had some experience following heimers disease? implies a description of
patterns for sewing, recipes for cooking, the experience that family members go
through in the process of giving care to A description of the study participants

their loved one. A descriptive design such or the sample. Most nursing studies in-
as one of the qualitative approaches dis- clude individuals, groups, families, or
cussed in Chapter 6 may be appropriate to communities as the elements of the
answer this query. The question, What is sample (see Chapter 15).
the relationship between the levels of fa- A description of the setting. Some re-
tigue in family caregivers of clients with search designs are distinguished by the
Alzheimers disease and available social natural setting in which the research is
support? implies a study of the relation- conducted. The investigator does not
ships between fatigue in caregivers and so- alter the conditions of the environment
cial support. A correlational survey design in any way during the data collection
(see Chapter 5) may be appropriate in this procedures (see Chapter 7). Other de-
example if fatigue and social support have signs require the investigator to manip-
been well described in the literature but ulate the environment in such a manner
the nature of the relationship between the as to control all of the variables except
two is not well established. the one major variable whose inuence
If the research phenomenon of interest the researcher is interested in measur-
has been well described in the literature ing (see Chapter 4). This latter type of
and its relationship to other factors is well articial environment is usually created
known, then an experimental design (see in a laboratory setting.
Chapter 4) may be appropriate. For exam- A description of the variables. Vari-
ple, the question, What is the effect of a ables are the main concepts the inves-
respite care program on the incidence of tigator intends to measure empirically.
depression in family caregivers of patients Research designs may be univariate
with Alzheimers disease? implies an ex- (one variable) or multivariate (two or
perimental design. Here the researcher more variables). Most nursing investi-
would be interested in testing the impact gations are multivariate, reecting the
of the respite care program on the inci- complexity of nursing practice (see
dence of depression in family caregivers Chapter 2).
who receive respite and comparing them A description of the data collection
with a group of family caregivers who re- methods. The data collection methods
ceive the usual care with no respite pro- may include the use of self-report data
gram. Chapters 4 through 8 discuss a collected by means of researcher de-
range of quantitative and qualitative developed instruments, questionnaires,
signs useful to addressing a variety of re- standardized instruments, or scales;
search questions. Each chapter discusses structured, semi-structured, or non-
in detail the components of a good re- structured interview guides; partici-
search design. As you read each chapter, pant or nonparticipant observations;
keep in mind your research question. This or biophysiologic measures. In some
will help focus your reading and allow you research designs, the data collection
to evaluate the designs relationship to the plan may also include the use of exist-
research question you wish to answer. ing records or data for the purposes of
a secondary analysis (see Chapter 13).
1. How to Describe a A description of the time dimension
Studys Design of the design. The time element is cru-
cial to the design in describing the fre-
Six basic components should be included quency (how often) and order (when)
in describing a studys design. These in- in which data collection will be con-
clude: ducted. If change over time is an im-
94 GETTING STARTED
portant component of the study de- participants, availability of research assis-

sign, it is critical to indicate the timing tants, and constraints imposed by ethical
of data collection procedures to cap- issues, to name a few. Pragmatic consider-
ture the rapid or slow patterns of ations should not supercede the primary
change. If data are collected at one consideration, which is the research ques-
point in time, it is referred to as cross- tion posed, but they do inuence the de-
sectional; if data are collected at dif- sign decision to a limited extent. In addi-
ferent intervals over a period of time, it tion to the research question posed, the
is referred to as longitudinal; if data following factors may be considered in se-
are collected from the present to some lecting the design:
future point in time, it is referred to as
prospective; and if data collection re- a. Purpose and Level of Knowledge
ects measures that occurred in the
past, it is referred to as retrospective. The research design needs to be consis-
A description of the role of the investi- tent with the research purpose and ques-
gator. The role of the investigator varies tion. For example, if a health-care facility
with the research design selected. In wants to evaluate the effectiveness of a
some designs, the investigator becomes school-based nurse practitioner clinic, it
an active participant in the research should try to produce scientic evidence
project; in other designs, the researcher that the clinic produces positive out-
actively manipulates the variables becomes, such as less substance abuse,
ing studied and assigns participants to fewer missed days because of illness,
different research groups or treatment lower incidence of teen pregnancy, fewer
conditions; and in still other designs, sexually transmitted diseases. The pre-
the researcher remains aloof and de- ferred design to get this information is a
tached from the variables being studied quasi-experiment using a school district
(see Chapters 4 to 8). with a nurse-run clinic and a matched con-
trol group school district without a clinic.
In describing the research design, it is
Data could be collected from the two
important to provide a detailed descrip-
groups and compared on the various out-
tion of each component, in much the same
come measures. On the other hand, if the
way as an architect provides a detailed
purpose was to assess student satisfaction
sketch of the blueprints for a house. This is
with services at the clinic, a descriptive
important in enabling other investigators
survey design may be more appropriate.
to replicate your study and determine un-
The level of knowledge about the topic
der what conditions the results of the in-
is closely related to the purpose and re-
vestigation are applicable in their setting.
search question. If little is known about a
topic, a descriptive or exploratory study
2. What Factors Guide Selection is appropriate. On the other hand, if con-
of a Study Design? siderable information is known about a
phenomenon and the factors that are re-
Many factors inuence the choice of a related to it, an explanatory or predictive
search design. Primarily, it is the nature of study using an experimental design may
the research question asked that deter- be appropriate.
mines which design is most appropriate.
One should note, however, that there are b. Nature of Research Phenomenon
also pragmatic considerations to keep in
mind such as the amount of time and re- Some research phenomena can only be
search funds available, access to research studied in natural settings, requiring the
use of eld study approaches. For exam- lect a large sample of women from two
ple, the immediate effect on families of rural communities to be interviewed
mining disasters must be investigated in about their birthing experiences at two
the settings where they occur. Therefore, a different health-care facilities. However,
researcher interested in investigating this you determine that it would not be possi-
phenomenon would select a design appro- ble to randomly assign women to a partic-
priate to the naturalistic setting, such as ular institution to receive care because of
one of those outlined in Chapters 6 and 7. their health insurance policies. Although
it may be desirable to use random assign-
ment, it is not possible in this situation be-
c. Ethical Considerations
cause of the context.
Some research questions cannot be ad-
dressed in humans because of ethical con-
siderations. For example, to deny clients f. Personal Preference
access to an available treatment for a par-
ticular disease would be unethical. In other The individual preference of the re-
cases, one may wish to study patterns of searcher needs to be considered in se-
illness or disease over time. It would be lecting a design because every project re-
unethical to induce illness in humans for quires time, energy, and intellectual
the purposes of research; therefore, the in- commitment on the part of the investiga-
vestigator must nd participants in which tor. For this reason, the researchers
the disease is occurring naturally. should consider his or her personal pref-
erences in selecting a design. For exam-
ple, if a researcher likes to explore con-
d. Practical Considerations cepts or make meaning out of peoples
experiences, then one of the qualitative
Issues such as researcher time, costs, designs such as phenomenology may be
available resources to conduct a study, appropriate. If, on the other hand, a re-
and available data may strongly inuence searcher enjoys studying relationships,
the selection of the research design. For determine cause-and-effect scenarios, or
example, if a study is limited by the redesign and evaluate interventions, then a
searchers time or is time sensitive be- quantitative approach such as compara-
cause of grant deadlines, it may be more tive surveys or experimental research
feasible to use ready-made instruments may be more appropriate.
and conduct a survey that can be group
administered to the respondents in one
sitting rather than select a design that re-
3. How to Choose
quires in-depth interviews. Similarly, if the
a Research Design
only data available to answer a research
question are past health records, then the Selection of the most appropriate re-
research design necessitates a secondary search design is a critical step in the re-
data analysis approach (see Chapter 5). search process. Now that you know what
the components of a design are and
e. Feasibility which factors to consider in selecting a
study design, how do you actually decide
Sometimes a particular design is the best what is the best design to answer your re-
one to answer the research question but it search question? The following summa-
is not feasible to use it. For example, you rizes some steps to help you decide
may decide that it is best to randomly se- which research strategy will be followed.
96 GETTING STARTED
Step 1. What have others done? Do a pre- process that texts on research meth-
liminary review of the literature to get ods present it to be.
some sense of what has been written As we have recommended previously,
about the topic that you wish to ex- the research design should match the re-
plore. As you read the literature, pay search question and purpose of the
attention to the research designs that study. To reiterate, research is generally
were used in the studies. Does the au- conducted for one or more of the follow-
thor specify a rationale for the design ing purposes: (1) to identify and describe
selected? Can you locate any articles concepts and phenomena of interest to
that reect the use of different designs nursing (descriptive surveys and qualita-
to explore the problem? tive designs); (2) to explore differences
Step 2. Review the chapter in this text and associations among variables (ex-
that introduces the design you have ploratory designs such as correlational
tentatively selected. The following and comparative studies); and (3) to ex-
chapters introduce different types of plain, predict, or control nursing phe-
research designs: nomena (experiments, case studies, and
Chapter 4: Experimental and Quasi- evaluation research).
Experimental Designs (randomized Flexibility and openness are important
clinical trials) to good research design. This point is im-
Chapter 5: Survey Designs (surveys, portant to remember as the study un-
secondary data, comparative stud- folds. Occasionally, problems arise that
ies, meta-analyses) are beyond the control of the investiga-
Chapter 6: Qualitative Designs (phe- tor. Experienced researchers will try to
nomenology,ethnography,grounded anticipate potential problems and take
theory) action to prevent them or solve them af-
Chapter 7: Field Study Approaches ter they occur. The following section pre-
(participant observation, in-depth sents solutions to common problems
interviews, focus groups, eld ex- that may emerge during the process of a
periments, naturalistic observa- research investigation.
tional studies)
Chapter 8: New Wave Contemporary
Applied Approaches (evaluation re-
search, action research, health pro- F. PREDICTABLE SETBACKS
motion research, feminist research) AND CHALLENGES
Recommended readings in each of the
chapters may be helpful in gaining ad- Few research investigations proceed
ditional insights for the design of your smoothly. Even after painstakingly evalu-
study. Students considering a qualita- ating your design options and carefully
tive approach should review Chapters selecting one, setbacks and problems will
6 and 7 in detail. emerge. Hence, it is prudent to prepare
Step 3. Be prepared to reconsider the re- yourself for some delays and challenges
search design used. As you review the along the way to answering your research
literature in detail, different questions question. Seasoned researchers always
may come to seem more relevant to consider alternatives when faced with
the purpose and goals of your project. difculties such as those listed below:
If this happens, be prepared to recon-
sider the type of research design to be Time (Rats, the project report is due
used in the study. Remember that rein 3 months!). Projects always take
search is never the clean and tidy longer than anticipated, so experi-
enced researchers will plan for delays be consulted and invited to join the
and unpredictable events in the time- research team as members or consul-
line of the research project. For exam- tants to the project. For example, a
ple, extra time should be allotted for statistician and a community health
subject recruitment, administration of nurse may be welcome additions if you
instruments or an intervention, and are studying patterns of communica-
follow-up measurement after interven- ble diseases in low-income families.
tion. It is always wise to allow extra Availability of research assistants
time at the end of a project to deal with (To do this project properly, we
any unexpected events that may have would need assistance and there are
caused a delay in the project. If it is not not enough qualied people available
needed, the researcher can relax and to do the interviewing and observa-
enjoy the extra time available to com- tions.) You may need to advertise
plete the nal report. widely and train your own research as-
Ethical issues (The Ethics Review sistants if the budget permits.
Board may not approve of efforts to in- Funding (We just dont have enough
crease recruitment into the study.) If money to do the project properly.)
the risk to participants is too high and Research expenses may include print-
appropriate steps cannot be taken ing costs, mailing, telephone or travel
to protect the participants, maximize costs, space and equipment rental,
benets, and minimize harm, the study salaries of research assistants, com-
may have to be redesigned (see Chap- puter time, and fees to participants.
ter 10). Researchers typically cast a broad net
Computer resources (Our institution in seeking research funds from gov-
does not have the programs that we ernment, private, philanthropic, and
need to do the analysis.) The feasibil- business sources.
ity of a project often depends on
Given difculties such as these, are
the availability of facilities and equip-
there other general alternative strategies
ment such as computer access, video
that might be used? It may be possible to:
cameras, recording devices, observa-
tion rooms with one-way mirrors, and Use data that have been generated by
so on. If resources are not available other researchers. You will need to get
within your institution, it may be pos- the material and permission to use it,
sible to borrow, rent, or have equip- and you will also have to deal with
ment donated from hospital research someone elses operationalizations.
units or university research centers. Use publicly available data. For ex-
This may also be an opportunity to col- ample, (1) university and college re-
laborate with researchers in centers search departments often maintain
that can provide the resources re- data banks; (2) government agencies
quired for your project. sometimes make data of interest avail-
Technical competence (Regression able to researchers; and (3) other pub-
analysis is beyond me, and to replicate licly available data sources exist such
the study, I would need to understand as the Human Relations Area Files (see
it.) Feasibility of a project depends on Chapter 6). Typically, data sets are
appropriate expertise in terms of re- available in the form of computer les.
search skill and knowledge of the sub- Use information published in hospital
stantive area under investigation. databases, medical insurance les,
When a researcher has deciencies newspapers, or books. Much seondary
in either of these areas, experts may data can be gleaned from regularly pub-
98 GETTING STARTED
lished information. For example, if you form outlined in the Publication Manual of
were interested in examining the rela- the American Psychological Association.
tionship between childhood immuniza- This manual is used by most nursing re-
tion patterns and childhood disabili- search journals such as Nursing Research,
ties, you could contact the National Canadian Journal of Nursing Research, Re-
Center for Health Statistics (NCHS) in search in Nursing and Health, and Image:
the United States; in Canada, contact The Journal of Nursing Scholarship.
Statistics Canada to get data appropri- For eld studies and other qualitative
ate to these variables. designs, in which the project usually
takes shape in process, proposals can
Finally, remember when you think you
discuss the rationale behind the study
have exhausted all research possibili-
and the relevance of the location chosen.
tiesyou havent! Persistence and deter-
In using Table 3.2, researchers are en-
mination are invaluable assets for re-
couraged to make realistic estimates of
searchers.
the amount of time it will take to com-
plete various elements. Experienced re-
searchers know that unanticipated de-
G. DETERMINING YOUR lays do often occur and will plan for them
READINESS TO START THE by adding in some extra time. The more
PROJECT people involved in the project (either
subjects, respondents, or informants),
Table 3.2 provides a checklist of items that the greater the number of delays that
need to be attended to before beginning a may be anticipated. But even projects
project. Not all items will be relevant to that involve few others (such as a content
your study, but it is worth going through analysis of nursing journals) may be de-
the list to see if all the things that should layed because getting some of the material
be done before beginning your project on interlibrary loan may take longer than
have been completed. As previously dis- anticipated. Try to avoid delays, but it is a
cussed, research projects frequently get good idea to assume that they will occur.
behind schedule for all sorts of unantici- How do you know when you are ready
pated reasons. The checklist may help an- to start the data collection phase of your
ticipate some of the problems. research? A rule to follow is: You are ready
The rst section of the checklist deals to commence data collection when all the
with the development of a research pro- relevant items noted in Table 3.2 have been
posal. Researchers preparing theses will realistically planned and when all the nec-
have been told by their supervisor that it essary written materials, permissions, in-
is possible to develop a draft of the rst struments, and equipment are in place.
three chapters of a thesis before any data There are several products available to
are collected. These chapters would in- assist researchers in managing projects.
clude: These range from those that assist in
maintaining a le of bibliographic infor-
A problem statement
mation to those that provide assistance in
A literature review
the overall management of projects. Stu-
A methodology section
dents may wish to examine the free
The methods section is usually divided demonstration program for Reference
into subsections that deal with the partic- Manager available from Research Infor-
ipants, the instruments or measurement, mation Systems (www:http://www.risinc.
and the procedures. Most nursing recom). This program prints indexed bibli-
search proposals and reports follow the ographies and imports bibliographic in-
Table 3.2 Project Initiation Checklist

Not Initiated Target Completed
Items to Check Relevant Date Date Date
Preparation of Research Proposal [] ________ ________ ________
Statement of problem [] ________ ________ ________
Literature review complete [] ________ ________ ________
Methodology statement [] ________ ________ ________
Written formal hypotheses [] ________ ________ ________
Ethics review committee submission [] ________ ________ ________
Funding application [] ________ ________ ________
Permissions
From subjects or guardians [] ________ ________ ________
For use of copyrighted material [] ________ ________ ________
For entry into country or group [] ________ ________ ________
For ofce or laboratory space [] ________ ________ ________
Project staff hired [] ________ ________ ________
Subjects [] ________ ________ ________
Method of contact established [] ________ ________ ________
Instruments Completed
Letters to respondents and others [] ________ ________ ________
Questionnaires [] ________ ________ ________
Pretesting [] ________ ________ ________
Pilot study [] ________ ________ ________
Recording forms [] ________ ________ ________
Other (list) [] ________ ________ ________
Equipment
Tape recorders [] ________ ________ ________
Computers and programs [] ________ ________ ________
Other (list) [] ________ ________ ________
Sampling procedures determined [] ________ ________ ________
Scheduling, Provision for
Training of staff [] ________ ________ ________
Holidays, bad weather days [] ________ ________ ________
Data collection period [] ________ ________ ________
Time to get last few cases [] ________ ________ ________
Data entry time [] ________ ________ ________
Data analysis time [] ________ ________ ________
Report writing time [] ________ ________ ________
100 GETTING STARTED
formation available from online and from inexpensive program provides basic tools
CD-ROM databases, such as CINAHL and for the overall management of research
MEDLINE. Among its many features, it ed- projects. A product specically designed
its style changes to references. Another for health-related projects is from the Na-
product that may prove a good starting tional Health Data Systems, Inc., known as
point in the search is Microsoft Project, Hii&trade. Information for this product is
available from Microsoft. This relatively available at www.nhds.com.
E X E R C I S E S
1. List nine sources of nursing research 7. Complete the relevant sections of
problems. Which factors would you Table 3.2 for a project that you wish to
consider in selecting a research prob- conduct.
lem for investigation?
2. Choose a dependent variable that is of

8. Write a problem statement including a
interest to you and draw a diagram
statement of purpose and a research
showing the variables that may be in-
question for each of the following re-
uencing variations in the dependent
search topics:
variable.
Homelessness in the elderly
3. Drawing on propositions of a nursing or
health behavior theory, derive a Osteoporosis in middle-aged women
testable hypothesis and draw a diagram
of the proposed relationship. You may Motherinfant bonding
combine axiomatic and replacement of
terms approaches in your derivations.
Try for a counterintuitive prediction. Tobacco use in adolescents
9. Write a research hypothesis state-

4. Propose and diagram a series of at
ment using the following terms:
least three alternative explanations for
greater than, less than, positively
a proposed relationship.
related, and negatively affects to
communicate the relationship be-
5. Choose an applied nursing problem tween two or more variables.
and diagram relationships that you
would explore to solve the applied
10. Choose a research question of inter-
problem.
est and write a brief review of the lit-
erature using variables identied in
6. Diagram the relationship (or relation- the research question. To complete
ships) that would be explored if you this assignment, do the following:
were to replicate an existing study and Identify two indexes used to identify
indicate and diagram one additional literature sources.
relationship you would explore in the
replication.
Identify three publications from Write a summary statement at the

each index appropriate to the re- end of the literature review.
search question.
Include articles with conicting re-

sults.
Brown, S.J. (1999). Knowledge for Health Care Martin, D.W. (1991). Doing Psychology Experi-
Practice: A Guide to Using Research Evi- ments (3rd ed). Pacic Grove: Brooks/Cole.
dence. Toronto: W.B. Saunders. This text Martin includes a useful chapter on know-
covers a broad spectrum of strategies for ing when you are ready to commence an
searching the literature and identifying and experiment.
relating relevant research evidence to clin- Sparkes, S., and Rizzoloo, M. (1998). World
ical questions. wide web search tools. Image: Journal of
Edwards, M. (1998). The Internet for Nurses and Nursing Scholarship, 30(2), 167171. A good
Allied Health Professionals. New York: overview of web sources to consider in a
Springer-Verlag. This is a useful guide for review of the current state of knowledge re-
searching nursing databases and informa- lated to a research topic.
tion sources that may be useful in com-
pleting a literature review.

Part 2
RESEARCH
DESIGNS AND
APPROACHES
P
art 2 introduces various research designs. Although there is
no consensus in the research community on the best way to
classify research designs, for purposes of simplicity, this text
will classify them into ve categories: experimental ( including
quasi-experimental), survey, qualitative designs, eld studies, and
new wave contemporary applied approaches.
Part 2 provides beginning researchers with insights into the major

research designs used in nursing investigations. As you read this sec-
tion, you will notice that designs are not always mutually exclusive and
each has advantages and limitations that inuence the researchers
choice. It is rarely possible to select the perfect design for your study;
however, careful selection from a variety of possibilities enables you to
choose a design that best ts your study. The experiment is a good
starting point because it offers the most straightforward demonstra-
tion of causal links among variables. However, because of practical or
ethical reasons, many things that nurse scientists want to study cannot
be studied experimentally, so other designs have an important place in
nursing research. Nonexperimental researchers need to understand
the experiment in order to heighten awareness of factors that may be
problematic in their own studies.
Chapter 4 introduces students to the basic principles of experi-

mental design and presents examples of experiments and quasi-
experiments. Chapter 5 presents survey designs. Chapter 6 intro-
duces popular qualitative research designs, including
103
104 RESEARCH DESIGNS AND APPROACHES
phenomenology, grounded theory, and ethnography. Chapter 7 discusses qualita-

tive data collection and analysis methods specic to the eld study approaches.
Chapter 8 presents various new wave contemporary applied approaches, including
health promotion, evaluation, action, and feminist research. At the end of each
chapter, the Recommended Readings help readers come to a fuller understanding of
how to implement these basic designs. Students can select a specic design that
can be studied in detail by reading the references provided.
Chapter 4
Experimental and
Quasi-Experimental Designs
CHAPTER OUTLINE
A. Basics of Experimental Designs 3. Between- and Within-Subject Designs

1. The Rationale for Experimental Designs Compared
2. Key Elements in Experimental Designs D. Quasi-experimental Designs
B. Pre-experimental Designs 1. Nonequivalent Control Group Design
1. Same Group: Pretest and Posttest Design 2. Panel Studies
2. Exposed and Comparison Group Design E. Advantages and Limitations of
C. Classic Experimental Designs Experimental Designs
1. Between-Subjects Designs 1. Pre-experimental Designs
2. Within-Subjects Designs 2. Quasi-experimental Designs
3. Classic Experimental Designs
KEY TERMS
Baseline data Experiment Quasi-experimental design

Baseline stability Experimental group Randomization
Between-subjects design External validity Random variable
Blocking Instrument decay Response bias
Classic experimental design Internal validity Selection
Confounding variable Hawthorne effect Single blind
Control by constancy History Statistical regression
Control group Maturation Testing
Control variable Mortality Treatment levels
Counterbalancing Panel study Treatment variable
Crossover design Precision matching Within-subject design
Double blind Pre-experimental design
105
A. BASICS OF EXPERIMENTAL designs to measure if a treatment ( inter-

DESIGNS vention) has a particular effect on some
outcome. It is important to ensure that
As we saw in Chapter 1, an experiment is a other possible reasons for variations in
scientic investigation that tests cause- the outcome have been ruled out. In
and-effect relationships while controlling short, we must be condent that it is the
for the inuence of other factors. A key fea- treatment and not something else that is
ture of the design is that measures are effecting the outcome.
taken on two or more groups both before If you were studying the effect of lis-
and after an experimental intervention. A tening to instrumental music on a pa-
well-designed experiment should indicate tients stress level on the day of surgery,
whether or not an intervention (e.g., study- one would want to make certain that fac-
ing with the radio on) would bring about a tors such as the severity of the illness,
measurable change in a dependent vari- gender, and age of the patient have been
able (e.g., an increase or decrease in grade- taken into account. These other factors
point average) with other things being might all be having an impact on reported
equal. This last phrase is critical. It means stress levels. The strength of experimen-
that an experiment takes into account any tal designs is in their ability to sort out
number of factors (e.g., hours of sleep, in- causal relations while taking into account
terest in the subject matter, previous per- other factors. Indeed, experimental re-
formance) to make certain that it is the searchers use the term internal validity
researchers intervention and not some to refer to the extent to which one can
other factor that is inuencing the change. demonstrate that ones treatment is hav-
ing an impact on a given outcome and
1. The Rationale for Experimental that other sources of inuence have been
Designs controlled. External validity, on the
other hand, refers to the extent to which
In A System of Logic, John Stuart Mill one can make extrapolations from the
(18061873) identied the Method of Dif- particular study to other groups in gen-
ference as a key tool in understanding eral.
causal relations. Although Mills rationale
for experimental designs was the rst to 2. Key Elements in Experimental
set out the logic of experimentation sys- Designs
tematically, experiments, in fact, go back
much earlier. Experimentation dates to at Suppose you wish to assess the effective-
least 1648, when Blaise Pascal (16231662) ness of a CD-ROM in promoting atten-
had his brother-in-law conduct an experi- dance in a university nursing program.
ment testing for the presence of atmos- Suppose that some students are exposed
pheric pressure (Palys, 1992, pp. 241244). to a CD-ROM about university attendance
A special container was taken up a moun- and others view a nature CD-ROM. Will
tain, and it was noted that the level of mer- seeing the CD-ROM about university
cury in the container declined as it was nursing education increase the number
moved higher. As a control, another mer- of high school students wishing to attend
cury dish was not taken up the mountain a university nursing program? You de-
and its level was monitored; its level re- cide an experiment would be in order.
mained unchanged throughout the obser- There are a number of elements to such a
vation period. study, including time, the choice of vari-
Nurse researchers use experimental ables, and the number of levels.
EXPERIMENTAL AND QUASI-EXPERIMENTAL DESIGNS 107
a. Time 1. Treatment variable. In an experiment,

the independent variable whose effect
A key feature of an experiment is that is being studied is known as the treat-
measures are taken at different points in ment variable. In our example, the
time (i.e., before and after the treatment type of CD-ROM showneither about
is administered). Because the experimen- nature or about a university nursing
tal design involves measures of the de- programis the treatment variable.
pendent variables at two points in time An experiment attempts to detect the
(time 1, time 2 measures), the procedure direct effect of the treatment on the de-
ensures that the measure at time 1 pre- pendent variable. Some experiments
cedes the one at time 2. Second, as we in- include more than one treatment vari-
troduce the treatment to the experimen- able. In such cases, the experimenter
tal group (the CD-ROM about a university must determine the effects of each
nursing program) but not to the control treatment used on the dependent vari-
group (who see a nature CD-ROM in- able. (We might consider the effect of
stead), and follow this by measuring the exposure to a multimedia CD-ROM
dependent variable at time 2, it becomes with audio, video, and animated graph-
possible to make an inference about the ics versus a CD-ROM with only video
impact of seeing the university CD-ROM as an additional treatment variable).
(treatment variable) on the desire to at- 2. Control variables. Control variables
tend a university program (the dependent are those specically taken into ac-
variable). Thus, because the CD-ROM was count in designing a study. In this case,
shown between the time 1 and time 2 mea- they would include other major factors
sures, if there is a change in the attitudes which may inuence a students plans
toward attending a university nursing to attend a university nursing program
program among those seeing the univer- such as gender, academic abilities, fam-
sity CD-ROM, we will have made the rst ily role models in health care, and
step in demonstrating that the CD-ROM whether the student comes from a fam-
caused the change in attitude. ily background in which university at-
tendance is expected. These variables
need to be controlled, or taken into ac-
b. Variables in Experiments count, in designing the study to ensure
that the study ndings accurately re-
Chapter 2 noted that a dependent variable ect reality. By taking control variables
is the effect in a cause-and-effect relation; into account we are able to eliminate
in the CD-ROM study, the dependent vari- possible alternative explanations for
able would be the subjects desire to at- the observed relationships between ex-
tend a university nursing program. In an posure to a CD-ROM and plans to attend
experiment, the dependent variable is the a university nursing program.
phenomenon measured in order to deter- 3. Confounding variables. Experimental-
mine if any change has taken place as a re- ists refer to confounding variables as
sult of some experimental treatment. known factors that may unintention-
The independent variables include all ally obscure or enhance a relationship.
the variables taken into account, or ma- (Survey researchers use the term
nipulated, by the researcher that may in- source of spuriousness to refer to the
uence the dependent variable. Four same idea.) For example, one would
types of independent variables are rele- not want to measure some students
vant to experimental studies: attitudes toward attending a univer-
sity nursing program on Monday control variable and a two-level control

morning, and others on Friday after- variable. For example, an experiment de-
noon because it is possible that stu- signed to test the effect of a university
dents interest in attending a univer- nursing CD-ROM versus a nature CD-
sity nursing program varies systemati- ROM on the desire to attend a nursing
cally during the week. If this possibility program for high, medium, and low aca-
is not taken into account, it may con- demic achievers of both genders is a 2
found the results of our study. (university versus nature CD-ROM) 3
4. Random variables. Random variables (high, medium, and low achiever) 2
are unknown factors that may have an (male versus female) design.
impact on our dependent variable.
They are controlled through the ran-
dom assignment to groupssome stu-
B. PRE-EXPERIMENTAL
dents see the university nursing CD-
DESIGNS
ROM, and others see a nature CD-ROM
(see randomization below). It is impor-
Pre-experimental designs are consid-
tant to control for these variables; oth-
ered to be of limited scientific merit be-
erwise, we may falsely conclude that it
cause they cannot rule out alternative
was our treatment that inuenced the
explanations for observed relationships.
dependent variable.
However, understanding the problems
of making causal inferences from data
c. Treatment Levels derived from pre-experimental design
studies help us to better understand the
Often, researchers examine two or three
rationale and value of experimental de-
different treatment levels. These refer to
signs. For this reason, pre-experimental
the different treatments to which sub-
designs are discussed first in this chap-
jects are exposed. In our example, there
ter as a beginning reference point to il-
are two different possibilities for the
lustrate weaknesses in these designs
type of CD-ROM shown. This two-level
that lead to threats to the internal valid-
study compares the effect of seeing the
ity of studies and difficulties in inter-
university nursing CD-ROM with the ef-
preting the findings.
fect of seeing the nature CD-ROM. A
A pre-experimental design is one that
study with three treatment levels might
does not permit clear causal inferences
compare the effects of seeing a short,
about the impact of a treatment on the de-
medium, or long CD-ROM. The various
pendent variable. Although these designs
control and confounding variables are
share some of the elements of the experi-
also exposed to the subject at various
ment, their designs inhibit clear causal in-
levels. The simplest multiple-variable de-
terpretations. We will examine two pre-
sign would be a 2 2 one; this refers to
experimental designs, again using the
a design in which there are two levels of
illustration exploring the impact of a CD-
the treatment variable (the rst 2) and
ROM on students desire to attend a uni-
two levels in a control variable (the sec-
versity nursing program.
ond 2). Thus, when experimentalists
talk about designs they might talk of a 2
2 2 design. This means that the treat- 1. Same Group: Pretest and
ment has two categories; similarly, the Posttest Design
two control variables each have two lev-
els. A 2 3 2 experimental design has Suppose you decided that you would
two levels in the treatment, a three-level need a measure of the students predis-
positions toward a university nursing ed- could be used to measure the change in
ucation before seeing the CD-ROM. Hav- the desire to attend a university nursing
ing completed this measure, the CD-ROM program.
could then be shown, and the difference We could diagram the proposed de-
between the pre- and posttest scores sign in the following way:
Suppose that at time 1, 57 percent indi- may have announced a new program
cated that they wished to attend a univer- to prepare nurse practitioners; or a
sity nursing program and at time 2, after professor may have just won a Nobel
seeing the CD-ROM, the percentage in- Prize. In the context of experimental
creased to 73. Could we argue that the CD- design, history refers to concurrent
ROM produced a 16 percent increase in events that, along with the experimen-
those wishing to attend the university? tal manipulation, may be inuencing
The answer is no. There are a number of variation in the dependent variable.
factors that may render such an interpre- 2. Maturation. People change over time.
tation incorrect. Campbell and Stanley, in Perhaps the students have become
their classic work, Experimental and Quasi- bored with school and, by the time the
Experimental Design for Research (1966), second measures are taken, systematic
identied threats to internal validity that changes in attitude may have occurred.
might confound our interpretations. Read- Such changes could, in part, be an inu-
ers should note that these threats also ence causing the different responses at
apply to nonexperimental designs and time 1 and time 2. By 3:00 P.M. on Friday
should be considered when interpreting afternoon, grade 11 students may expe-
study ndings. rience a profound drop in their motiva-
tion to attend a university. Maturation
1. History. Any number of events may refers to any changes that occur in an
have happened in addition to the CD- individual subject over the course of an
ROM. For example, a local universitys experiment, which may, along with ex-
nursing school may have established a perimental manipulation, inuence the
new community-based health program outcome of the experiment.
for high school youth; the universitys 3. Testing. Even asking identical questions
basketball team may have won a na- at both tests may inuence the re-
tional championship; the university sponses. Some respondents may want
to appear consistent and, therefore, BOX 4.1 Regression Effect:

give the same responses in both tests. Reading Scores
Others, suspecting that the study is
meant to demonstrate how good the Suppose a clinical nurse specialist wants to
CD-ROM is, might want to help the re- try out a new program for teaching com-
searcher and exhibit response bias by munication skills to health professionals
being more positive the second time. working in a hospice setting. If the clinical
4. Instrument decay. Suppose we asked nurse specialist tests the communication
students to indicate their preferences skills of all health professionals in the hos-
pice setting and then selects the lowest
by the strength with which they
scoring 10 percent of health professionals
squeezed a hand dynameter. (A dyna- for the program, puts them through it, and
meter is a mechanical device with a cali- then retests them, almost certainly the
brated dial that permits a researcher to scores will increase. However, the amount
read the indicated intensity.) If, for ex- attributable to the program versus the
ample, the spring in the device has amount attributable to regression effects
weakened, the second set of readings would remain uncertain.
will be slightly higher. Instrument de-
cay refers to the fact that the dyname-
ter no longer measures reliably.
5. Statistical regression. To explain this
would, on average, be slightly more
idea, we will need to alter the example
positive on the second testing. The ex-
slightly. Suppose, after our pre-test
planation is that inaccuracies in mea-
scores had been taken, that we decide
surement at the time of the rst mea-
to show the CD-ROM just to those with
surement have tended to distort the
the most negative attitudes toward at-
data negatively. In short, those sub-
tending a university nursing program.
jects with very low scores may have
Could we legitimately say that any gain
simply given more negative responses
in the scores of these students is a re-
on that day because they were in a bad
sult of the CD-ROM? The answer is
mood. Box 4.1 presents another exam-
again no. The reason is that when a
ple of a regression effect.
sample is selected on the basis of ex-
treme scores, retesting will tend to
show a statistical regression toward 2. Exposed or Comparison Group
less extreme scores. That is, the most Design
negative scores are likely to change in
the direction of the mean simply be- After considering the aws in the previ-
cause extreme scores (either positive ous design, suppose you alter it to pro-
or negative) are unstable. Even without vide for a comparison group, as in the fol-
any CD-ROM, the students responses lowing diagram:
In the above design, one group is ex- any difference in test results may simply
posed to the CD-ROM promoting univer- reect initial differences or differences
sity nursing education, and the other is that emerge during the study but are un-
not. Can we legitimately conclude that related to the impact of the CD-ROM. Two
the difference in attitude between the ex- additional confounding factors are noted
posed group and the unexposed group by Campbell and Stanley (1966):
represents the impact of the CD-ROM? (If
75 percent of the students exposed to the 1. Selection. It is possible that the stu-
CD-ROM expressed a desire to attend a dents most inclined to go to a univer-
university nursing program and only 60 sity nursing program choose to go to
percent of those students who did not see the CD-ROM; thus, if they score
see the CD-ROM wished to go to a univer- higher in the test, this result may sim-
sity nursing program, could we conclude ply reect their initial predisposition
that the CD-ROM produced this 15 per- to study nursing at a university. Selec-
centage point difference in measured at- tion thus refers to subjects selecting
titudes?) Again, there are problems. The themselves into a study.
major difculty is that we do not know if 2. Mortality. Just as people choose to be-
the groups were the same to start with: long to a group, they may choose not to.
Some of those who see the CD-ROM may contains three crucial elements of the clas-
withdraw from the study before their sical experiment: control over extraneous
attitude toward attending a university variables, methods of dealing with pre-
nursing program has been measured. treatment similarity of groups, and manip-
The effect of subjects withdrawing from ulation of a treatment. A between-subjects
the experiment may be that those who design involves a control and an experi-
leave are less interested in higher edu- mental group. Measures are taken from
cation and hence the proportion for members in both groups before treatment
those who stay in the experiment are and repeated after the treatment has been
more likely to indicate a desire to attend experienced. Whereas the control group is
a university nursing program. Mortality exposed to a neutral treatment (seeing the
thus refers to subjects selecting them- nature CD-ROM in the illustration we have
selves out of a study. Therefore, mortal- been using), the experimental treatment is
ity may systematically distort the re- the focus of the study (the university CD-
sults of the study. ROM). Data analysis reveals whether there
are differences in the dependent variable
As we have seen, pre-experimental de-
as a result of the intervention, while taking
signs have serious problems in making
into account other factors.
clear causal inferences because of the in-
herent lack of adequate controls in these The experimental group is the one that
designs. Among the problems is whether is exposed to the treatment interven-
they have adequately demonstrated: (1) tion. The researcher attempts to ensure
that the test groups were similar before that each subject experiences the inter-
the introduction of the treatment vari- vention in as similar a fashion as possi-
able; (2) that they have a way of dealing ble to deal with extraneous factors that
with random, confounding, and control may be inuencing the outcome of the
variables. Classic experimental designs study.
deal with these issues. The control group is established so
that comparisons can be made between
the experimental group and the control
group. Normally, the control group is
C. CLASSIC EXPERIMENTAL
exposed to a placebo or some neutral
DESIGNS
treatment that is similar in the time it
takes and the kind of experience it rep-
To solve the problems in causal inference
resents. In the case of the CD-ROM ex-
faced by pre-experimental designs, clas-
ample, the control group is exposed to
sic experimental designs may be used.
a CD-ROM having a nature theme.
We will consider only the most basic of
these in this presentation: the between-
subjects and the within-subjects designs. a. Achieving Equivalence of Control
When properly executed, these designs and Experimental Groups in a
are powerful causal inference vehicles. Between-Subjects Design
A key factor in the success of a between-

1. Between-Subjects Design subjects design is that the treatment and
the control groups (completely different
A between-subjects design, also known as groups) are made as equivalent as possi-
the pretest, posttest control group design, ble before the experimental group is ex-

posed to the experimental condition. or control group. Your rule might be to

There are three ways of achieving equiv- assign odd numbers to the experimen-
alence: tal group and even ones to the control
group. Using this rule, if the rst num-
1. Through randomization ( in which indi- ber generated was a 7, you would as-
viduals are randomly assigned to ei- sign the rst subject to the experimen-
ther the treatment or to the control tal group. You would continue through
group) the random numbers and subjects un-
2. Through precision matching (e.g., til all are assigned. Also, the statistical
matching pairs according to gender, package SPSS has a procedure for se-
socioeconomic status, and grades, and lecting cases at random and may also
then assigning one person from each be used to assign cases to experimen-
of the matched pairs to the control tal or control groups.
group and the other person to the ex- Tables of random numbers may be
perimental group) used in a similar fashion. Table 4.1 is a
3. Through a combination of the above table of random numbers. Here the pro-
two: using precision matching to cedure is to close ones eyes and, with
match pairs and then randomly assign pencil in hand, mark the table of ran-
one member of each pair to the exper- dom numbers, note the number nearest
imental group and the other one to the the mark; establish a rule for proceed-
control group ing (perhaps, read down column to bot-
tom of page then start at top of next col-
( i ) Randomization
umn and read down). Again, using an
Randomization is a process of assigning
odd or even rule, you could begin group
subjects to a treatment or a control group
assignment; if the start number is even,
so that each subject has an equal chance
subject 1 is placed in the control group;
of being assigned to either group. This pro-
if the second random number is also
cess of random assignment should not be
even, then subject 2 also goes into the
confused with random selection, which is
control group. Continue assignment un-
discussed in Chapter 5. If the numbers are
til all subjects are assigned.
large enough, randomization provides
Table 4.2 illustrates the use of a table
control over both known and unknown
of random numbers or a computer-
factors (the control and random vari-
generated set of random numbers to
ables). A variety of techniques may be
assign subjects. In this case, the rule
used to accomplish the random assign-
established was that even numbers
ment. In all cases, we avoid allowing sub-
lead to an assignment to the control
jects to choose their group assignment.
groups, odd ones to the experimental
The method selected should assure that
group. Thus, if the numbers were odd,
each subject has an equal chance of being
odd, odd, even, odd, even, even. The
assigned to the experimental or the con-
subjects would then be assigned, as
trol group. Ways of accomplishing random
shown in Table 4.2.
assignment include:
A deck of playing cards may also be
Random numbers. Computers have used to assign subjects to groups.
random number generators built into Again, establish a set of rules (e.g.,
them and these may be used to assign hearts and diamonds mean experi-
individuals to groups. The rst num- mental group, clubs and spades mean
ber shown by the random number gen- control group assignment). Shufe the
erator is used to assign the rst person cards well. Turn cards up one at a time;
in the subject pool to the experimental if the rst card is red (hearts or dia-
Table 4.1 Table of Random Numbers

53 85 34 13 77 36 06 69 48 50 58 83 87 38 59
24 63 73 87 36 74 38 48 93 42 52 62 30 79 92
83 08 01 24 51 38 99 22 28 15 07 75 95 17 77
16 44 42 43 34 36 15 19 90 73 27 49 37 09 39
60 79 01 81 57 57 17 86 57 62 11 16 17 85 76
03 99 11 04 61 93 71 61 68 94 66 08 32 46 53
38 55 59 55 54 32 88 65 97 80 08 35 56 08 60
17 54 67 37 04 92 05 24 62 15 55 12 12 92 81
32 64 35 28 61 95 81 90 68 31 00 91 19 89 36
69 57 26 87 77 39 51 03 59 05 14 06 04 06 19
24 12 26 65 91 27 69 90 64 94 14 84 54 66 72
61 19 63 02 31 92 96 26 17 73 41 83 95 53 82
30 53 22 17 04 10 27 41 22 02 39 68 52 33 09
03 78 89 75 99 75 86 72 07 17 74 41 65 31 66
48 22 86 33 79 85 78 34 76 19 53 15 26 74 33
60 36 59 46 53 35 07 53 39 49 42 61 42 92 97
83 79 94 24 02 56 62 33 44 42 34 99 44 13 74
32 96 00 74 05 36 40 98 32 32 99 38 54 16 00
19 32 25 38 45 57 62 05 26 06 66 49 76 86 46
11 22 09 47 47 07 39 93 74 08 48 50 92 39 29
31 75 15 72 60 68 98 00 53 39 15 47 04 83 55
88 49 29 93 82 14 45 40 45 04 20 09 49 89 77
30 93 44 77 44 07 48 18 38 28 73 78 80 65 33
22 88 84 88 93 27 49 99 87 48 60 53 04 51 28
78 21 21 69 93 35 90 29 13 86 44 37 21 54 86
41 84 98 45 47 46 85 05 23 26 34 67 75 83 00
46 35 23 30 49 69 24 89 34 60 45 30 50 75 21
11 08 79 62 94 14 01 33 17 92 59 74 76 72 77
52 70 10 83 37 56 30 38 73 15 16 52 06 96 76
57 27 53 68 98 81 30 44 85 85 68 65 22 73 76
20 85 77 31 56 70 28 42 43 26 79 37 59 52 20
15 63 38 49 24 90 41 59 36 14 33 52 12 66 65
92 69 44 82 97 39 90 40 21 15 59 58 94 90 67
77 61 31 90 19 88 15 20 00 80 20 55 49 14 09
38 68 83 24 86 45 13 46 35 45 59 40 47 20 59
Table 4.2 Using Table of Random Numbers to Assign Subjects to

Experimental and Control Groups
Odd or Even from Treatment
Subject Random Number Table Condition
Subject 1 Odd Experimental
Subject 4 Even Control
monds), the rst subject is assigned to subject into the opposite group, altering
the experimental group; if the second back and forth as subjects arrive. As
card turned over is black, the second long as there are no systematic biases in
subject is assigned to the control arrivals, this technique should provide
group. Continue through the deck as- each subject with an equal chance of be-
signing each subject. ing assigned to the experimental group.
An equal number of slips of paper
These procedures are some examples
with Experimental Group or Control
of how to increase the likelihood that the
Group written on them are placed in a
treatment and control groups are equiva-
container to be drawn out one at a
lent before the introduction of the treat-
time. As they are drawn out of the con-
ment. However, because experiments are
tainer, the subjects are assigned to ei-
often carried out on fairly small samples,
ther the treatment or control group
researchers often try to enhance the
based on the slip of paper.
equivalence of the two groups by using a
Other techniques could also be used,
method known as precision matching.
but with caution. You could not, for ex-
ample, assign every second person to ( ii ) Precision Matching
the control group if your sample was Precision matching is a method of achiev-
made up of married couples with the ing equivalence between control and ex-
mans name rst. Even if you start ran- perimental groups by ensuring that the
domly, you would end up with an all- groups are matched on certain key vari-
male experimental group and an all- ables. The key variables vary depending
female control group! Be careful that the on the topic being investigated. In the CD-
technique you use provides each indi- ROM study, one might decide to match on
vidual with an equal chance of ending previous computer experience and on
up in the experimental group and that gender. A preliminary measure would be
there are no systematic biases built into taken on the subjects computer experi-
the order of your subject pool. If you ence and the subjects gender would be
choose a random number (or a heads or noted. Subjects could then be arranged
tails coin ip) to determine whether the from highest to lowest computer experi-
rst subject would be placed in the ence for the women, and then produce a
treatment or in the control group, you similar listing for the potential male sub-
could then proceed by placing the next jects. The top two women on the com-
puter experience measure would then be duction of the treatment. Although the
randomly assignedone to the experi- treatment and control groups may be
mental group, the other to the control equivalent on many measures as a result of
group. The next pair of women would randomization, they nonetheless may be
then be randomly assigned. And so on somewhat different in terms of their mea-
through each pair, followed by using the sures on the dependent variable. Experi-
same procedure for the male pairs. Fol- menters therefore collect baseline data so
lowing this procedure ensures equiva- that we have a stable starting point against
lence in the average computer experience which we can make comparisons as we re-
in the treatment and control groups and peat the measures of the dependent vari-
also matches them by gender composi- able through the course of the study. Base-
tion. Rather than simply trusting that luck line data is simply the score a subject has
(using the randomization process) will before the introduction of the treatment
balance the two groups, precision match- variable. In some circumstances, several
ing ensures equivalence on the two key measures are taken and we dont intro-
variables. We rely on the random assign- duce the treatment variable until the mea-
ment to balance the groups on other con- sures achieve stability. This is referred to
trol and random variables. as baseline stability. So even if the average
baseline scores of the experimental group
( iii ) Blocking
compared with the control group are
Another technique used in between-sub-
somewhat different, we are able to mea-
jects designs to reduce within-group vari-
sure how much the scores change from the
ations is blocking. A blocked design
baseline levels when the measures are re-
refers to one in which subjects have been
peated after the treatment has been ad-
grouped together on some variable that
ministered.
needs to be controlled, and subjects are
then randomly assigned to treatment and
control groups. For example, if you were b. Analyzing the Data
testing for the effectiveness of a new
health promotion program and you were To estimate the impact of the CD-ROM,
working in an area with many immigrant we could do the following computation
children, you might want to use blocking on the percent wanting to attend a uni-
in your experimental design. You would versity nursing program before and after
assign immigrant children randomly to seeing the CD-ROM (Table 4.3):
either the experimental group (new pro- At time 1, 57 percent of the experimen-
gram) or to the control group (old pro- tal group planned to attend a university
gram); you would then do the same for nursing program compared with 55 per-
nonimmigrant children. The blocking
does not allow you to measure the inu-
ence of immigration status on learning Table 4.3 Percent Wanting to Attend
but does increase your chances of de- University Nursing Program by Exposure and
tecting a signicant effect on your new Nonexposure to CD-ROM
program. The reason for this is that you Patients Wanting to Attend a
have reduced the variation between the University Nursing Program, %
treatment and the control groups. Group Time 1 Time 2 Difference
( iv) Baseline Data Treatment 57 73 16
Another way in which the pretest equiva- Control 55 61 6
lence is handled is by taking measures of
Estimated impact of CD-ROM: 10
the dependent variable before the intro-
cent of the control group members (base- 1. Changes in the treatment variable
line measures). The time 2 measures indi- must occur before changes in the de-
cate that the experimental group (after pendent variable. Because the be-
seeing the CD-ROM about attending a uni- tween-group design involves measures
versity nursing program) now had 73 per- of both the treatment and the depen-
cent planning to attend a university nurs- dent variables at two points in time
ing program compared with 61 percent for (time 1, time 2 measures), the proce-
the control group (viewers of a nature CD- dure ensures that the measure at time 1
ROM). To estimate the impact of the uni- precedes the one at time 2. Second, as
versity CD-ROM on a persons desire to at- we introduce the treatment to the ex-
tend a university nursing program, we perimental group (the CD-ROM about a
subtract the time 1 percentage from the university nursing program), but not to
time 2 percentage. Here we note a 16 per- the control group (who see a nature CD-
cent difference for those in the experi- ROM instead) and follow this by mea-
mental group compared with a 6 percent suring the dependent variable at time 2,
difference among those in the control it becomes possible to make an infer-
group. We subtract the control difference ence about the impact of viewing the
(6 percent) from the treatment difference university CD-ROM (treatment vari-
(16 percent) and arrive at an estimate of able) on the desire to attend a univer-
the impact of the CD-ROM amounting to sity program (the dependent variable).
some 10 percentage points. Thus, because the CD-ROM was shown
The seven factors Campbell and Stan- between the time 1 and time 2 mea-
ley (1966) identied as confounding in- sures, if there is a change in the atti-
terpretations are dealt with in between- tudes toward attending a university
subjects designs through: nursing program among those seeing
the university CD-ROM, we will have
Attempting to make certain that the
made the rst step in demonstrating
groups are similar to begin with
that the CD-ROM caused the change in
Noting that the various factors should
attitude.
inuence the treatment and the con-
2. The variables must be associated; as
trol groups equally
values on the treatment variable are
Thus, between-subjects designs rely on increased, the dependent measures
establishing pretreatment similarity of vary correspondingly. A variety of
control and experimental groups to mini- techniques may be used to illustrate
mize the effects of history, maturation, that the variables are associated. For
testing, instrument decay, statistical re- example, we might show that those
gression, selection, and mortality. There- who saw the CD-ROM are more likely to
fore, some form of randomization is a key indicate that they are planning to at-
tool in the experimenters quest to estab- tend a university nursing program than
lish unambiguous causal relationships. those who did not see the CD-ROM. To
say that the dependent variable varies
systematically indicates that either:
c. Demonstrating a Causal The proportion of respondents say-
Relationship in a Between-Subjects ing that they plan to attend a uni-
Design versity nursing program is higher
among those exposed to the univer-
To show a causal relationship in any re- sity CD-ROM; or
search, three conditions must be demon- The proportion of respondents say-
strated: ing that they plan to attend a uni-
versity nursing program is lower not having an impact on the results. For
among those exposed to the univer- example, if you were testing the impact
sity CD-ROM; or that the intensity of color (treatment
The proportion of students saying variable) has on the speed of recovery
that they plan to attend a university from a major depressive episode, you
nursing program increases as the should not simply move systematically
length of the CD-ROM increases. from low color intensity (soft white,
However, there may appear, for ex- beige) through to intense color (bright
ample, to be a leveling off in the pro- yellows, bright pinks, lime greens). Sub-
portion of students planning on at- jects will probably recover from their
tendance when the CD-ROM is depressive reaction as they go through
longer than 23 minutes. their regular treatment regimen, and
3. Nothing but the treatment variable you should therefore vary the presenta-
can have inuenced the dependent tion of different color conditions for
variable. To ensure that it is only the each subject randomly. In this way, the
treatment variable that is causing vari- order of presentation would not in itself
ation in the dependent variable, steps systematically inuence the speed of re-
must be taken to rule out, or control, covery. If you do not deal with this po-
various sources of contamination: tential confounding factor, you would
Step 1: Ensure that the context in which not be able to tell how much of the in-
the experiment is carried out is the crease of speed in recovery was caused
same for all subjects. To guarantee the by increasing the intensity of color in
similarity of the experimental situa- the clients environment and how much
tion, researchers take great pains to was caused by a response effect to the
ensure that physical conditions (e.g., regular protocol of care for depressive
temperature, humidity, lighting condi- clients. Confounding variables of this
tions) all remain the same; similarly, sort are dealt with by:
instructions to the subjects are stan- Varying the way the treatment is ad-
dardized, even recorded, to be certain ministered to the subjects; one
that each respondent gets the same in- strategy for doing this is known as
formation in the same manner. As counterbalancing. Here a treat-
much as possible, all subjects should ment level is introduced, changed,
have a similar experience in partici- maintained, and then returned to
pating in the experiment. the rst level to control for effects of
Step 2: Balance the background charac- learning on the subjects perfor-
teristics of the subjects. Typically this mance.
control is accomplished through some Statistical controls in which sub-
combination of precision matching groups are analyzed separately to
(matching a woman with a woman; take into account potential con-
high computer experience person with founding variables (perhaps analyz-
another who is also experienced with ing the male subjects separate from
computers) and randomly assigning the female subjects).
each person to the control or to the ex- Step 4: Deal with random variables. Ran-
perimental groups. The goal is to try to dom variables include all other un-
populate both groups with similar known factors that might inuence the
kinds of people. dependent variable, none of which you
Step 3: Neutralize any confounding vari- can control. Subjects are assumed to be
ables. The researcher must be careful different on all dimensions. Knowing
that the experience of the experiment is this, experimenters do not let subjects
choose which group they will go into expose each subject to the different treat-
(they might think it more fun to be in a ments. Because the subject is the same
experimental group than in a boring person, the background characteristics,
control group; or they might choose to attitudes, and intelligence are all perfectly
be in the same group as their friends). controlled. In a control group design, on
The randomization process used in the the other hand, the researcher counts on
assignment of subjects to groups in- randomized assignment to groups (or
creases the likelihood that the two precision matching) to adjust for known
groups are similar on both known and and unknown variations between the two
unknown variables before the introduc- groups. With large samples, the assump-
tion of the treatment variable. When tion that sources of contamination have
sample sizes exceed 30 in each group, been satisfactorily addressed is reason-
random assignment is likely to result in able; with smaller samples, one has less
reasonable balancing of the groups on faith in the ability of randomization to
random factors. deal with such contamination.
The key advantage of a within-subject
Box 4.2 presents a classic between-
design is that it provides convincing evi-
subjects experiment in nursing, Ciliskas
dence for the impact of the treatment
investigation of two nondieting interven-
variable on the dependent variable. The
tions for obese women. This study is of
evidence is all the more convincing be-
interest for its design, its anticipated nd-
cause, by using the same subject under
ings, and its clarity of presentation. It
different levels of the treatment variable,
uses a randomized trial to determine the
one can be condent that most of the pos-
efcacy of new weight paradigm inter-
sible extraneous inuences, such as gen-
ventions as alternatives for dieting in
der, age, socioeconomic status, type of
obese women.
background, and values, have been con-
trolled. Sometimes the control achieved
2. Within-Subjects Designs in within-subject designs is called control
by constancy. The term stems from the
An alternative to the between-subjects de- idea that because the same subject expe-
sign is a within-subject design (also known riences different levels of the treatment,
as repeated measures designs). Sometimes the subject acts as his or her own control.
these designs use a single subject, some- In contrast, the between-subjects design
times just a few subjects, and sometimes has to try to control these differences by
large samples. What is the idea behind this random assignment or precision match-
kind of design? ing. Especially in studies using relatively
We have stressed the importance of small samples, such controls are some-
controlling possible sources of contami- what suspect because there are always
nation. Matching groups and randomly as- going to be some random uctuations,
signing subjects to treatments were noted which may mask true effects.
as ways of ensuring pretreatment similar-
ity. Without such controls, it would be im- a. Crossover Designs
possible to distinguish treatment effects
from the effects of other factors. Suppose we were interested in the effect
Sometimes, however, one can use an of two smoking cessation treatments
alternative strategy, which provides for ( i.e., the nicotine patch and Zyban) on
the ultimate in control of extraneous fac- number of cigarettes smoked by cardiac
tors. Suppose that instead of assigning clients. A within-subject design could be
people to different treatments, we instead chosen to study this effect. In this design,
EVALUATION OF TWO NONDIETING INTERVENTIONS FOR

OBESE WOMEN
Many women have a long history of failure at at- were excluded. A total of 78 subjects had data
tempted weight loss and experience an accom- collected on the primary outcome variables at
panying self-appraisal of being fat and a failure. pre- and posttest time. There were no statisti-
Recently, feminist critiques of the biomedical cal differences in the intervention and control
paradigm have contributed to a new paradigm groups in mean ages, mean weights, mean
that encourages no further attempts at weight height, mean body mass indexes, or means of
loss but instead focuses on self-acceptance, im- percent average weight, indicating that random-
proved body image, better nutrition, and more ization was successful.
physical activity. Because little research exists The treatment interventions consisted of
on the effectiveness of such nondieting ap- meeting once a week for 12 weeks. The educa-
proaches, the primary purpose of this project tion group consisted of a lecture format with 16
was to evaluate them. to 21 people that met for 1 hour per week and
A randomized trial design was used in which included only the educational content such as
consenting women were randomized, using a the cultural imperative for women to be thin,
table of random numbers, into three groups: strategies to achieve a nondieting style of eating,
education alone, psychoeducation, or a no and the importance of regular physical activity.
treatment (wait-list) control group. Measures The psychoeducation group was a more inten-
were taken at pretest, posttest, and 6 months sive, small group (6 to 8 women) that met for 2
and 1 year posttests, although this article re- hours every week and included the educational
ports only the pre- and posttest results. content plus cognitive therapy strategies, as-
Three hypotheses were tested: sertiveness, body image exercises, and a group
support component. The control group received
1. Participation of obese women in a group ed-
no intervention.
ucation intervention program increases self-
In relation to the three hypotheses tested,
esteem, decreases body dissatisfaction, and
results suggest:
decreases restrained eating patterns.
2. Participation of obese women in a group 1. Participation of obese women in a group
psychoeducation intervention program in- education intervention did not increase self-
creases self-esteem, decreases body dissatis- esteem or reduce body dissatisfaction, nor
faction, and decreases restrained eating pat- did it decrease restrained patterns of eating
terns. compared with a control group.
3. Participation in a psychoeducation interven- 2. Participation of obese women in a group
tion will result in increased self-esteem, de- psychoeducation intervention increased self-
creased body dissatisfaction, and decreased esteem, decreased body dissatisfaction, and
restrained eating patterns, which are of decreased restrained patterns of eating in
greater magnitude than results of participa- comparison to a control group.
tion in the education intervention. 3. Participation in the psychoeducation inter-
vention resulted in increased self-esteem and
A variety of scales were used to measure self-
decreased restrained patterns of eating,
esteem; feelings of social inadequacy; body dissat-
which are of greater magnitude than results of
isfaction; restrained eating patterns; and factors in
participation in the education intervention.
eating behaviors, including cognitive restraint,
The changes in body dissatisfaction did not
disinhibition, and susceptibility to hunger. After
reach the level of statistical significance.
signed, informed consent was obtained, the pre-
test questionnaires were completed. The same In conclusion, this study established that a
questionnaires were used after testing, immedi- 12-week group intervention can be effective in
ately after the 12-week intervention session. improving self-esteem, encouraging a nonre-
Inclusion criteria included: women, age 20 strained pattern of eating, and, in the short
years or older, 120 to 200 percent average term, improving body dissatisfaction. For a
weight. Women who were pregnant or had a complete description of this study, consult the
medical or psychiatric disorder self-disclosed original source.
SOURCE: Summarized from Ciliska, D. (1998). Evaluation of two nursing interventions for obese women. Western
Journal of Nursing Research, 20(1), 119135.
both treatments are given to the same in- ban), and the other half (group 2) receive
dividuals rather than separately to two treatment B, (Zyban) followed by treat-
different samples. An obvious advantage ment A (the nicotine patch). This distrib-
in this type of design is the saving in the utes the carry-over effects equally across
number of subjects required because the all conditions of the study, thus canceling
same group is used twice (or more often them out. It can be argued that any differ-
if necessary). Because the subjects serve ence found between groups 1 and 2 can-
as their own controls, the sample size re- not be attributed to the fact that treat-
quired to detect a signicant effect is con- ment A always followed treatment B or
siderably less. vice versa. To prevent an effect related to
A drawback does exist in within-sub- time, the same amount of time needs to be
ject designs: Because two treatments are allocated to each treatment. The design
being tested, there may be an effect from must also allow for a washout period. This
the rst treatment that the subjects un- is an adequate interval of time between
derwent. One may argue that the subjects the two treatments to eliminate the effect
who receive the second treatment are not of the rst treatment. Figure 4.1 illustrates
exactly the same as the group of subjects a within-subject, crossover design.
who received the rst treatment. To can- In this design, a group of heavy smokers
cel out the effects of any carry-over from are selected from a cardiac clinic. They
treatment A to treatment B, a within-sub- are randomly assigned to two groups, 1
ject crossover design (also called a coun- and 2. Before the experiment begins, an
terbalanced design) is used. Crossover de- initial baseline assessment of the number
signs assign subjects randomly to a of cigarettes smoked by each subject is
specic sequencing of treatment condi- made over a 1-week period. The baseline
tions. In this case, half of the subjects measure is an essential element in any ex-
(group 1) receive treatment A (the nico- perimental design because it represents
tine patch) followed by treatment B (Zy- the situation existing before any experi-
Group 1 Group 2
Nicotine Patch
Wash
Out
Period
Group 2 Group 1
Zyban
Week 1-4 Week 5 Week 6-9

Figure 4.1 Within-subjects, crossover (counterbalanced) design.
mental treatment is introduced. Group 1 ming the lights in the workplace would be
then receives the nicotine patch for 4 the way to increase productivity. How-
weeks followed by Zyban for 4 weeks. ever, when the experimenters then in-
Group 2 receives Zyban for the rst 4 creased lighting intensities to the baseline
weeks followed by the Nicotine patch for 4 levels, they noted that worker productivity
weeks. The design specied that each sub- continued to increase. Woops! Something
ject would crossover to the alternate other than the intensity of illumination was
smoking cessation treatment after a 1- having an impact on the speed of work. A
week washout period. Throughout the 9 possible interpretation of this result is that
weeks of the study, the subjects recorded the workers knew that they were being ob-
the number of cigarettes smoked per day. served, so they tried hard to please the ob-
If all the smokers had received Zyban servers by increasing their productivity
followed by the nicotine patch and if Zy- (Roethlisberger and Dickson, 1939). When-
ban had proved to be superior in reduc- ever an outcome cannot be reversed, you
ing the number of cigarettes smoked, the have to suspect that some factor (or fac-
conclusion would be open to criticism tors) other than the treatment is inuenc-
that it was simply the fact that Zyban was ing the dependent variable (in this case,
tried rst that explained the results. Giv- the speed of work).
ing half the sample Zyban rst and the In nursing research, an investigator
other half of the sample the nicotine may have to contend with a double
patch rst overcomes this criticism. This Hawthorne effect. For example, if an or-
solution to the problem of order and thopedic unit is interested in testing the
carry-over effects is known as counter- impact of a new ambulation protocol on
balancing. A disadvantage to this design client recovery rates after hip replace-
is the fact that subject dropouts may in- ment surgery, nurses, patients, and other
crease because the data collection pe- hospital staff may be aware of their par-
riod is longer. ticipation in the study. This may lead to
both groups altering their behavior ac-
cordingly. This is why double-blind ex-
b. Hawthorne Effect periments, in which neither the subjects
nor those who administer the treatment
A limitation of the experimental design are aware of who is in the experimental or
that may inuence the study design in Fig- the control group, are so powerful. Un-
ure 4.1 is a phenomenon known as a fortunately, it is often difcult to disguise
Hawthorne effect. This phenomenon re- nursing interventions in the same way
fers to any differences in the dependent that medication or medical treatments
variable that are not the direct result of can be disguised. Hence, for some nurs-
changes in the treatment variable. This ing investigations, a double-blind ap-
phenomenon is named after one of the ex- proach is not feasible.
periments done in the Hawthorne Western
Electric plants in Chicago by Roethlis-
berger and Dickson (1939). In their study, 3. Between- and Within-Subjects
workers output was monitored under Designs Compared
varying intensities of lighting. (The hy-
pothesis was that worker productivity Lets compare a between-subjects ap-
would increase with increasing levels of il- proach with a within-subject approach to
lumination.) The experimenters noted that the same research question: What is the
as lighting intensity decreased, worker out- efcacy of therapeutic touch (a specialized
put increased. It looked as if simply dim- type of healing touch often used in nursing
practice) on situationally induced stress in tive in reducing stress that is situation-

healthy adult women? We will approach ally inducedthere is simply too much
the question by looking at two different variability in the pressure values. Had
between-subjects studies and one within- one or two persons ended up in the op-
subject study. Table 4.4 compares cti- posite group, the results might change
tious data on these studies. substantially. (A test of signicance con-
rms our suspicion because the differ-
Student nurses comparison: a be- ence between the diastolic readings
tween-subjects design. Suppose nurs- achieved with therapeutic touch and
ing students in a research methods the diastolic readings of subjects re-
course are asked to volunteer for an ex- ceiving casual touch is not statistically
periment. They are asked to report to signicant. See Chapter 12 for a discus-
the nursing school laboratory on Satur- sion of what these tests tell us.)
day afternoon. Students are given basic Experienced nurse practitioners: a be-
instruction on the procedure of thera- tween-subjects design. The second ex-
peutic touch, a demonstration of it, and periment is identical to the previous
an opportunity to practice it on each one, except we will use expert nurse
other for 20 minutes. The students are practitioners who have been specially
then randomly assigned to either labor- prepared to administer therapeutic
atory A or B. In laboratory A are healthy touch and have varying years of experi-
adult women who will receive therapeu- ence in doing so. Using each nurses his-
tic touch from student nurses while tory of administering therapeutic touch
being subjected to articially induced permits us to rank-order all the nurses
stress. In laboratory B are healthy adult by skill level in therapeutic touch. As-
women who will receive casual touch signment to laboratory A and B is done
from student nurses while being sub- in pairs: using a coin ip, the ranked
jected to articially induced stress. The nurses are randomly assigned to either
women receiving the touch therapy are laboratory A (therapeutic touch) or lab-
not aware of the type of touch they will oratory B (casual touch) until 20 nurses
be administered. This is called a single- have been assigned. Once again, each
blind study and is used to minimize the expert nurse records the blood pres-
possible effects on the subject of know- sure reading for the subject and then ad-
ing which intervention is being used. ministers the touch intervention to the
Each student rst records a blood pres- participant who has been subjected to
sure reading (a physiological indicator articially induced stress, and then
of stress) for the subject. Students then records the blood pressure of the sub-
administer the touch intervention to the ject a second time. Section II of Table 4.4
women who have been subjected to ar- shows much less variation than was ap-
ticially induced stress and then record parent among the student nurses.
the blood pressure of the subject a sec- Nonetheless, some of the variation be-
ond time. The results of the difference in tween the expert nurses may still be
diastolic pressure readings are shown caused by differences in nursing skill,
in section I of Table 4.4. Note that there the female subjects, or the touch inter-
is considerable variation in the scores. ventions. The results indicate that the
Even though the average reduction in reduction in diastolic blood pressure
diastolic blood pressure readings for readings of subjects who received ther-
subjects in the therapeutic touch group apeutic touch are somewhat greater
is higher, we are not comfortable in con- than those of subjects who received ca-
cluding that therapeutic touch is effec- sual touch. The nagging problem, how-
Table 4.4 Results from Three Studies to Test Therapeutic Touch versus Casual Touch on Blood
Pressure
Therapeutic Touch Casual Touch
Experiment Name Change in BP Name Change in BP
I. Student nurses: Kathleen 10 Paula 12
Between-subjects Li 9 Marlies 7
design; random Danielle 11 Vanny 6
assignment to Kim 8 Paula 6
groups Kevin 12 Sandra 11
Mary 4 Marius 1
Yvonne 5 Chris 2
Tara 7 Andrea 4
Ursula 8 Tony 7
Holly 10 Carol 8
Mean 8.40 6.40
Standard deviation 2.547 3.502
Mean difference 2.00
II. Expert nurses: Eve 12 Mary 11

between-subjects Joan 12 Ted 11
design; pairs Freda 11 Gladys 10
matched on Carlie 10 Ronelda 9
known skill level; Jill 12 Molly 10
random assignment Neil 11 Emily 10
to groups Hazel 10 Sara 9
Iris 11 Edwina 9
Frank 10 Zoe 9
Betty 11 Brenda 9
Mean 11.00 9.70
III. Expert nurses: Jamie 12 Jamie 10

within-subject design Petra 11 Petra 10
Margo 11 Margo 11
Jill 13 Jill 12
Sandra 13 Sandra 12
Dale 12 Dale 11
Paula 11 Paula 10
Teresa 11 Teresa 10
Linda 11 Linda 10
Melanie 12 Melanie 11
Mean 11.70 10.70
ever, is whether differences in the sub- tic touch versus casual touch). To make a
jects themselves and their reaction to study as convincing as possible, we should
stress may have been responsible for have lots of variations between the cate-
some of the differences in the diastolic gories (average diastolic pressure should
readings. However, because of the small vary consistently by type of touch) but lit-
sample of only 10 expert nurses in each tle variation within the category (diastolic
group, we lack condence with the re- pressure readings for those subjects re-
sults. (The test of signicance, however, ceiving the same type of touch). In the
does indicate a statistically signicant three designs presented here, note how
difference in diastolic blood pressure the rst design had a lot of within-group
for the two touch interventions.) variation (the nursing students varied con-
Expert nurse practitioners: a within- siderably in their therapeutic touch skills);
subject design. Given some uncertainty the second and third designs had less
about the results, could we get even within-group variation (all expert nurses).
greater condence in our results by Although it is not always possible to
changing our design? Suppose we sim- use a within-subject design, the advan-
ply have each nurse run two trials with tage of such a design over a between-sub-
the same subject, once administering jects design is that you need fewer sub-
therapeutic touch and once administer- jects, require less time to train and run
ing casual touch? We now have control your experiment, and exercise almost
over the subject who will be subjected perfect control over extraneous inu-
to articially induced stress, as well as ences on your dependent variable.
the skill of the nurse. We have control
by constancythe nurse and the sub-
ject are the same for the test with ther- D. QUASI-EXPERIMENTAL
apeutic touch and casual touch. As be- DESIGNS
fore, the subjects are not aware of
which type of touch they will receive Often we want the power of an experimen-
rst (single blind). They are randomly tal design in making causal inferences but
assigned to the different orderings of nd that it is not possible to conduct an ex-
the touch treatment. The results, re- periment. Sometimes this happens be-
ported in section III of Table 4.4, show cause we cannot manipulate the required
about the same variability as the re- variables for ethical or for practical rea-
sults of the previous experiment, but sons; sometimes we cannot randomly as-
now we have greater condence that sign people to different treatments. When
we have controlled both nurse and sub- these situations arise, researchers modify
ject variability. (The differences by the experimental approach and design a
touch intervention are statistically sig- quasi-experiment. A quasi-experimental
nicant.) As a nurse practicing in the design is one in which it has not been pos-
specialized and controversial eld of sible to meet all of the following require-
therapeutic touch, which of these tests ments:
would you trust more?
Randomly assign subjects to a treat-
ment or control group
The goal in experimental research is
to isolate and measure treatment effects.
Control the timing or nature of the
treatment
To demonstrate an effect convincingly,
one wants clear differences in the depen- An important feature of the quasi-
dent variable (diastolic blood pressure experimental design is its attempt to com-
values) by treatment categories (therapeu- pensate for the absence of either the ran-
dom assignment or the control group ple, nursing students at two different uni-
component by introducing other controls. versities or clients from two tness clubs
The quasi-experimental design comes as are used. The Smeenk study is a good il-
close as possible to the experimental de- lustration of how the limitations of a
sign in order to measure the impact of a quasi-experimental design can be ad-
treatment. dressed in a study. As you read the sum-
There are a number of quasi-experi- mary of the study in Box 4.3, pay particu-
mental designs available to nurse re- lar attention to the means used by the
searchers. These include the nonequiva- investigators to deal with the issue of the
lent control group, the one-group pretest- nonequivalence of the two groups and
posttest (also called pre-experimental), possible confounding variables.
the time series design, and the one-time This study is convincing because it is a
case study design, to name a few. This quasi-experimental study that compen-
section discusses the most basic quasi- sates for the fact that it is not a classic ex-
experimental design: the nonequivalent perimental design. The investigators could
control group. To illustrate this approach, not assume the equivalence of the two
we will examine a study using this design. groups because they were drawn from two
different natural collectives and not ran-
domly assigned. To address this, the re-
1. Nonequivalent Control Group searchers made a point of showing the
Design similarities between the two geographical
areas from which the sample was drawn.
To illustrate the nonequivalent control The areas were similar in terms of popula-
group design, we will examine an imagi- tion and accessibility to hospitals and
native study done by Smeenk and associ- home care services. Hence, it is unlikely
ates (1998) that makes an important con- that the different residential areas could
tribution to nursing knowledge in the explain the ndings. Furthermore, a broad
area of palliative care and quality of life. spectrum of patient and direct caregiver
The objective of this study was to inves- characteristics were collected at time 1.
tigate the effects of a home care interven- The groups proved to be fairly comparable
tion program for terminal cancer patients on these characteristics. The data were ex-
on the quality of life of direct caregivers, amined for confounding variables. Poten-
compared with standard care programs. tial confounders were entered into the re-
It is an important investigation because it gression analyses with the preintervention
is the rst documented controlled study OQOLI (at time 1). As a result of these pre-
that has investigated the effects of such cautions, one can have condence that the
an intervention program on the quality of nal results of the study are convincing.
life of direct caregivers of terminally ill
cancer patients.
In a nonequivalent control group de- 2. Panel Studies
sign, the control and experimental groups
are not drawn from the same population. Panel studies monitor specic organiza-
It is used in situations in which subjects tions or individuals over time. Because
cannot be randomly assigned to the ex- the participants in panel studies are mon-
perimental and control group. Conse- itored over time, these designs are in-
quently, the researcher cannot assume cluded in this section on quasi-experi-
equivalency of the two groups. Usually in mental designs. Panel studies frequently
a nonequivalent control group design, the use survey data collection techniques
groups are naturally occurring; for exam- (questionnaires or interviews) to investi-
TRANSMURAL CARE OF TERMINAL CANCER PATIENTS:

EFFECTS ON CAREGIVERS
Few studies in the nursing literature have in- transmural home team, (3) a collaborative home
vestigated the effect of a home care program on care dossier (case file), and (4) protocols de-
the quality of life of direct caregivers of cancer signed by a multidisciplinary team to standardize
patients. This article describes an intervention and support caregivers when using specific skills
program called transmural care, which was de- and actions such as intravenous therapy or
veloped specifically to meet the individual needs epidural and spinal pain relief.
of patients with cancer, and was offered by pro- The outcome variable was the direct care-
fessional caregivers from hospitals and primary givers quality of life. It was measured in a mul-
care teams in the Netherlands. Its purpose was tidimensional way using four instruments that
to provide supportive care at home for patients measured four dimensions, namely, daily func-
and their direct caregivers. tioning, fear, loneliness, and general well-being.
It was hypothesized that transmural care Data were collected 1 week before (T1), 1
may lead to a better quality of life for direct week after (T2), and 4 weeks (T3) after the pa-
caregivers of terminal cancer patients. Because tients discharge from the hospital, as well as 3
of practical and ethical reasons, the study used months after the patient had died (T4). Dis-
a quasi-experimental design. charge from the hospital marked the starting
The sample consisted of 116 direct care- point of the intervention. Two trained inter-
givers of terminal cancer patients whose esti- viewers using structured questionnaires col-
mated prognosis was less than 6 months. Pa- lected the data. They were blinded to the study
tients and their caregivers were included from design. To yield an aggregate score for the qual-
January 1994 until February 1995. All caregivers ity of life measurements, a factor analysis was
living in Eindhoven, the Netherlands, were as- conducted at each assessment stage. This
signed to the intervention group (n 79), and yielded one factor and was considered the pri-
those living in the urban surroundings of Eind- mary outcome variable called the overall qual-
hoven were assigned to the control group (n ity of life index (OQOLI).
37). (The experimental and the nonequivalent Before assessing whether the intervention
control group were natural collectives.) Consid- contributed positively to the direct caregivers
ering the study design, one should note that the OQOLI, a search for possible confounding vari-
two areas are very similar. The population of ables in the relationship between the experi-
both Eindhoven and the suburban area is mental condition and the outcome measure
200,000 each. The maximum travel time to a was performed. Correlations between the di-
hospital in both sites is approximately 30 km rect caregiver and patient characteristics and
(20 miles), and the availability of home care ser- the outcome measure yielded only two poten-
vice in each area is comparable. Furthermore, a tial confounders: the patients being treated
pilot study using the database of the Compre- with chemotherapy and the direct caregivers
hensive Cancer Center showed no significant also being the patients partner. To control for
difference in the incidence of the various cancer the confounding variables they were added to
types, mean age, and genders of cancer patients the experimental condition as independent
in these two populations. variables in the regression analyses.
After discharge from the hospital to the To assess whether the intervention con-
home, subjects in both groups received the stan- tributed positively to the direct caregivers
dard care available in the Netherlands. In addi- OQOLI, hierarchical backward multiple regres-
tion, patients and caregivers in the experimental sion analyses were performed on the outcome
group received the intervention program. This measure. A total of 71 direct caregivers withdrew
consisted of four main elements: (1) access to a from the study at various times because they
specialist nurse coordinator, (2) a 24-hour tele- found it too burdensome. No significant differ-
phone service in the hospital with access to a ences could be found when comparing dropouts
TRANSMURAL CARE OF TERMINAL CANCER PATIENTS:

EFFECTS ON CAREGIVERS (CONTINUED )
with the direct caregivers, who could be evalu- In conclusion, this study supports the value
ated on their initial characteristics. The results of transmural care of terminal cancer patients
suggest that the intervention contributed posi- and their caregivers. Such programs can be ef-
tively to the direct caregivers OQOLI at T2 (1 fective in lessening the burden of care for ter-
week after hospital discharge) and T4 (3 months minal cancer patients and in reducing the nega-
after patients death). At T3, no significant differ- tive effect on the direct caregivers quality of
ence was found between the two groups. The life. For a complete description of the study,
lower number of direct caregivers that could be consult the original source.
analyzed for this time period may explain this.
SOURCE: Summarized from Smeenk, F., de Witte, L., van Haastregt, J., et al (1998). Transmural care of terminal
cancer patients: Effects on the quality of life of direct caregivers. Nursing Research, 47(3), 129213.
gate a particular group but collect data study and that the two groups compared
about the respondents a minimum of two over time be exposed to different types of
times. Panel studies are also referred to employers. A carefully designed panel
as longitudinal studies. Panel studies al- study provided a naturalistic experiment
low the researcher to assess not only and led to a better understanding of the
change over time but also the differences impact of the workplace on workers
and similarities in two or more cohorts or (Clairmont and Jackson, 1980).
groups over time. Harton and Latan (1997) used a panel
study to examine the impact of social in-
uence (dened as popularity) on devel-
a. Rationale for Panel Studies and opment of mature lifestyle attitudes in
Examples ve cohorts of school children, age 9 to
15 years. The main purpose of the study
Why would one choose to do a panel
was to examine the development of ma-
study rather than an experiment or a sur-
ture adolescent attitudes in older chil-
vey? The answer is that experimentation
dren and to look at how these attitudes
is frequently not possible, or judged not
relate to popularity with the same and
to be relevant, for the examination of
other gender. (See Box 4.4.)
some relationships. Surveys, on the other
hand, might not provide a sufcient basis
for making causal inferences. Clairmont b. Tips for Keeping Track of
and Jackson (1980) were interested in Respondents in a Panel Study
studying the impact of working for large
companies rather than small companies. On occasion, respondents are contacted
In this case, the researchers needed to es- at different points in time. When the time
tablish what the individuals were like at between contacts is a year or two and the
time 1; next, the researchers needed to population being studied is fairly mobile,
monitor changes in these individuals over special problems are posed for the re-
time. These are complex issues that researcher. With an initial sample of 538,
quire the individuals being compared to Clairmont and Jackson (1980) were able to
be roughly similar at the beginning of the locate and interview 96 percent of respon-
SOCIAL INFLUENCE AND ADOLESCENT LIFESTYLE ATTITUDES

In this panel study, mature attitudes were de- graders, and 32 seventh graders, subtracting
fined as favorable attitudes toward activities out the graduating eighth graders from year 1),
that adolescents become increasingly inter- representing 62 percent of the total eligible
ested in as they age. Both deviant behaviors population, and 76 percent of those for whom
such as smoking cigarettes and consuming beer permission had been given in year 1.
and wine (illegal for children in the study age Results indicated that approval of both
group) were considered, as well as normal de- healthy and deviant mature lifestyle attitudes in-
velopmental activities such as wearing make-up creased linearly with grade for both boys and
and kissing. The study examined longitudinally, girls. By eighth grade, gender differences in atti-
fourth- through eighth-grade students lifestyle tudes had almost vanished. Students with more
attitudes and sociometric status (popularity), mature attitudes were more popular with the
once each semester. Group-administered ques- opposite gender, but there was no relationship
tionnaires were used to collect data on lifestyle between same-sex popularity and attitudes. In
attitudes and sociometric status. The current this situation, the panel study design enabled
study reports on data collected over 2 years. use of the repeated measures over 2 years to
The study examined grades and gender ef- study attitude measures, to trace the develop-
fects on lifestyle attitudes and popularity. ment of mature lifestyle attitudes over time and
Specifically, it examined whether popularity is simultaneously to test measures of same-sex
related to lifestyle attitudes and whether in- and cross-sex popularity and its relationship to
creased popularity leads to greater adoption of development of mature lifestyle attitudes in five
lifestyle attitudes. Complete data for both se- cohorts of students. On the basis of this inves-
mesters of year 1 were obtained for 201 stu- tigation, the researcher can generate several
dents, and a total of 193 students completed hypotheses about adolescent attitude develop-
both semesters of year 2. Complete data from ment and the role of social influence. The same
all four waves were obtained from 128 students panel design could be used for purposes of test-
(30 fourth graders, 31 fifth graders, 35 sixth ing these hypotheses in future studies.
SOURCE: Summarized from Harton , H.C. & Latan, B. (1997). Social influence and adolescent lifestyle attitudes.
Journal of Research on Adolescence, 7(2), 197220.
dents in the second wave of interviews, connected, the operator may be able
some two years after the initial interview. to provide the new number.
The following techniques were used in 3. Phone directories for the original year
that study to locate the respondents: and for the current one are helpful;
check how the person was listed in the
1. At the time of the rst interview, re- original directory. If the person is still
quest that the respondent provide the in the area, the chances are that the
name of a relative or friend who will al- name will be listed identically.
ways know how to get in touch with him 4. Contact the employer of the respon-
or her. These names prove invaluable dent. Also contact coworkers in an ef-
later in efforts to contact individuals. fort to locate the individual.
2. Try the original phone number; even 5. Contact neighbors. Especially in smaller
though the respondent has moved, the cities and towns, neighbors can be help-
same phone number may have been ful in tracing respondents.
maintained. If the phone has been dis- 6. Now you are desperate! Call people in
the directory with the same last name Sometimes quasi-experimental designs
and hope you will get lucky by nding involve representative samples. In such
a relative of the person involved who studies, a group of individuals (usually
might be able to give you an address or called a panel) is monitored over time. The
phone number. panel may have been selected as a repre-
sentative one and thus the results may en-
able the researcher to make sound causal
E. ADVANTAGES AND arguments than can, at the same time, lead
LIMITATIONS OF to extrapolations to the larger population.
EXPERIMENTAL DESIGNS Quasi-experimental designs thus can some
times fuse strong causal inferences and
The forte of the experiment is in demon- also be strong on external validity.
strating that a treatment variable has an
impact on an outcome measure. To do
b. Limitations
this well, an experimentalist has to rule
out possible sources of contamination, Given limited control over the context in
deal with control variables, and deal with which the study is conducted, quasi-
the effect of random variables. These experimental studies lack the causal in-
challenges are met in greater or lesser de- ference power of the classic experiment.
grees by experimental designs. It is sometimes not possible to control the
treatment variable or to place individuals
1. Pre-Experimental Designs in experimental and control groups. Thus,
these studies often represent a compro-
a. Limitations mise between what the ideal study would
look like and what is feasible.
The pre-experimental design involving a
same group: pretest and posttest discussed
earlier in this chapter had problems in that 3. Classic Experimental Designs
in this design all kinds of other random fac-
tors may have inuenced the change in The classic experimental designs seek to
outcome measures. On the other hand, the remedy the problems of causal infer-
exposed comparison group design was seen ences by establishing randomly matched
to be awed because it could not demon- comparison groups (experimental and
strate that the experimental and control control groups) along with before-treat-
group were similar before the introduction ment and after-treatment measures on
of the treatment. In short, the groups may the dependent variable. These designs
have been different from one another, and are referred to as between-subjects de-
the treatment effect cannot be convinc- signs. A second approach discussed dealt
ingly documented. with problems of causal inference by ex-
posing the same subjects to different
treatments. Control over random vari-
2. Quasi-Experimental Designs ables is achieved in the latter design
through control by constancythe use of
a. Advantages
the same subject in the comparisons.
Quasi-experimental designs are typically
done when the topic dictates that an ex- a. Advantages
periment is not possible. Thus, they are
often the best that can be done given cer- The advantage in both these designs is
tain ethical or practical constraints. that they can provide powerful, convinc-
ing evidence for the effect of a treatment should not conclude that therapeutic
variable. Through statistical analysis and touch performed by nurses of all ages is
through the use of randomization and pre- effective in reducing stress.
cision matching, control over random and
( ii ) Limits to Predictability in
confounding variables is enhanced. A
Experimental Studies
nurse researcher who wishes to assess
The fact that experimental studies do a
whether a given nursing intervention is
better job than surveys in predicting out-
better than some current practice may
comes should not be misinterpreted. For
well turn to an experimental design to
example, suppose that a treatment vari-
shed light on the question. Thus, classic
able X (therapeutic touch) controls 10
experimental designs produce the clearest
percent of the variability in the depen-
view of causation. This is accomplished
dent variable Y (stress level). Ideally, an
through the application of a treatment to
experiment is designed to control all ex-
subjects under conditions in which extra-
ternal factors that inuence the depen-
neous variables are controlled. Further-
dent variable. Surveys can only exercise
more, through the use of double-blind
such control through statistical adjust-
designs, in which neither subjects nor re-
ments. Thus, the ability of experimental
searchers know which subjects are under
and nonexperimental designs to predict
what condition, there is considerable con-
outcomes is not comparable. In principle,
trol over bias on the part of subjects and
experiments should produce higher pre-
researchers. However, there are some lim-
dictability than that achieved by non-
itations of experimental designs.
experimental designs using the same
variables. A robust variable in an experi-
ment, however, may be relatively impo-
b. Limitations tent outside the laboratory setting.
( i ) External Validity ( iii ) Articiality Issue
Few experiments are conducted using The third major concern is that there is
samples that can be viewed as represen- an element of articiality in laboratory
tative of a larger population; therefore, experiments that is difcult to detect, in-
one cannot make extrapolations to the terpret, or control. In the nursing world,
general population. This is the issue of ex- the natural environment in which the
ternal validity. In short, can a study on a client is situated bears heavily on health
single ward in Kansas City be extrapo- outcomes. Controlling all factors, as you
lated to wards across the country? do in a classic experiment, may not yield
Clearly one would need to repeat the results that are applicable in the real
study in many different locales to deter- world, where you are unable to control
mine the extent to which valid general- the same set of factors.
izations may be drawn from a single In addition, a good subject may try to
study. In our therapeutic touch experi- respond by giving the researcher what he
ment, signicant differences were found or she thinks the researcher wants. The
in the stress levels of subjects who re- subject may also be intimidated and act
ceived therapeutic touch. These differ- atypically. To what extent do people
ences, however, may well hold true only modify their behavior when they know
for older nurses experienced in thera- that they are being studied? Although
peutic touch protocols but not for more is said about this topic in Chapter 9,
younger ones (all of the nurses in the test researchers need to take into account the
were older than age 40 years). Thus, we effects of expectations on the responses
Table 4.5 Advantages and Limitations of Alternate Designs

Research Design
Category: Experimental
Designs General Validity Causal Inference Multivariate Probing
Pre-experimental
Experimental
Quasi-experimental
Panel studies
In each category, a plus sign means that this is an advantage of the technique; a minus sign means this is a possible
limitation; a plus or minus sign () means that in some conditions it is a limitation, but in others, it is an advantage.
General refers to the extent to which extrapolations to larger populations may be made using each of the design or
data collection procedures.
Validity is the extent to which indicators clearly measure what they are intended to measure.
Causal Inference refers to the ease with which inferences about causal relations among variables may be made.
Multivariate refers to the ease with which information on many variables is collected, leading to the possibility of
multivariate analysis.
Probing refers to the extent to which responses may be probed in depth.
of subjects. Because subjects can man- (v) Experiments Cannot Study All
age their presentation of themselves to Topics
the experimenter, questions can be It is neither practical nor ethical to ex-
raised about the validity of some experi- periment on all aspects of human behav-
mental results (e.g., are you measuring ior; therefore, other designs remain im-
what you claim to measure?). portant tools for nurse researchers. For
example, in a study exploring the effec-
( iv) Limitations on Number of tiveness of a new birth control method, it
Variables would be impractical and entirely uneth-
For practical reasons, experimental de- ical to randomly assign half the clients at
signs can only deal with a few variables si- a well womens clinic to use no birth con-
multaneously. (Multiple-variable designs trol while the other half would be as-
are possible, but with several treatment signed to use the new method. Beginning
and control levels, one would soon need researchers should have some under-
far too many subjects to run the experi- standing of experimental designs be-
ment. For example, a 3 3 3 3 design cause their great strength is their ability
produces 81 treatment conditions and, if to control many of the confounding fac-
there were 10 subjects per condition, tors that inuence outcomes. When ex-
would require 810 subjects.) When a perimentation is not possible, it is more
large number of variables require simul- difcult to produce convincing infer-
taneous analysis, it is advisable either to ences about causal relationships. Table
use a different experimental design or to 4.5 summarizes the advantages and dis-
simplify the experiment by using fewer advantages of experimental and quasi-
variables or fewer levels in each variable. experimental designs.
E X E R C I S E S
1. Suppose you wished to investigate the 4. Outline a quasi-experimental project.
impact of a school nutrition program Be sure to identify all variables and in-
on the grade performance of students. dicate how the dependent variable will
What research design or designs be measured. How will control over
would be appropriate to deal with possible sources of contamination be
such a question? Outline the rationale achieved?
for your choice.
5. Review a current nursing research

2. Outline a within-subjects experimental journal and select a pre-experimental,
project. Be sure to identify all variables experimental, or quasi-experimental
and the number of levels that will be study of interest to you. Identify the
used and indicate how the dependent type of design and the major steps in
variable will be measured. Comment the design. Also, compare the advan-
on the method by which controls will tages and disadvantages of the design.
be achieved on possible contaminants. Formulate design remedies to deal with
potential disadvantages or weaknesses
in the design.
3. Outline a between-subjects experi-
mental project. Be sure to identify all
variables and the number of levels that
will be used and indicate how the de-
pendent variable will be measured.
How will control over possible sources
of contamination be achieved?
Brink, P.J., and Wood, M.J. (1998). Advanced nal of Nursing Research, 20(1), 119135.
Design in Nursing Research (2nd ed). Lon- Ciliskas experimental design compares
don: Sage. This advanced text explores two non-dieting interventions to the prob-
three levels of research based on knowl- lem of obesity in women.
edge of the topic: experimental, survey, Mitchell. M., and Jolley, J. (1997). Research De-
and exploratory-descriptive. The text be- sign Explained (3rd ed). Orlando: Harcourt
gins with the experimental design and sub- Brace Jovanovitch College Publishers. This
sequent designs are contrasted with the is an excellent text that explains experi-
classic experiment. mental design and the analysis of results.
Campbell, D.T., and Stanley, J.C. (1966). Exper- Smeenk, F., de Witte, L., van Haastregt, J., et al.
imental and Quasi-Experimental Designs for (1998) Transmural care of terminal cancer
Research. Chicago: Rand McNally. This is patients: Effects on the quality of life of di-
considered by many to be the classic state- rect caregivers. Nursing Research, 47(3),
ment on confounding variables in different 129136. An example of a quasi-experimen-
types of research designs. tal design testing the hypothesis that trans-
Ciliska, D. (1998). Evaluation of two nursing in- mural care may lead to a better quality of life
terventions for obese women. Western Jour- for the direct caregivers of cancer patients.
Chapter 5
Survey Designs
CHAPTER OUTLINE
A. A Short History of the Survey E. Comparative Studies

B. The Rationale for Surveys 1. Examples of Comparative Nursing
1. Appropriateness of Surveys Studies
2. Validity of Surveys 2. Challenges in Comparative Research
C. Steps in Survey Research F. Secondary Data Analysis
D. Guidelines for the Administration of 1. Purpose
Survey Questionnaires 2. Sources of Data
1. General Rules for the Administration of 3. Challenges in Conducting Secondary
Questionnaires Data Analysis
2. Individually Delivered Questionnaires 4. Examples of Research Using Secondary
3. Group-administered Questionnaires Data Analysis
4. Mailed Questionnaires G. Advantages and Limitations of Survey
5. Phone Surveys Designs
6. Interviews
KEY TERMS
Comparative research studies Meta-analysis Retrospective study

Cross-sectional Panel study Salience of topic
Ex-post facto study Phone survey Sample
Group-administered Population Secondary data analysis
questionnaire Quality control monitoring Structured interview
Individually delivered system Survey design
questionnaire Questionnaire Unilineal model
Interview schedule Response rate
Mailed questionnaire
135
A
survey design is one in which a review the ndings of many researchers on
searcher collects information from some topic, frequently use survey analyti-
respondents on some topic at one cal techniques; therefore, this category of
point in time. It is similar to a snapshot of research is also included in this chapter.
the phenomenon under investigation. Un-
like an experiment, which involves some
treatment and includes measures both
before and after the treatment, survey A. A SHORT HISTORY OF THE
designs do not include an experimental SURVEY
manipulation. Instead, surveys measure
many variables, and statistical techniques Surveys have been around for a long
are used to make inferences about causal timethere are biblical references to
relations among variables. A survey design the counts of the children of Israel, and
is similar to the pre-experimental exposed censuses were taken of the population of
or comparison group design reviewed in the ancient Egypt. However, Charles Booth,
Chapter 4. Much nursing research falls into who conducted three major surveys,
this category. laid the foundation for the modern sur-
The most common forms of the survey vey in 19th century England. Booth was
design include censuses, polls, and a concerned with providing an accurate
whole range of situations in which respon- count and description of poor people liv-
dents are asked to provide answers to a ing in London. He was the first to work
set of xed questions. A survey design is out operational definitions (the indica-
generally used to collect data from large, tors he used to measure poverty); the
representative samples on the prevalence, first to draw color maps to reflect the so-
distribution, and interrelationships of vari- cial characteristics of an area; and the
ables within the sample. The variables of first to attempt to show how variables
interest can be classied as knowledge, at- were related to one another, thus begin-
titudes, attributes, opinions, or behaviors. ning efforts to understand the associa-
Survey data are typically collected by tion between social variables.
means of questionnaires or interviews. Ex- More recently, survey research devel-
amples of survey topics include lifestyle oped in this century partly through the ef-
practices of middle-aged adults, health- forts of pioneer pollsters George Gallup
care services accessed by disadvantaged and Elmo Roper to provide an accurate
women, nutrition practices of female ado- prole of Americans, partly from efforts of
lescents, factors inuencing public opin- market researchers to understand con-
ion on euthanasia, health-care needs of sumer behavior better, and partly from
clients with AIDS, attitudes of high school the interests of journalists, government
students toward abortion, and family plan- agencies, and political organizations, all
ning practices in low-income communi- of whom understood the advantages that
ties, to name a few. could be gained if one could accurately
Survey designs are sometimes used to gauge public opinion. Survey research
do comparative studies between coun- was just too powerful a tool to be left to
tries or other groupings, and these gener- scholars alone.
ally rely on secondary data (these are data During the postWorld War II era,
collected by other researchers). But be- Samuel A. Stouffer and Paul F. Lazarsfeld
cause the techniques used to analyze such did much to make survey research a legit-
data are modeled after survey research, imate academic and practical pursuit. The
they are discussed in this chapter. In ad- Peoples Choice (Lazarsfeld, Berelson, and
dition, meta-analysis researchers, who re- Gaudet, 1948) was a sophisticated analy-
SURVEY DESIGNS 137
sis of voting intentions and behavior. It up of a series of set questions and either
was the rst study to interview members provides a space for an answer or offers
of a panel several times leading up to an a number of xed alternatives from which
election. The Peoples Choice marks the the respondent makes a choice. Ques-
beginning of voting studies in political sci- tionnaires can be completed in group set-
ence and was the rst study to take con- tings, mailed to respondents, or read to
trol variables into account systematically. respondents by interviewers, either over
From the 1950s onward, survey re- the phone or in person.
search became a key approach in nursing As indicated in Chapter 1, surveys are
and several social science disciplines. typically associated with the positivist
Many universities support survey re- approach to knowledge. But it is to be
search centers. These centers typically noted that both Marx and Weber used
offer consulting services and coordinate surveys. Marx developed a questionnaire
surveys being conducted by university to measure French worker experiences
scholars. They are excellent resources for (using the conict approach; cited in Bot-
professionals, community groups, and be- tomore, 1988), and Weber surveyed atti-
ginning researchers who need assistance tudes toward work (using the interpre-
with aspects of survey design. tive approach).
In the new millennium, survey designs Surveys can be a relatively inexpensive
will remain particularly important to so- method of collecting much information
cial, health, and nursing research investi- from a large number of people. Survey re-
gations because they can economically searchers frequently draw a sample to
provide a wealth of surprisingly accurate make estimations about some population.
data on a wide range of phenomena im- As used by researchers, the term popula-
portant to professionals working in these tion refers to a collection of individuals,
areas. Currently, academic researchers communities, or nations about which one
in many disciplines, including psychol- wishes to make a general statement. To
ogy, political science, sociology, educa- save time and money, the researcher
tion, and business, also commonly use draws a sample from the population,
surveys. Public opinion pollsters, com- which is interpreted to represent the
munity groups, and professionals doing population. Although including the whole
evaluations of applied programs all make population could prove to be more accu-
frequent use of survey methods. rate (as in a census), the costs may be pro-
This chapter discusses the major types hibitive. If a researcher wishes to make ex-
of surveys and provides practical guide- trapolations from a sample to a larger
lines for using them. Let us rst examine population, then a fairly large sample is re-
the rationale for surveys to see how it quired; therefore, it is likely that a survey
contrasts with that of experiments. design will be used. Because human be-
havior is often highly complex and subject
to the simultaneous impact of many vari-
B. THE RATIONALE FOR ables, a research strategy that measures
SURVEYS many variables simultaneously is often ap-
propriate.
Survey designs often involve collecting
information to describe, compare, or ex- 1. Appropriateness of Surveys
plain knowledge, attitudes, attributes, or
behaviors. To do so, researchers fre- Surveys are appropriate for descriptive
quently have respondents complete a and correlational studies. Unlike experi-
questionnaire. A questionnaire is made ments, in which researchers are interested
in demonstrating a link between a treat- rst diagnosed?) And because question-

ment and an outcome of that treatment, naires probe into sensitive areas, they
the survey presents a picture of events, permit respondents to manage their re-
people, and phenomena as respondents sponses so they can appear in a favorable
report them. Thus, in a survey, a re- light. Suppose one was attempting to
searcher explores the prevalence and in- measure attitudes toward minorities. In
terrelationships among variables in a pop- such cases, one has to understand that
ulation without introducing a treatment some people may try to appear toler-
variable. This is again in contrast to exper- antperhaps more tolerant than they ac-
iments, in which researchers introduce a tually are. There is not a one-to-one rela-
treatment variable (e.g., a low-fat diet ver- tionship between what people say they
sus a high-fat diet) that is manipulated and believe and how they actually behave
then measure its effect on another variable when confronted with real situations. For
(serum cholesterol level). Notice that the example, adolescents who are concerned
same research question, on the relation- about public perceptions or negative
ship between a low-fat diet and serum cho- consequences of behavior may underes-
lesterol levels, can be studied using a sur- timate their engagement in unhealthy
vey. Such a survey may include a large lifestyle practices such as consumption
number of participants who already con- of alcohol or drug products, or con-
sume a low-fat diet and a similar number versely, overestimate their engagement
from a population known to consume in healthy practices on surveys to make a
foods high in fat content. Results of data favorable impression.
analysis may indicate that those who con- An initial eld study conducted by
sume foods high in fat content have a Richard T. LaPiere (1934) highlights the
higher serum cholesterol level than those discrepancy between what people say
who consume low-fat diets. In this situa- and what they do. In his classic work, he
tion, the researcher did not introduce par- showed that only one overnight lodging
ticipants to a low-fat diet or manipulate the establishment out of 251 directly refused
amount of daily fat intake; the researcher accommodations to a Chinese couple.
simply investigated the relationship that Yet when LaPiere asked the same busi-
occurred after fat was consumed. This nesses some 6 months later in a mailed
type of survey is known as a retrospective questionnaire whether or not they would
or ex post facto study, meaning that after accept as guests members who were Chi-
the fact, the researcher attempts to link nese, only one business indicated that it
present events (serum cholesterol) to past would. Deutscher (1966) also reminds us
events (amount of fat in diet). that the relationship between attitudes
expressed on a questionnaire and subse-
quent behavior can be problematical
2. Validity of Surveys (Box 5.1). Sometimes there may be little
relation between the two. This is an im-
With both experimental and survey de- portant point for nurses and other health
signs, there can be difculty in establish- professionals to consider if they are de-
ing the validity of the measures used. In pending on the results of survey data to
the case of surveys, respondents are guide program planning, design, and eval-
asked to report their own attitudes, be- uation.
havior, and backgrounds. Some of the Nonetheless, questionnaires can pro-
data requested require respondents to duce reliable and valid responses on
recall episodes from their pasts (e.g., many issues. And if one wishes to mea-
How anxious were you when you were sure attitudes, there is no good alterna-
SURVEY DESIGNS 139
about a variety of points in time, such as

the year of the respondents birth, the
type of community the respondent grew
up in, or how old the respondent was
when getting his or her rst full-time job.
Once again, although problems do exist
with recall data, sometimes there is no
practical alternative for measuring some
variables.
Questionnaires are restrictive be-
cause they can only be used with a liter-
ate population. The wording of questions
must be straightforward so that allor
Image/Text rights unavailable at least mostof the respondents are
able to handle the language. Also, the re-
quirement that all respondents be able
to understand the questions prevents
certain areas from being probed in
depth. Everyone is given the same set of
questions, and although it is advanta-
geous to have all respondents reply to
the same question, it does mean that in-
teresting responses cannot be pursued.
For in-depth probes, a personal inter-
view is necessary.
Students frequently conduct surveys.
Box 5.2 lists some recent studies com-
pleted by nursing and sociology stu-
dents under the supervision of the au-
thors of this text. As can be seen by the
titles of the projects, many topics can be
addressed. The questionnaires were de-
signed and the data collected in consul-
tation with the instructor. Many of these
questionnaires may be seen by checking
the Web site http://www.stfx.ca/people/
wjackson. By clicking on Questionnaires,
some 160 project questionnaires may be
tive to asking people about them. One viewed. Appendix B includes a question-
simply has to live with the problems of naire designed by nursing students.
measurement if the variables are to be This chapter provides a general orien-
measured at all. According to conven- tation to survey research. Anyone who
tional wisdom, survey designs are consid- wishes to know how to carry out a survey
ered cross-sectional studies in that they needs to consult two additional chapters
provide point-in-time data and are there- in the book: Chapter 14 provides guide-
fore poor at measuring changes over lines for developing a questionnaire, and
time. Although it is no doubt risky to as- Chapter 15 discusses how one goes about
sume that people will recall their past re- selecting a sample and determining how
liably, most questionnaires include items large it should be.
BOX 5.2 Student Surveys

The following is a list of recently completed sur- mitted the best paper in the research team is
veys conducted by sociology or nursing stu- identified as the author. Copies of the ques-
dents. Each project was designed by a research tionnaires may be viewed at the Web site http//
team in consultation with the instructor in a re- www.stfx.ca/people/wjackson. Click on Question-
search methods course. The student who sub- naires.
Project Author
Premarital Sex Among University Students Michelle Chisholm (1999)
Contraceptive Usage Among Males Andrea Brophy (1999)
The Effect of Program of Study on Health Christina Vanhorn (1999)

Seeking Behaviors
Attitudes Toward Voluntary Active Euthanasia Donna Beiswanger and Cathy Richard (1999)
and Physician-Assisted Suicide: Are Nursing
Students Supportive of These Practices?
Experiences of Sexual Harassment Among Jan Murray (1999)

Females Living on Campus
The Self-Acceptance of Homosexuality Michelle Murray (1999)
Occupational Ranking: Prestige and Cultural Ray MacIssac (1999)

Diffusion
The Consequences of Shiftwork: Physical, Michelle Chisholm (1998)

Psychological, Social, and Economic
Health Promoting Behaviors at St. Francis Sarah LeBlanc (1998)

Xavier University
Alcohol Consumption of University Students Alison Fisher (1998)
Attitudes Toward Women in Competitive Alanna MacNeil (1998)

Sports
The Punitiveness of Students at St. Francis Kristine Cameron (1998)

Xavier University
The Trends and Variations in Smoking Amy Gillis (1998)
Illegal Drug Use Among Young Adults within a Wendy Chisholm (1998)
University Setting
Eating Disorders Linda Jane Liutkus (1998)

SURVEY DESIGNS 141
BOX 5.2 Student Surveys (Continued)
Project Author
Nurses Attitudes Toward Student Nurses Shauna M. Grant (1997)
Students Attitudes Toward Campus Health Tanya M. LeBlanc (1997)

Care Facilities
Homesickness Among First-Year University Elsa Arbuthnot and Laura Rogal-Black (1997)
Students
Illegal Drug Use at University Tracey Pye (1996)
Perception of Risk of Sexual Assault Among Alicia Van De Sande (1996)

Freshman Women at St. F. X. University
Suicidal Ideation on Campus Laurie Fraser (1996)
Attitudes Toward Euthanasia Tracey MacMillan (1996)
The Effects of Gender on Self-Esteem Kathleen Lumsden (1996)
C. STEPS IN SURVEY RESEARCH 6. Develop sampling procedures to se-

lect a representative and appropri-
There are various types of survey re- ately sized sample (see Chapter 15).
search. The main differences are deter- 7. Design data collection procedures.
mined by how the data are collected. The 8. Plan for data analysis (see Chapter 16).
range may include face-to-face inter- 9. Pilot test the approaches to data col-
views; phone interviews; individually de- lection and data analysis procedures.
livered, group, or mailed questionnaires; 10. Modify the procedures as necessary.
and secondary data studies. Regardless 11. Collect and analyze the data (see
of the data collection method used, the Chapters 16, 17, and 18).
steps in survey research are the same. 12. Write descriptive, comparative, or
Wilson (1993) outlines a simple series of evaluative ndings and draw conclu-
steps to follow in survey design. sions (see Chapter 19).
1. Formulate the research question.
2. Determine that the survey is the ap-
propriate design to address the re- D. GUIDELINES FOR THE
search question. ADMINISTRATION OF
3. Select the type of survey to be used SURVEY QUESTIONNAIRES
(e.g., interview, questionnaire, phone
survey). In this section, general rules are provided
4. Translate the objectives of the survey for administering questionnaires in a va-
into categories of questions or items riety of ways. In all cases, the suggestions
(see Chapter 14). should be used with common sense be-
5. Identify the population of partici- cause there are times when they should
pants or settings (see Chapter 15). be violated. A summary account of a
nursing survey study is provided in Box be provided; otherwise, the researcher

5.3 and comments on the application of makes it less likely that the respon-
administration guidelines as they relate dents will cooperate in the future.
to this study are noted. Rule 4. Pay respondents. When reason-
Permissions from parents, school ad- able and nancially possible, pay
ministrators, and ethics review commit- respondents for their time and coop-
tees are typically required in order to dis- eration. The fee helps establish the
tribute questionnaires; researchers need legitimacy of the study as well as a rec-
to take this fact into account in planning iprocal relationship with the respon-
surveys. Pursuing permissions takes time dent. Such payments appear to have a
and usually requires a statement of the modest impact on the willingness of re-
problem under investigation and a copy of spondents to participate (Heberlein
the proposed questionnaire. Researchers and Baumgartner, 1978). Payments
must plan for extra time. help establish reciprocity between re-
searchers and respondents and help
to avoid the respondents perception
1. General Rules for the that they have been ripped off for
Administration of their data.
Questionnaires Rule 5. Do not pressure respondents to
participate. Although researchers have
The following rules are intended to in-
powerful interests in getting everyone
crease response rates for all types of sur-
selected to complete the survey, it must
veys. Later sections suggest approaches
be indicated that although cooperation
for particular types of surveys.
in completing the questionnaire or in-
Rule 1. Establish legitimacy. Establish terview is appreciated, it is, nonethe-
the legitimacy of the research by not- less, optional. Particularly in face-to-
ing who is sponsoring it, describing face encounters, unintended pressure
why it is being done, and presenting it may be placed on individuals to partici-
in such a way as to make it seem credi- pate in the study. In the case of ques-
ble and competent. The cover letter tionnaires administered to a gathering
and the questionnaire must look proof individuals, there may be consid-
fessional. erable informal pressure from peers in
Rule 2. Keep it simple. Keep question- the room to complete the survey. Re-
naires, interview guides, and phone in- searchers must exercise self-discipline
terviews as simple and as nonthreaten- and avoid putting undue pressure on
ing as possible. Questionnaires should individuals in an attempt to coax par-
be easy to respond to and should avoid ticipation. (See the discussion on re-
asking questions that pry unnecessar- search ethics in Chapter 10.)
ily into the respondents personal af- Rule 6. Do not reveal research hypothe-
fairs. (Chapter 14 provides additional ses to those involved in a study. Gen-
ideas for making questionnaires easy erally, it is not advisable to inform
to complete.) respondents, interviewers, or data col-
Rule 3. Provide a report to the respon- lection assistants about the hypothe-
dent. When individuals are to be inter- ses of the study. Interviewer ex-
viewed more than once during the pectancy bias (see Chapter 9) may be
course of a study (as in a panel study), reduced if the hypotheses are not
report ndings to respondents. In all known. Respondents who are made
cases in which a report has been aware of the research hypothesis may
promised to the respondents, it must also try to help the researcher by
SURVEY DESIGNS 143
WILL EVIDENCE-BASED NURSING PRACTICE MAKE PRACTICE

PERFECT?
Little is known about the kinds of evidence that riential followed by nursing school, workplace
nurses use in their practice. This study explores sources, physician sources, intuitions, and what
the sources and types of knowledge on which has worked in the past. Literature, either in
nurses base clinical decisions. Important discov- textbook form or journal articles, was found in
eries are revealed about both the nature and the bottom five for frequency.
structure of nursing knowledge as it is applied in This survey study has important implications
practice. This information is important to the for nursing. First, it suggests that nurses use a
profession as it strives to adopt evidence-based broad range of practice knowledge, much of
nursing (EBN), and base decisions on empirical which is experientially based rather than re-
evidence. search based. Secondly, we can infer that tradi-
The Alberta Association of Registered tional scientific journals are not very effective as
Nurses (AARN) membership list was used to dissemination vehicles. A third issue is that ba-
draw a random sample of 1500 nurses for a sic nursing education continues to play an im-
cross-sectional mail survey. All participants portant ongoing role as a source of practice
were required to be actively engaged in the de- knowledge for nurses.
livery of nursing care. Following initial receipt of
Comments on Mailed Questionnaire
the survey instrument reminders were mailed
at approximately 3, 6, and 9 weeks. With the 1. This survey provided a relatively cheap way
six-week reminder a replacement question- of contacting a large number of nurses.
naire was included. The final sample consisted 2. A cover letter was included to explain the
of 600 (40%) questionnaires. The sample of survey to the respondents.
nurses was representative of the population 3. The survey topic was likely to be salient and
from which it was drawn based on a compari- of interest to the participants.
son of it with demographic data about nurses in 4. The questionnaire was brief, easy to respond
Alberta. to, and professionally presented.
The 16-item survey questionnaire developed 5. Additional contacts were made by the re-
for this investigation was adapted from the 12- searcher with the participants at 3, 6, and 9
item Research Utilization Questionnaire (Baess- weeks after the original mailing to increase
ler et al., 1994). The 16 questions began with the the response rate.
transition statement: The following questions 6. A replacement questionnaire was included
relate to the kind of knowledge you use in your at week 6 to nonrespondents.
nursing practice. A pilot study was conducted on 7. A 40% response rate was achieved and the
a convenience sample of 23 postbaccalaureate researcher noted that a comparison of the
and masters level nursing students. sample with the total population of 15,000
Mean scores were reported for each item nurses on major demographic and related
and scores were ordered from most to least variables suggests that the sample was rep-
frequent. Results indicated that the two most resentative of the population from which it
frequently used knowledge sources were expe- was drawn.
SOURCE: Summarized from Estabrooks, C.A. (1998). Will Evidence-Based Nursing Practice Make Practice Perfect?
Canadian Journal of Nursing Research, 30(1), 1536.
skewing their responses in a direction the presentation can be improved. The

to favor the hypothesis. Respondents quality of ones research can be no bet-
have the right to know why the study ter than the quality of the data col-
is being done and who is sponsoring it. lected. Monitor the process carefully.
Rule 7. Do quality control spot checks. It
As previously discussed, there are
is critical to do spot checks to ensure
many variants of the survey, but only the
that administrative procedures are be-
major approaches are explored here.
ing followed. Research directors are
Each method has a distinctive set of ad-
sometimes negligent on this point.
vantages and disadvantages, as well as
Conducting interviews and handing
rules that apply to its administration. Let
out questionnaires is not many peo-
us begin with the questionnaire that is
ples idea of fun. Research assistants
personally handed to a respondent.
occasionally cut corners; these range
from ignoring the random sampling
procedures that should be used to se- 2. Individually Delivered
lect which person in a household is to Questionnaires
complete the questionnaire to invent-
ing respondents and their answers. Individually delivered questionnaires
Checks can be run on the representa- are delivered to a respondent by a re-
tiveness of the sample (to see how well searcher. A brief explanation is offered,
the respondents and their answers questions are answered, and arrange-
match known characteristics of the ments are made for the return of the com-
target population) or to see if the per- pleted questionnaire. This method of
son who was supposed to have been handing out questionnaires is typically
interviewed actually was. However, by used in community surveys in which the
the time checks are run, eld research form is dropped off at selected houses, in
funds may be expended and you may college dormitories, where questionnaires
not be able to redo the work. And if any are handed to selected respondents (usu-
data have been falsied, it will take a ally the researcher tries to obtain a sys-
lot of time to distinguish the genuine tematic sample of campus dormitory
data from the bad. Be cautious of in- rooms), and in studies of organizations
terviewers who are doing much more (e.g., surveys of hospital staff, university
than other interviewers. Watch out for faculty, or employees of a private rm) in
systematic differences in response which the target respondents are ap-
rates to sensitive questions. Some- proached individually.
times, for example, interviewers are In cases in which a survey of a sys-
too embarrassed to ask about some- tematic sample of patients in a hospital
ones income, so they preface their is being conducted, care should be
question by saying, You dont have to taken to provide everyone with an equal
answer this one if you dont want to, chance of participating in the survey if
but could you tell us your family in- there is a mixture of single and double
come last year? By giving this cue to rooms. This means that, in the case of
the respondent, it is easier for the re- double rooms, both patients should be
spondent to say, Id rather not answer asked to complete the questionnaire. In
that. So, if an individual interviewer is this way, all patients have an equal
missing data more often than other in- chance of being selected to participate
terviewers, try to go through the part in the survey. (In this case, rooms are
of the questionnaire affected to see if the units being selected. See more de-
SURVEY DESIGNS 145
tails on the systematic sampling proce- tion. It is critical to record where ques-
dure in Chapter 15.) tionnaires have been dropped off and
when they are to be picked up. Pick up
the questionnaire on time; respondents
a. Rules for Individually Delivered
will be annoyed if it is not picked up. A
Questionnaires
form for recording this information
should be developed and then used
In addition to the general rules for con-
faithfully (Table 5.1). Besides informa-
ducting surveys (see previous section),
tion regarding place and time, the form
there are some rules that should be con-
should also have space to list dates
sidered in situations where respondents
when an attempt was made to contact
are approached on a one-to-one basis.
the individual and to record times when
Rule 1. Make personal contact with the it would be convenient to return to
respondent. When feasible, contact re- meet the person. After the data have
spondents in person to explain the sur- been collected, these sheets will prove
vey and let them know when you will invaluable in calculating the response
pick up the completed form. In a door- rate to the survey and in identifying the
to-door survey, it should be possible to problems that were encountered.
get more than 80 percent to agree to Rule 4. Provide envelopes to help main-
complete the form. Avoid having third tain privacy. Generally it is a good idea
parties handing out your question- to provide respondents with an enve-
naires; a member of the research team lope into which they can seal their
can better explain the survey and an- completed questionnaires. Sometimes
swer questions that might be raised. In questionnaires are left around waiting
particular, avoid having teachers, work- to be picked up, and unwanted eyes
ers supervisors, or coworkers hand out may peruse the responses. A sealed
questionnaires. The extra effort needed envelope foils most snoopers. In intro-
to have a member of the research team ducing the survey to the respondent,
hand out the questionnaires avoids the researcher can indicate that the
many problems associated with a third- envelope is a means of protecting the
party delivery. In short, the greater the anonymity of the respondent. This
personal contact, the greater the re- makes respondents feel safe.
sponse rate. Rule 5. Use a slotted return box. To help
Rule 2. Avoid mailed and drop-box re- convey the sense of anonymity, it is a
turn methods. When possible, avoid good idea to use a box with a slot cut
mailed returns or having respondents in one end (a box measuring 9 12
drop their completed questionnaires is a good size), and respondents ques-
into a box in a dormitory, staff ofce, or tionnaires can be slipped into the box
other convenient spot. If at all possible, as they are returned. In especially sen-
the researcher should pick up com- sitive studies, this return procedure
pleted forms at a time agreed to with can be pointed out when the question-
the respondent. Such arrangements naire is delivered to the respondent.
encourage respondents to complete
the forms by the prearranged time. Do 3. Group-administered
not be tempted to violate this rule; if Questionnaires
you do, you will pay a heavy price in
lost and missing questionnaires. Group-administered questionnaires al-
Rule 3. Record place and time informa- most always have good response rates.
Table 5.1 Sample Drop-Off Form

# Address Try 1 Date Try 2 Date Try 3 Date Pick Date Pick Time Done Notes
1 ________ ________ ________ ________ ________ ________ ____ _____

2 ________ ________ ________ ________ ________ ________ ____ _____
3 ________ ________ ________ ________ ________ ________ ____ _____
4 ________ ________ ________ ________ ________ ________ ____ _____
5 ________ ________ ________ ________ ________ ________ ____ _____
6 ________ ________ ________ ________ ________ ________ ____ _____
7 ________ ________ ________ ________ ________ ________ ____ _____
8 ________ ________ ________ ________ ________ ________ ____ _____
9 ________ ________ ________ ________ ________ ________ ____ _____
10 ________ ________ ________ ________ ________ ________ ____ _____
11 ________ ________ ________ ________ ________ ________ ____ _____
12 ________ ________ ________ ________ ________ ________ ____ _____
13 ________ ________ ________ ________ ________ ________ ____ _____
14 ________ ________ ________ ________ ________ ________ ____ _____
15 ________ ________ ________ ________ ________ ________ ____ _____
16 ________ ________ ________ ________ ________ ________ ____ _____
17 ________ ________ ________ ________ ________ ________ ____ _____
18 ________ ________ ________ ________ ________ ________ ____ _____
19 ________ ________ ________ ________ ________ ________ ____ _____
20 ________ ________ ________ ________ ________ ________ ____ _____
This form of administration involves convenience; they cannot be taken as

handing out questionnaires to an assem- representative of some larger population.
blage of personsbe it a class, at a public
meeting, or any other setting where peo-
ple can be asked simultaneously to com- a. Rules for Group-administered
plete a questionnaire. There is consider- Questionnaires
able informal pressure (from peers in the
In addition to the general rules for the ad-
room) to cooperate with the researcher
ministration of surveys, there are some
and, normally, between 90 and 100 per-
rules that apply to situations in which a
cent of potential respondents will com-
researcher distributes questionnaires to
plete questionnaires in group settings.
an assembled group.
One caution when using group-admin-
istered questionnaires is that typically Rule 1. Indicate the voluntary nature of
probability sampling procedures (see survey. Researchers should acknowl-
Chapter 15) are not used and, therefore, edge this informal pressure and be sure
the data collected cannot be used to ex- to inform the potential respondents of
trapolate to some larger population. their complete freedom to refuse to an-
What this means is that the groups se- swer any or all questions; participation
lected are usually done on the basis of is voluntary. The person administering
SURVEY DESIGNS 147
the questionnaire can briey explain generally less disruptive to administer

what it is about and be available to an- them at the end of the session.
swer any questions that may be raised. Rule 5. Take steps to identify bad ques-
Rule 2. Arrange well in advance. It is fre- tionnaires. One hazard to watch out
quently necessary to gain permission for is that a few within the room may
from the person in charge of the group decide to make a joke of the question-
meeting to have a questionnaire ad- naire and start making silly responses.
ministered, so it is a good idea to make This happens rarely, but watch out for
arrangements well in advance. And such responses. Normally any ques-
just before the questionnaire is to be tionnaires that appear not to have
administered, it is also a good idea to been taken seriously are reviewed by
remind the person concerned that you the research team and withdrawn if
will be coming. judged to be frivolous.
Rule 3. Explain the survey to those pres-
ent. The researchers should explain
who is doing the research and why it is 4. Mailed Questionnaires
being done, and respondents should
be encouraged to ask any questions Although researchers try to avoid doing
about the survey in general or about so, often there is no choice but to use the
particular questions. mail system. Long-distance phone inter-
Rule 4. Administer the questionnaire at views may be too expensive and travel
the end of the session. For practical rea- costs would quickly eat up a research
sons, it is usually best to administer budget if the researcher attempted to de-
questionnaires at the end of a meeting, liver the questionnaires by hand.
group session, or class, rather than at Mail surveys are popular because they
the beginning. If, for example, one goes provide a relatively cheap way of con-
into a classroom with a questionnaire at tacting a large number of respondents.
the beginning of a class, problems will And despite the reputation mail surveys
arise because not all the students will have for producing low response rates, it
nish at the same time. Hence, from the is possible to have the majority of ques-
teachers point of view, valuable class tionnaires returned. The response rate
time will be wasted as the researcher measures the percentage of delivered
waits for the last forms to be completed. questionnaires that are returned. In mail
Similarly, at a meeting, avoid handing surveys, we normally deduct from the to-
out a questionnaire before the meeting tal number of questionnaires sent out the
begins; administer it at the end, or number that are returned because of an
before a break during the meeting. This incorrect address. In other words, the
allows people who work at different number of delivered questionnaires is
speeds to complete the questionnaire equal to the number sent out minus the
without feeling rushed. However, ad- number returned as undeliverable.
ministering a questionnaire at the end Because our major concern with the
of a meeting or class probably loses a mail survey is the response rate, we will
little in terms of the quality of replies. At consider the factors that inuence whether
the end of sessions, respondents may a questionnaire will be returned. John Goy-
be tired or bored and wish to leave as der has done extensive work in examining
soon as possible. But, given the re- survey research response rates. Using a re-
searchers desire to maintain cordial gression-based method of analysis (see
relations with those giving permission Chapter 18), he developed a nine-variable
to administer the questionnaires, it is model for predicting nal response rate
(Goyder, 1982, 1985a, 1985b; Goyder and letter should be included that explains
Leiper, 1985c; Box 5.4). Two factors are in- the survey to the respondents. The legiti-
volved: those largely beyond the control of macy of the survey is enhanced if the
the researcher and those the researcher questionnaire is well presented, the spon-
can control. soring agency identied, and the worthi-
ness of the research established.
Among the variables examined, the ev-
a. Factors Beyond the Control of the idence indicates that monetary incentives
Researcher do increase response rates. Follow-up
contacts in the form of letters, postcard
These factors are of interest in trying to
reminders, registered mail, and long-
predict the likely response rate to a
distance phone calls all enhance the like-
mailed questionnaire. The type of respon-
lihood of a positive response. However,
dent receiving the questionnaire is im-
with each contact, one can expect slightly
portant; as Heberlein and Baumgartner
reduced effectiveness. Registered mail
have noted (1978), students, employees,
and long-distance phone calls seem to im-
and military personnel are more inclined
press on respondents the importance of
to return a mailed questionnaire than are
the study and their role in it; using these
members of the general public. The type
approaches pays off well in increased par-
of sponsoring agency also has an impact
ticipation. One of the follow-up contacts
that favors government-sponsored re-
should contain a replacement copy of the
search over market research. (Perhaps
questionnaire in case the rst one has
the response rates in government-spon-
been misplaced. Although follow-up con-
sored projects are higher because some
tacts are worthwhile, one must always be
citizens may believe that they are legally
careful not to harass potential respon-
required to participate in the same way
dents.
that they are required to participate in the
Given the many factors involved, it is
census.) Finally, we need to consider the
difcult to estimate a response rate with
salience of the topic to the respondent;
precision before the survey is under-
subjects that are important to the re-
taken. However, a rst-round response
spondent are more likely to produce a
rate of about 50 percent should be con-
positive response than those of less im-
sidered good; three follow-up contacts
portance to the respondent.
can be expected to increase the response
rate to about 75 percent. Any response
b. Factors Under the Control of the rate above 75 percent should be consid-
Researcher ered excellent. In Canada, one can expect
somewhat lower response rates (Eichner
Although the quality of the questionnaire and Habermehl, 1981; Goyder, 1982).
is not identied as a factor in the Goyder Goyder has suggested that there may
model, it should be noted that his re- well be cultural factors working to lower
search group was examining published response rates to mail questionnaires in
studies that had passed various reviews Canada. His research indicates that in
before publication, so it would be reason- Canadian studies, researchers should an-
able to assume that all were highly pro- ticipate a response rate about 7 percent
fessional. Common sense dictates that lower than is likely in the United States
every effort be made to make the ques- (Goyder, 1982, 1985a).
tionnaire look as professional as possible. It is useful for the researcher to esti-
For the convenience of the respondent, in- mate response rates by using Goyders
clude a stamped return envelope. A cover formula presented in Box 5.4. To illustrate
SURVEY DESIGNS 149
BOX 5.4 Eight-Variable Model for Estimating Response Rate to a

Mailed Questionnaire or Interviews
John Goyder and his associates identified fac- (ERR). This box will provide you with the steps
tors that influence response rates to mailed necessary to do the estimation for the example
questionnaires or interviews. His work is useful in the textbook. A worksheet is provided in
to anyone planning to do interviews or a mailed Box 5.5 to do the calculation for the study you
survey to calculate an estimated response rate propose.
Response Rate Factors (A) Coefficient (B) Scoring (A) (B) (C) Total
Constant 1.146 1 1.146
Post-1970 eld work (constant) 0.059 1.000 0.059
a. Data collection mode (0,1) 0.349 0 0.000
b. Number of contacts (logged) 0.856 0.477 0.408
c. Salience of topic (0,1,2) 0.272 1.000 0.272
d. Special third contact (0,1,2,3) 0.130 1.000 0.130
e. Incentive (0,1,2,3,4) 0.148 0.000 0.000
f. Type of sponsor (0,1) 0.260 0.000 0.000
g. Population type (0,1) 0.099 0.000 0.000
Total 1.897
Steps in Completing the worksheet. f. Type of sponsor, where 0 nongovernmen-

Step 1. For the proposed study, assign a scor- tal; 1 government
ing value for points a through g; using the fol- g. Population type, where 0 special subgroup
lowing guide, enter the score in column B. 1 general population
a. Data collection mode: 0 mailed question- Step 2. For each row, multiply Coefficient (A)
naire; 1 interview by Score Value (B), enter result in Total Col-
b. Log value of number of contacts, where 1 umn (C). Note that the first two rows are al-
0.000; 2 0.301; 3 0.477; 4 0.602 (be sure ready entered; treat these as constants but in-
to include your initial contact as part of the clude them in your calculations.
total number) Step 3. Add the values in column (C).
c. Salience, where 0 not salient; 1 possibly Step 4. To calculate the Estimated Response
salient; 2 salient Rate (ERR), use the following equation (Goy-
d. Special third contacts, where 0 none; 1 der, 1985c: 58): TOTAL 2 arcsine (ERR ),
regular mail; 2 special mail; 3 telephone where TOTAL is the total value calculated by
or personal contact summing the values in the table above. (In the
e. Incentive, where 0 none; 1 less than 25; case of our example, the value is 1.897.)
2 25; 3 50; 4 $1.00 or more

BOX 5.4 Eight-Variable Model for Estimating Response Rate to a

Mailed Questionnaire or Interviews (Continued)
Step 5. Using a scientific calculator do the fol-

lowing calculations:
Notes:*

1. TOTAL 2 arcsine (ERR) Goyder formula

2. 1.897 2 arcsine (ERR) Use value calculated above (1.897)
1.897

3. arcsine (ERR) Divide by 2
2

4. 0.9485 arcsine (ERR) To get rid of arcsine, take sin of value (use calculator
radian mode)

5. sin (0.9485) ERR

6. 0.8125 ERR Square both sides
7. 0.66 ERR 66% expected to respond
Step 6. After the ERR is calculated, decide if Goyder coefficients to decide which proce-
you need to alter procedures to increase the dures would be most feasible in your case.
expected response rate. To do so, examine the
*The use of the arcsine in computing the response rate prevents the predicted response rate from exceeding 100
percent. This can be problematic if steps are not taken to take into account the diminishing returns of additional
contacts with respondents. Heberlein and Baumgartner (1978) regression estimate of response rate, for example,
can lead to estimates of returns of over 100 percent.
SOURCE: Adapted from John C. Goyder (1985b). Face-to-face interviews and mailed questionnaires: The net differ-
ence in response rate. Public Opinion Quarterly, 49, 234252.
the use of Goyders formula, suppose a ceptable. If the researcher thinks that the
study is designed to survey graduates of a estimated response rate will be insuf-
university nursing program to determine cient, then additional steps should be
their job experiences since graduation. taken to increase the likelihood of a re-
The survey is sponsored by the university sponse. An examination of the last col-
school of nursing, the subject is regarded umn of Box 5.4 suggests that the most im-
as possibly salient to the respondents, portant factors in determining response
and two follow-up contacts are planned. rate include:
In this case, we estimate the response rate
by simply plugging in the values pre- The number of contacts the researcher
sented. The predicted response rate is has with the respondents
shown in Box 5.4. The importance of the subject matter
Meeting the predicted response rate to the respondent
should be considered an excellent result. Special third contacts such as phone
To be within 20 percent of the predicted calls or special delivery letters
response rate should be considered ac- The use of a monetary incentive
SURVEY DESIGNS 151
One might wish to consider trying to ble. Avoid the mass produced look; do
make the questionnaire more salient for not use mailing labels.
the respondents, using phone call follow- 4. A stamped envelope for the return of
ups, or include a 25-cent coin as a token of the completed questionnaire should
appreciation. Although many researchers be enclosed with the original material.
would feel uneasy (the authors included) 5. If the questionnaires are to have identi-
about sending money to try to encourage cation codes placed on them, place
a positive response because some poten- them on the top right-hand corner of
tial respondents would be insulted by the the rst page and indicate in the ac-
gesture, it would, nonetheless, probably companying letter that the number is
be effective in increasing the response there to assist in following up on re-
rate. It has been shown, for example, that spondents who have not returned the
the use of incentives as small as ten cents questionnaire. Do not use secret codes.
increase the response rate. For example, 6. If an incentive is being used, use new
in one classic study of top corporate ex- currency enclosed in plastic envelopes.
ecutives, 40 percent of those receiving no 7. You can follow up by sending a post-
incentive responded and 54 percent of card, thanking respondents if they
those receiving 10 cents responded. How- have returned the questionnaire, and
ever, among those who received a 25-cent reminding them that returning the
piece, 63 percent returned their question- form would be much appreciated if
naires (Erdos, 1983). Would you have this has not already been done.
guessed this result if you knew that the 8. A second follow-up, including a copy of
value of the token sent was all that differ- the questionnaire, may be sent 3 weeks
entiated the various surveys and that the after the original has been mailed.
respondents were among the highest- 9. A third follow-up after 6 or 7 weeks, us-
paid executives in North America? ing either registered mail or a phone
call, is worthwhile and increases the
response rate. Most researchers do
c. Tips for Increasing Response Rate not go beyond the third follow-up.
Assuming that the questionnaire looks Generally, returns will be quick at rst
professional and that the appropriate and then slow down. After 1 week, expect
cover letter is included, the following tips to receive about 30 percent of all the
are suggested as methods for increasing questionnaires that will be returned, and
the likelihood of response to a mailed expect to receive about 85 percent within
questionnaire: 2 weeks. By the end of 4 weeks, about 96
percent of the questionnaires that will be
1. The envelope should identify the spon- returned will have arrived (Erdos, 1983;
soring organizations name. By identi- Jackson, 1999).
fying the sponsor, an effort is made to Box 5.3 reports on a study by Esta-
increase the perceived legitimacy of brooks (1998) that used a mailed survey to
the project. collect information from nurses in Alberta,
2. The name should be typed or even Canada, on the use of evidence-based de-
handwritten using the full name, rather cision making. The survey achieved a 40
than initials. percent response rate (600 out of 1500 re-
3. The mailing should be sent by rst class turned the questionnaire), and the results
mail and should use stamps rather than indicated that in this random sample, nurs-
metered postage. The idea is to make ing practice was informed more by experi-
the package seem as personal as possi- ence than by research-based knowledge.
BOX 5.5 Worksheet for Estimating Response Rates for Mailed

Questionnaires and Interviews.
The following worksheet may be used to do the mine the scoring for each of your proposed ap-
calculations to estimate the response rate for proaches, see Box 5.4.
mailed questionnaires or interviews. To deter-
Response Rate Factors (A) Cofficient (B) Scoring (A) (B) (C) Total
Constant 1.146 1.000 1.146
Post-1970 eld work (constant) 0.059 1.000 0.059
a. Data collection mode (0,1) 0.349
b. Number of contacts (logged) 0.856
c. Salience of topic (0,1,2) 0.272
d. Special third contact (0,1,2,3) 0.130
e. Incentive (0,1,2,3,4) 0.148
f. Type of sponsor (0,1) 0.260
g. Population type (0,1) 0.099
TOTAL
The values for the constant and post-1970 field work should be left as is and included in your calculations. See Box
5.4 for details on calculating the estimated response rate (ERR).
Furthermore, the study indicated that recorded into the computer. Today most
there are problems in how scientic evi- phone interviews use computers to pres-
dence is communicated to practitioners. ent the questions to the interviewer on the
screen as well as the possible response
categories into which respondents an-
5. Phone Surveys swers are to be tted. Care must be taken,
however, to monitor the reliability of in-
Like all surveys, phone surveys rely on in-
terviewers in placing respondents an-
formation reported by respondents and
swers in the categories provided.
are therefore vulnerable to image manage-
ment. Questions and response categories
must be kept simple because they are pre- a. Advantages and Disadvantages of
sented verbally. In-depth probes are dif- Phone Surveys
cult and, as with other surveys, it is always
difcult to make causal inferences. The in- Phone surveys are gaining in popularity.
terviewers expectations may inadver- They are widely used by polling organiza-
tently inuence the responses that are tions and academic and applied research-
SURVEY DESIGNS 153
ers, and they represent a technique of data Phone interviews also have some spe-
collection that will almost certainly in- cial problems related to assessing re-
crease in years to come. Phone surveys are sponse rates. It is not always easy to de-
a relatively cheap and quick way to collect termine how many numbers called are
data. Because no travel time is included, connected to live phones; there are a
phone interviewers can do many more in- fair number (generally about 20 percent)
terviews in a day than would be possible if of phones attached to businesses. Fur-
the interviewer had to travel to each re- thermore, phone interviews have lower
spondents home. Moreover, phone inter- completion rates than questionnaires;
viewing can provide cost-effective access typically, the completion rate does not
to people with whom it is very difcult to exceed 70 percent (Jackson, 1999). Stud-
arrange interviews (such as physicians) or ies based on rural populations generally
with those that are not concentrated in have greater success in phone surveys
one area (such as the blind). because lower levels of mobility mean
In national studies, it is estimated that that fewer phones are disconnected.
phone surveys cost about 45 percent as
much as personal interviews (Jackson, b. Creating a Sample
1999). Recently, there has been some
troubling news regarding the willingness In conducting a phone survey, one may
of people to participate in telephone in- work with a list of potential respondents
terviews. In 1993, some 7 out of 10 re- (e.g., a list or from the names listed in the
spondents in the Toronto area refused to phone book). In such cases, one usually
cooperate in pre-election polls. If these proceeds by using a systematic sampling
levels of nonparticipation become wide- procedure (see Chapter 15 for this sam-
spread, it will be more difcult to justify pling method).
using the phone for collecting such infor- It is also possible to create a sample by
mation (Sheppard, 1993; Fisher, 1993). identifying the various residential phone
But there are disadvantages to phone exchanges in the area and then using a
surveys. First, they are not the best table of random numbers to determine
means of gathering data if probing is re- the numbers to be called. Typically phone
quired or if complex response categories numbers are assigned in ve-digit blocks,
are to be presented. Second, the distribu- the rst three determining the exchange.
tion of phones is uneven. The less well off The numbers might start 86321xx. A
and the mobile are less likely to have a table of random numbers may be used to
phone or a phone number listed. How- determine the last two digits to be used. If
ever, as phones become more universally it is possible to get the information from
available, there is less need to avoid the phone company, one attempts to nd
phone surveys on principle as long as re- out the percentage of phones in each
searchers recognize the possibility of block and then a sample is drawn to rep-
sample distortions. Indeed, it is possible resent each block proportionally (Abra-
to weight samples to adjust for underrep- hamson, 1983; Jackson, 1999).
resented categories in a survey. Respon- Computer-assisted telephone inter-
dents interviewed over the phone are viewing is an important tool for polling
slightly less at ease than respondents be- organizations and market researchers. A
ing interviewed in person. As a result, computer dials a sample of respondents
phone interviews generally produce and then guides the interviewer through
slightly higher refusal rates on sensitive the data collection by presenting the
issues, such as income or political pref- questions on the screen and, depending
erence (Jackson, 1999). on the response, it proceeds to show the
next appropriate question. The answers sponse categories simple. Although it

are then recorded into the machine. is possible to conduct lengthy inter-
views by telephone, keep the ques-
tions simple, and if necessary, break
c. Rules for Phone Surveys complex questions into smaller, sim-
pler ones (Sudman and Bradburn,
In addition to the general rules identied
1983; Jackson, 1999). Respondents are
earlier, a few additional rules should be
slightly more likely to select a neutral
considered in designing a phone survey.
response category and there is also a
Rule 1. Begin with interesting, salient, tendency to choose the last response
and yet simple questions. This is a category presented; therefore, re-
tough rule to follow, but attempt to be- searchers frequently vary the order of
gin a phone interview with questions the response categories.
that are simple to answer, nonthreat-
ening, and that will be considered im- Phone interviews are relatively inexpen-
portant. Because phone respondents sive and can be used to generate a repre-
are a little more likely than those in- sentative sample. However, the represen-
terviewed in person to terminate the tativeness of the sample has to be
interview, it is especially important to monitored carefully because phones are
ease them into the discussion. Al- not evenly distributed among the popula-
though phone interviews can last up to tion and, as mentioned earlier, there is a
half an hour, they should be kept as growing disinclination to agree to be in-
short and as simple as possible. terviewed over the phone.
Rule 2. Supply interviewers with rules for
determining who is to be interviewed.
Procedures must be provided for deter- 6. Interviews
mining who within the household con-
tacted is to be interviewed. It is not ac- The nal data collection procedure dis-
ceptable to interview the person who cussed here is the interview. Generally,
happens to answer the phone. If this two kinds of interviews exist: (1) the struc-
were done, the survey would overrep- tured interview, which is a questionnaire
resent those who are most likely to an- that is read to the respondent; and (2) the
swer the phone. The interviewer should interview schedule, which outlines the ar-
not be making convenient decisions. eas to be probed by the interviewer.
Table 5.2 presents such a sample form.
Rule 3. Monitor quality. A quality con-
trol monitoring system must be in a. Structured Interview
place to ensure that interviewers are
following established procedures for Structured interviews involve face-to-
selecting respondents, asking ques- face interviews in which questions are
tions, and entering the data. In partic- read to respondents. Such interviews or-
ular, some interviewers nd it difcult dinarily provide for in-depth probes on
to resist the temptation to interview some of the questions. Interviews also al-
the person who answers the phone or low the respondents to ask questions to
someone who is immediately avail- help clarify any ambiguities. Basically,
able. Callbacks are the bane of the in- structured interviews are made up of set
terviewers life. questionnaire items, and the interviewer
Rule 4. Simplify response categories. is expected to read the questions, exactly
Phone interviews must keep the reas worded.
Table 5.2 Choosing Phone Respondent
# Phone # List People in Household1 Skip2 Interview Names3 Name/Time4 Done5 Name/Time Done Name/Time Done
1 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
2 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
3 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
4 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
5 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
6 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
7 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
8 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
9 ________ ________________________ _____ _________________ ____________ _______ ____________ ______ ___________ ______
1Get initials of members of household older than age 19 years who normally reside in the household, from oldest to youngest.
2Start with person X from this household; if skip interval is three, do one, skip two, and do one. Carry over skip from interview to interview.
3Write down name of persons to be interviewed.
4Write down name of rst person and time interview to be done.
SURVEY DESIGNS
5Check when interview completed.
155
b. Interview Schedule time spent working for NORC, high grade

point averages in high school, liking two or
Interview schedules outline the major more science subjects, intelligence, and
questions that are to be raised. The inter- the completion of college. In addition,
viewer has greater autonomy in exploring those who scored high on need achieve-
questions in detail. Interviews require ment and manipulativeness (Machiavel-
much skill on the part of the interviewer lianism scale) are more likely to do well
and care must be taken not to lead the when interviewing. Of note is that happi-
respondent. Furthermore, the responses ness, nancial need, religious behavior,
are ltered through the interviewer; there- perfectionism, and size of home commu-
fore, if there are a number of interviewers, nity were not found to be related to the
one must realize that some of the varia- quality of interviewing that a person does
tions in response will be caused by differ- (Sudman, 1967; Jackson, 1999).
ences between interviewers and not solely Interviewers need to be trained. They
by differences between the respondents. need to gain knowledge about ethical is-
Interview schedules are used for in-depth sues, the survey being conducted, appro-
interviews in eld studies (see Chapter 7). priate dress, how to introduce them-
selves to the respondent, gaining rapport,
organizing the interview setting, how to
c. Advantages and Disadvantages of present questions, how to react to re-
Interviews sponses, which issues to probe and how
to probe them, how to keep the respon-
Because interviews are expensive, they
dent on topic, and how to gracefully end
are normally done when not too many are
the interview. In addition to some of these
required and when in-depth information
issues, research directors need to pro-
is needed. One major advantage is that
vide potential interviewers with experi-
good rapport is often built up between in-
ence in a few simulated interviews.
terviewers and respondents so that if re-
Interviewers are frequently paid on a
peated interviews are required, as in a
per interview basis. This method is often
panel study, it will be possible to main-
preferred because it allows researchers to
tain high response rates. A second major
control costs. Also, it seems that many in-
advantage of interviews is that they per-
terviewers burn out after 6 or 8 weeks.
mit the respondents to clarify any ques-
Interviewing is an especially challenging
tions that they have about the interview.
task, requiring great concentration, and it
One of the disadvantages is that inter-
is not easy to remain alert after having
view studies are expensive (more than
walked many respondents through the
double the cost of phone interviews) and
interview process. Expect high turnover
time consuming (Jackson, 1999).
among interviewing staff.
d. Selection and Training of

Interviewers
E. COMPARATIVE STUDIES
Although it is beyond the scope of this
book to discuss the selection and training Whether we are doing an experiment com-
of interviewers, we will make some brief paring time 1 measures on some variable
comments on this subject. Research done with time 2 measures, or whether we are
by NORC (National Opinion Research Cen- trying to gure out why inner city crime
ter) indicates that the quality of work done rates are higher than those of suburban
by interviewers is related to the length of rates, we make comparisons. Thus, com-
SURVEY DESIGNS 157
parison is integral to the research process ing has become increasingly concerned
itself. Comparative surveys involve com- with providing culturally competent care
paring two or more samples on one or in all sectors, such studies have taken on
more variables, at a single point in time, greater importance. And although these
and they use standard survey techniques studies do not have a distinct methodol-
of analysis; therefore, they are included in ogy associated with themthey use his-
this chapter. But there is an important dis- torical material, surveys, secondary data,
tinction between experiments, surveys, experiments, and eld studieswhat
and other designs that allow us to draw they have in common is that they attempt
comparisons and comparative research to make comparisons between cultures
studies whose very purpose is to compare. or within cultures over time.
It is this latter group of studies that we ex- Because nurses are frequently curious
amine here. about the efcacy of different approaches
Comparative research may be quantita- and treatments to client problems and
tive, qualitative, or a combination of both. conditions, comparative studies enable
Comparative researchers use the full them to test these differences as well as
range of standard techniques, including compare and contrast the impact of vari-
surveys, interviews, eld studies, and ex- ous interventions on different client
periments, but, in particular, they use pub- groups. For example, a comparative study
lished information. Cross-cultural studies may be used to answer the following
and historical studies are two of the most types of research questions:
common examples of comparative stud-
ies; these look at the similarities and dif- What caring behaviors are identied
ferences between cultures or within the by oncology nurses compared with
same culture over time. oncology patients on an active treat-
When the modern social sciences were ment unit?
emerging in the 19th century, comparative Do clients who receive care for hyper-
studies were a central concern of social re- tension in a nurse-run clinic do better
search. During the early years of the 20th in reducing their blood pressure levels
century, after arguments against a unilin- than clients who receive care through
eal model (that proposes that the same their doctors ofce?
patterns of development are followed by What is the experience of school-aged
all societies), comparative studies shifted children admitted to a hospital day
to intensive studies of exotic cultures by care unit compared with those admit-
anthropologists such as Bronislaw Mali- ted to a general pediatric unit for treat-
nowski (1925) and Margaret Mead (1935). ment of asthma?
In the closing decades of the 20th century, Do psychosocial differences exist in
many comparative studies have been un- adaptation to pregnancy for Latin-
dertaken. Later in this chapter you will American, African-American, and Anglo-
read about a study comparing the experi- American ethnic groups?
ence of girls coming of age in Jerusalem What differences exist in the career in-
with those in Toronto (Box 5.6). terests of nurses educated in a bac-
There is, then, a long comparative tra- calaureate program, an associate de-
dition in social science research. Because gree, and a diploma program?
many early nurse scientists received What is the pattern of folk healing be-
their research training in schools of so- liefs and practices among Korean and
cial science, it is not surprising that a Chinese refugees who move to the
number conducted studies using cross- United States?
cultural comparative analyses. As nurs- What differences exist in the health
COMING OF AGE IN THE METROPOLIS

Findings manding, relenting domestic chores and care-
The results indicate similarities and differences giving tasks, differentiating themselves clearly
between girls from the two cities. The most from their brothers, who were free of regu-
striking similarities included: lar responsibilities.
Ideals and aspirations. The girls in Toronto
Requests from girls in both cities for contin- were preoccupied with wealth and Barbie
uing, systematic opportunities for guided dis- dolls. This may suggest some deprivation in
cussion on their personal development and their lives, compensated by fantasizing about
well-being the lives of TV soap and drama characters.
A shared perspective on the social function The girls in Jerusalem cited a fashion model
of schools as their ideal and also stressed intelligence
The need to be trusted, to have greater free- and inner strength.
dom of action, and not to be overprotected Fathers. The girls from Jerusalem described
or controlled some egalitarian tasks and roles of fathers;
The perspective, whether realistic or not, of the girls from Toronto mentioned fathers
professional aspirations mainly as authoritarian figures.
The need and desire to be free of teasing and Discussion style. Responses from the girls in
harassment by boys Toronto were short, sometimes almost
The embarrassment about and rejection of monosyllabic; group dialogue and interaction
the role of grandparents in their lives were minimal. The group from Jerusalem
freely engaged in debate and discussion. It is
The differences were equally striking: difficult to know if this difference is cultural
or situational.
Primary concerns. Despite their existence in
Conclusions
an unstable part of the world, girls in
Jerusalem expressed concerns mainly about A better understanding of this critical period of
their positive biopsychosocial development. life is needed to support these young girls
More than anything else, the girls from inner through this developmental phase toward their
city Toronto articulated their insecurities future goals as women. Policy development that
and risks concerning rape, AIDS, pregnancy, is relevant to age and gender is needed to
and sexual pressures. strengthen and enhance health and lifestyle de-
Tasks and chores. The girls in Jerusalem could cisions for the girls themselves but also because
not discuss this even when prompted. This women everywhere are the major health-care
suggests that they were negligible. The girls in providers and users. This would ultimately con-
Toronto, without exception, reported de- tribute to the health status of countries.
SOURCE: Summarized from Paltiel, F., Ross, E., and Neill, M. (1998). Coming of age in the metropolis: The Toronto
experience. The Canadian Nurse, 94(10), 2230.
promotion role for nurses working in ing contrasted. For example, if a nurse re-
urban versus rural hospital settings? searcher is interested in conducting a
comparative survey of health promotion
When conducting a comparative study, practices of nurses working in large urban
researchers must pay special attention to hospitals and nurses working in small
the sampling procedure used. Usually the rural hospitals, two samples would be ob-
sample for a comparative study is drawn tained, one from a population of nurses
from two or more populations that are be- working in urban hospitals and one from
SURVEY DESIGNS 159
a population of nurses working in small, conducted to identify cross-cultural uni-

rural hospitals. This is in contrast to other versal and culturally unique needs of
surveys in which the researcher is inter- young adolescent girls. Data were col-
ested in studying how variables are re- lected by means of nurse-facilitated focus
lated (correlational surveys) rather than groups that lasted for 2.5 hours. The rst
how populations differ on a variable. For focus group was conducted in Jerusalem
example, if the researcher were simply in- with 19 girls from the ages of 10 to 14
terested in examining the factors or vari- years. These girls shared aspects of the
ables related to health promotion prac- biological, social, and psychological as-
tices by nurses, the study sample could pects of their lives. The second focus
be drawn from a single population of group was conducted in Toronto with 24
nurses. In a comparative survey, the em- girls ages 12 to 14 years (see Box 5.6).
phasis is on how two or more populations A discussion guide was developed for
differ with respect to the variables of in- use in the focus groups by the principal
terest. It usually involves measurement of investigator. The guide covered a wide
one or more variables in two or more range of topics, including education,
groups at one or more points in time health, friendship, sexuality, ideal women,
(Woods and Cantazaro, 1988). spirituality, recreation, and family. The fo-
Let us consider some examples of cus groups were intended to provide a fo-
comparative studies in nursing. rum in which the experiences, feelings,
perceptions, values, and judgments of the
adolescent girls could be expressed. In
1. Examples of Comparative each focus group, the nurse facilitator
Nursing Studies could adapt the guide for ease of use with
the local group. The opening question in
Two nursing studies are presented to il-
each group was a general one, What are
lustrate the comparative approach. The
the major concerns of girls your age? A
study by Paltiel and associates (1998)
round robin technique was used to pro-
compares perimenarchal girls from two
mote discussion. Girls could pass if they
different cultures. This qualitative study
did not wish to speak. In Toronto, the fo-
was reactive in nature, using focus
cus group took place in a large inner city
groups to elicit data for comparison pur-
school. In Jerusalem, the focus groups
poses. We also include an exploratory
were held in a health center. In addition to
study using open-ended interviews to col-
the adolescent girls, the principal investi-
lect data on prenatal adaptation to preg-
gator, the facilitator, and two observer-
nancy of women from three different cul-
recorders were present in the focus
tural groups.
groups.
The Toronto sample was selected from
a. Coming of Age in the Metropolis: An one school and reected the multicultur-
International Study on the Needs alism of Toronto and, especially, the
and Concerns of Girls 10 to 14 neighborhood in which the school was lo-
cated. The group was multiethnic with
Palteil and associates (1998) investigated origins from Burma, the Philippines, the
the differences and similarities among Caribbean, Vietnam, China, Iran, India,
two urban groups of early adolescent and African countries. Two of the 24 girls
girls residing in different parts of the were of European descent. The Jerusalem
world in terms of their identied needs sample came from four different schools
and major concerns. A collaborative, in four different neighborhoods. The so-
qualitative, action-oriented study was cioeconomic range of the Jerusalem sam-
ple was wider because of the range of ment of a psychosocial adaptation ques-
neighborhoods used for selection. tionnaire for different cultural popula-
tions; therefore, response categories were
prepared for all questions after the inter-
b. Adaptation to Pregnancy in Three views were completed by creating mutu-
Different Ethnic Groups: Latin- ally exclusive categories based on subject
American, African-American, and responses. Results and conclusions of this
Anglo-American study are provided in Box 5.7. The results
have implications for providing culturally
This comparative study used a qualitative sensitive and relevant health care to preg-
approach to investigate adaptation to nant women in the three ethnic groups
pregnancy in three ethnic groupsLatin- studied. By identifying the reproductive
American, African-American, and Anglo- concerns of minority ethnic groups and
American women (Lederman and Miller, providing culturally competent care to re-
1998). The research question addressed spond to these needs, the anxiety and
was: Do psychosocial differences exist in stress experienced by pregnant women
adaptation to pregnancy for Latin-American, can be reduced. This study suggests that
African-American, and Anglo-American eth- culturally informed data be used for re-
nic groups? The purpose of this study was search development, policy decisions,
to produce knowledge that would improve and program implementation.
the prenatal health care and the counsel-
ing of women from different cultural 2. Challenges in Comparative
groups. The goal was to create a survey Research
that would target differences in cultural
groups so that the success of future inter- Comparative research, particularly stud-
ventions could be enhanced (Box 5.7). ies that focus on cultural differences, has
Thirty Latin-American, 34 African- potential to contribute signicantly to
American, and 30 Anglo-American women nursings body of knowledge. Culture has
were recruited from a low-risk prenatal an important impact on ones interpreta-
clinic at a university medical center in the tion of health and illness and ones re-
southwestern United States. Participants sponse to health care. Nurses and all
were interviewed face-to-face in the latter health-care professionals must make a
half of pregnancy by trained interviewers commitment to expand their knowledge
using open-ended questions. The inter- base and competence in cultural con-
views ranged from 45 to 90 minutes in du- cepts and care. A critical area for com-
ration. The interview schedule focused on parative research in nursing is the impact
ve dimensions: acceptance of pregnancy, of cultural diversity on health outcomes
identication with motherhood role, rela- in various populations. Nursing actions
tionship with mother, relationship with must be based on systematic investiga-
partner or husband, and preparation for tions of cultural groups and not on cul-
labor. The interview schedules were trans- tural myths. Cherry and Jacob (1999) call
lated into Spanish using a forward-and- for research that examines normal life
backward method of translation (from processes of growth and development,
English to Spanish and back to English). birth, death, pregnancy, as well as well-
The interviewers of the Latin-American designed studies that examine the bio-
participants were uent in Spanish; most logical, psychological, sociological, and
interviews, however, were conducted in spiritual differences within, between, and
English. among cultural groups. Comparative de-
A goal of this study was the develop- signs can yield such information.
SURVEY DESIGNS 161
ADAPTATION TO PREGNANCY IN THREE DIFFERENT

ETHNIC GROUPS
Results pected to receive help from their extended fam-

Women in all three ethnic groups were pleased ily and other family members, whereas Anglo-
with their pregnancies, despite the unplanned and Latin-American women depended more
nature of the event and the associated changes upon their partner. Latin-American women re-
in life plans that resulted. Less than 30 percent ported that they wanted to be like their own
of the pregnancies were anticipated. Many mother although they thought less about the
women commented that their feelings toward kind of mother they wanted to be.than the other
the pregnancy changed as they progressed groups.
through the various stages moving from disap- In reflecting upon childhood patterns of seek-
pointment to anticipated pleasure. It appears ing help it was noted that the Latin-American
the women had come to accept their pregnancy women sought assistance most often from their
by the time the interviews were conducted in mothers, Anglo-Americans from their fathers,
the third trimester. and African-Americans turned to a range of fam-
Childcare was an issue for all three groups. ily members.
Family, friends, and day-care centers were Patterns of sexual activity underwent changes
sources used to provide assistance. Contrary to for all three groups. An increase occurred for a
reports in the literature about support for child- small number of African- and Anglo-American
care in Latin-American families, the women in women but for the majority in all groups there
this study were not able to depend upon family was a decrease reported. Anglo-Americans re-
care to the same extent. A possible explanation ported the greatest satisfaction with the fre-
for this finding may be that the women were first quency of intercourse, while the other two
generation Americans and hence did not have groups of women would have preferred a fur-
easy access to their families who resided else- ther decrease. Different forms of pleasure were
where. explored most frequently by Anglo-Americans.
Differences were noted among the groups in
terms of aspects of pregnancy that brought plea- Conclusions
sure. The Latin-American women were happier This study is important because it addresses an
than women in the other two groups with their area understudied in nursing research. It high-
pregnancies and stated more frequently that their lights the unique concerns and needs of three
families were happy. Whereas, Anglo-American ethnic groups of pregnant women. This infor-
women made comments that reflected a valuing mation can be used to provide culturally sensi-
of their children; African-American women did tive and responsive counselling and health
not reflect this sentiment in their comments. teaching in the area of family nursing and pre-
The ability to stay at home and provide child natal care. It points out that future research
care was explored, as well as anticipated life should address ethnic and cultural differences
changes that would result from the birth of a relevant to prenatal care should address repre-
child. Plans to stay at home were expressed sentative samples of populations at risk and
strongly by Anglo-Americans and less so by prominent psychosocial and health problems,
African Americans. Anglo- and Latin-American and should use related assessment and inter-
women anticipated more life changes than vention approaches to produce culturally in-
African-American women as a result of giving formed data for research development, policy
birth. African-American women more often ex- decisions, and program implementation.
SOURCE: Summarized from Lederman, R. and Miller, D. (1998). Adaptation to Preganancy in Three Different Ethnic
Groups: Latin-American, African-American, and Anglo-American. Canadian Journal of Nursing Research, 30(3), 3751.
Although it has much to offer nursing, compare health status: given different
comparative research also presents some denitions, recording methods, and, in-
interesting challenges because it deals deed, the different interests of the parties
with different cultures and often a differ- involved in the data collection (would
ent language. A number of authors iden- they benet by showing an increase in
tify issues concerning equivalence in con- morbidity?), can we legitimately make
cepts, indicators, and language (Warwick comparisons across jurisdictions? For
and Osherson, 1973; Jackson, 1999). most measures, there are problems with
a lack of equivalence indicators as we
move from one jurisdiction to another.
a. Equivalence of Concepts How might we go about limiting the ef-
fects of such disparities? One suggestion
Although ideas such as incest and health
is that we use data trends rather than ab-
are cultural universals, there is no pre-
solute measures (e.g., the rate of crime,
cise agreement about what is meant by
the school drop-out rate). For example, if
these concepts. What is incest in one cul-
you are comparing the United States and
ture may be dened as a preferred mar-
Canada in school drop-out statistics, you
riage partner in another; what is consid-
could minimize the effects of alternate
ered normal, healthy behavior in one
measures by simply using the change in
culture may be labeled as bizarre in an-
school drop-out rate over periods of time.
other. So although we may have similar
Then, even if denitions vary between the
concepts, their content and meaning may
countries, at least they are comparable
vary considerably from culture to cul-
across time spans within the country (un-
ture. Moreover, even within a culture
less, of course, denitions or procedures
there may be subtle variations in how
have changed within the country). Just as
concepts are dened.
we standardize data within a country by
In all research, but particularly in com-
calculating rates and ratios to deal with
parative research, it is important to dene
units of unequal size, when between-
concepts carefully, noting any variations
country data are required, consider com-
between the cultures being studied. The
puting trend data. And although such
search for appropriate indicators is facili-
trend data would not provide the ab-
tated by attention to the denition of the
solute measures desired, at least one
concepts. Only if a researcher is satised
should be able to detect whether the
that there is an equivalence of concepts
rates are converging or diverging.
can direct comparisons between cultures
be drawn.
c. Equivalence of Language
b. Equivalence of Indicators
In the study Adaptation to Pregnancy in
The evidence collected in different coun- Three Different Ethnic Groups, the inves-
tries is rarely based on identical deni- tigators attempted to provide for equiva-
tions and collection procedures. For ex- lence of language by having the interview
ample, if one country denes a school guides translated into Spanish using a
drop out as anyone under age 14 years forward-and-backward method of trans-
not attending school but another denes lation. In addition, the interviewers of the
a drop out as anyone under age 16 years Latin-American participants were uent
not attending school, do we dare use the in Spanish. This greatly facilitated the
information to compare the two soci- participants ability to communicate ap-
eties? Similar problems arise if we wish to propriately with the interviewer. Such
SURVEY DESIGNS 163
precautions, however, are no guarantee d. The Problem of Selecting Evidence

of successful equivalency of language, as
the following anecdote suggests. Although the problem of researchers
Box 5.8 describes a problem that can choosing among alternate indicators is rel-
arise when trying to provide for language evant to all types of research projects, this
equivalence. Several years ago, Clairmont is especially the case in comparative stud-
and Jackson were doing a study in Monc- ies. Given the vast amount of information
ton, New Brunswick, Canada. They were available, what is the best way, for exam-
comparing people who worked for large ple, to measure health status in a country?
companies and government agencies with (Some may use morbidity statistics as
a matched sample of workers in smaller number of hospital admissions, days spent
companies. Because Moncton has a bilin- in the hospital, or rates of chronic illness
gual community, the researchers hired a as indicators; others may select com-
translator to develop a French version of munity health outcomes such as infant
the interview. A number of bilingual inter- or child mortality rates or immuniza-
viewers were employed to conduct the in- tion rates; others may use health-related
terviews with some 600 residents in Monc- quality of life indicators such as quality-
ton. Box 5.8 reports on the experience. adjusted life years [QALYs]; and still oth-
ers may use social outcomes that relate to
health such as school absence rates and
status indicators such as employment, in-
come, and living arrangements). The prob-
lem is that it is possible to demonstrate al-
most anything if you are free to search
around for possible indicators. The gen-
eral point here is that the selection of evi-
dence is an especially problematic issue in
comparative research. Not only can we
question whether the indicators reect
what they are intended to, but we can also
question whether the same variable will
Image/Text rights unavailable have the same meaning in different cul-
tural contexts. This was demonstrated in
the cross-cultural study of adolescent girls
in Toronto and Jerusalem concerning fam-
ily chores. Despite the best prompting ef-
forts by the group facilitator, the girls in
Jerusalem would not discuss this, leading
one to conclude that they were either neg-
ligible or the participants did not under-
stand the meaning of the question. The op-
posite situation existed for the girls in
Toronto, who spoke without exception
about the demanding family chores and
domestic responsibilities they were ex-
pected to assume.
It would be difcult for most nurse re-
searchers to avoid selecting indicators
that produce results conforming to the re-
searchers preferred outcomes, ruling out of committing oneself to operational pro-

the indicators not selected as insufcient cedures and operational hypotheses.
in some way. There is no easy solution to
this problem. One check is simply that re-
searchers be forthright about which indi- F. SECONDARY DATA
cators were used and perhaps even public ANALYSIS
about which indicators were rejected and
why they were rejected. It would not be A range of researchers, including nurses,
practical to ask researchers to explain all economists, political scientists, histori-
of their decisions or to list all the alterna- ans, and, to a lesser degree, sociologists,
tives considered. To do so would be akin anthropologists, and psychologists, fre-
to asking a chess player to explain all the quently conduct research based on avail-
alternatives considered before making a able material. These sources may include
move. The chess players information pro- virtually any data, including published
cessing is highly complex and the player, statistical data, national census track sur-
in fact, probably could not provide much veys, medical records, health data from
useful information about how the choice personal health records, business data,
was made. The same would be true of the and unpublished diaries of important his-
researcher. But if we are up front about torical gures. With the exception of
the choices that have been made, then those who record oral histories, histori-
critics can reexamine the issue using the ans rely on secondary data sources and
same or alternate indicators. the resulting studies are based entirely on
Researchers have not always paid suf- secondary data analysis. Secondary data
cient attention to identifying precise indi- analysis is the analysis of an existing data
cators before beginning data analysis. set, or documents, for some research pur-
Such precision is required because other- pose other than the one originally in-
wise the researcher may inadvertently tended. It involves re-analysis of data col-
bias the outcome of the study. Moreover, lected by another investigator to answer
the exact cut-points that are to be used the same research question or a different
in collapsing categories should also be question or to apply a different method of
identied before beginning analysis (e.g., analysis. Secondary data analysis is in-
where will the line be drawn between large cluded in this chapter because it most re-
and small communities when looking at sembles survey designs in terms of the an-
the relationship between prevalence of alytical procedures used.
sexually transmitted diseases and com- Meta-analysis is the statistical analysis
munity size?). Again, if commitments are of a large collection of results from indi-
not made, the researcher may inadver- vidual studies for the purpose of integrat-
tently select cut-points that result in an ing ndings (Onyskiw, 1996). The subject
analysis showing what the researcher an- and the unit of analysis in a meta-analytic
ticipated or wanted. study is the individual research report.
Realistically, these two suggestions are Meta-analysis is often referred to as analy-
unlikely to be applied routinely by scien- sis of analysis because it depends on the
tists. It is sometimes a challenge to get re- ndings of primary research for its data.
searchers to state hypotheses formally, This approach provides a mechanism for
let alone provide details of exact opera- nurses to make sense out of the growing
tional procedures. At present, there is lit- body of nursing research so it can be used
tle awareness that there is a problem in to guide practice. For example, Gillis
this area. Issues raised in Chapter 12 and (1993) conducted an integrated review and
in Chapter 9 are also related to the issue meta-analysis of the research literature on
SURVEY DESIGNS 165
determinants of a health-promoting life- inal study. For example, Humenick and

style. The meta-analysis allowed her to associates (1998) used secondary data
summarize mathematically the results of analysis to partially replicate and ex-
the studies to determine the effect of a tend Mortons (1994) research to ex-
range of independent variables on the de- amine additional risk factors with me-
pendent variable of health-promoting be- diating potential on the relationship of
havior. The ranges and pooled correlation elevated breast milk sodium (BM
values were determined for the indepen- [Na]) to breastfeeding outcomes.
dent variables to provide an index of effect The primary research question was,
magnitude. For a discussion of meta-analy- Do elevated levels of BM [Na] at
sis using the correlation statistic (r), inter- postpartum day 6 by themselves serve
ested readers are referred to Rosenthal as a marker to predict low frequency of
(1984). The review summarized the nd- breastfeeding at week 4? In addition,
ings of 23 separate research studies on de- Humenick et al. asked a new question:
terminants of health-promoting behavior Do the psychosocial risk factors of (1)
and showed interesting patterns across perceived insufcient milk supply
the studies. (IMS); (2) the number of best friends
Box 5.9 lists some recent projects in who breast fed their babies in the past
secondary data analysis completed by 3 years; and (3) the mothers own per-
our students in a one-term course. ception of her breastfeeding skills me-
diate the relationship between ele-
vated BM [Na] and the breastfeeding
1. Purpose outcome?
There are several purposes for conduct-
To use a unit of analysis that is differ-
ent from that used in the original data
ing secondary data analysis (Jackson,
analysis. For example, in the study on
1999; Woods and Catanzaro, 1989; Hinds
health-promoting lifestyles in child-
et al., 1997). These include:
bearing families, a researcher may
The identication and focus on a sub- elect to focus on the mothers lifestyles
sample or subset of cases not consid- and subsequently see the mother
ered in the original study. For example, rather than the family as the unit of
in a study of health-promoting lifestyles analysis.
in childbearing families, you may wish To use different analytical strategies
to do a secondary analysis of the life- from those used in the original study.
style proles of only the adolescent For example, in a study titled Examin-
girls in those families sampled in the ing Emotional, Physical, Social, and
original study. Spiritual Health as Determinants of Self-
The analysis of variables or concepts rated Health Status, the researchers
not analyzed in the original study. For Ratner and associates (1998) used dif-
example, if the original study focused ferent statistical procedures to con-
on the impact of social support as a duct a secondary analysis of data from
critical determinant of success in the Yukon Health Promotion Survey.
smoking cessation for single mothers, The purpose of the secondary analysis
you may wish to analyze the impact of was to determine if individuals per-
a related variable, the peer network ceptions of their emotional, physical,
that appeared to be discussed in the social, and spiritual health constitute
original study but was not analyzed. elements of their self-rated health sta-
The investigation of relationships or tus operationalized with a commonly
hypotheses not considered in the orig- used single indicator (the EGFP, or ex-
BOX 5.9 Student Researchers at Work
SECONDARY DATA ANALYSIS PROJECTS

Project Student or Students Brief Summary
Conformity to Thinness in Kelly Yorke (1995) This project was a content

Women analysis of television ads
in terms of the degree of
thinness by type of
commercial.
Portrayal of Women in Robyn MacConnell (1995) This project examined the

Seventeen magazine female models used in the
1960s, 1970s, 1980s, and
1990s in terms of body
type.
Pay, Performance, and Sean Barret (1995) Using published informa-

Productivity tion, the project related
team pay levels to win
and loss records.
Change in the Portrayal of Kathleen Lumsden (1995) Using Macleans ads

Women, 19501995 published over 4 decades,
the study examined
changes in the portrayal
of women.
Personal Ads Shelley Conners (1996) Using the Halifax Chronicle

Heralds personal ads, the
study did a content analy-
sis (what people were
looking for) in hetero-
sexual ads, 19941995.
NHL Penalty Minutes Scott Gardner (1996) Using published hockey

results material, the
project examined the
relationship between
penalty minutes and the
success of teams.
Romance in Alcohol, Car, Deborah E. Farr (1996) Continue analysis of the use
and Cologne Ads of romance to promote
cars and cologne.
Content Analysis of Logann McNamara (1997) This project examined

Advertisements in Vogue changes in the extent to
Magazine, 1982 to 1991 which women in adver-
tisements were portrayed
in sexually permissive or
dominant positions.
Faceism and Gender Roles Wendy Chisholm, Rita Gillis, This project examined
Michelle Waye, and changes in the extent to
Flora Murphy (1997) which women in adver-
tisements are portrayed
with full body shots but
men are more likely to use
face shots: Time maga-
zine in the 1950s, 1970s,
and 1990s.
SURVEY DESIGNS 167
cellent, good, fair, poor indicator). The vey; Statistics Canada National Census
sample was made up of 742 women and Data; the National Population Health Sur-
713 men. The key nding of the analy- vey; and The National Institute of Nutri-
sis was that the EGFP indicator, as a tions Tracking Nutrition Trends Survey. In
measure of self-rated health status, the United States, comparable surveys ex-
measures only physical health status ist such as the National Health Survey; the
and does not appear to tap the emo- Annual Survey of Hospitals conducted by
tional, social, and spiritual dimensions the American Hospital Association; the
of health. National Survey of Personal Health Prac-
tices and Consequences; and the Health
2. Sources of Data Records Survey, which contains informa-
tion from medical and residential institu-
Secondary analysis involves using data in tions providing health care, to name a few.
a variety of forms and from a variety of
sources. Both quantitative and qualitative
data can be subjected to secondary analy- 3. Challenges in Conducting
sis. It may be raw data in the form of tapes, Secondary Data Analysis
interview transcripts, videos, completed
questionnaires or interview schedules, a a. Using Clinical Databases
database, or archival material. Computer-
Most large health-care institutions have
ized databases are becoming popular
some type of computerized data collec-
sources for secondary analysis of data in
tion system. These databases are rich re-
health-care settings. They are excellent
sources for answering nursing questions,
data sources for nurses wishing to explore
but some issues need to be dealt with be-
clinical questions, particularly those that
fore using them for research purposes.
require information over time, without
With careful planning, many of the limita-
gathering new data (Querker, 1997).
tions of databases can be overcome, mak-
A vast range of health information and
ing them viable and economical sources
data sources is available to nurse re-
of data as research funding continues to
searchers for secondary data analysis. Ex-
shrink. Querker (1997) identies the fol-
amples include the databases of national,
lowing issues:
provincial, and state nursing associations;
clinical databases from health mainte- The research question must be framed
nance organizations (HMOs); Canadian in a manner that enables the researcher
Medical Services Insurance databases; to answer it using the available data.
records of Medicaid and Medicare (USA); The researcher must be familiar with
reportable disease registers of the Centers the nature of the data collected, as well
for Disease Control and Prevention and as who collected it and how. This helps
the World Health Organization, as well as the researcher determine if the data set
those of provincial and state health de- is appropriate to answer the research
partments; local and regional hospital question. To fully understand the limi-
databanks; the Vital Statistics Registers, tations of the database, the researcher
which record events such as births, should be familiar with the clinical set-
deaths, divorces, and marriages; and an ar- ting from which the data emerged.
ray of national databases. In Canada, sev- Terminology of the database may be id-
eral national population surveys have iosyncratic. For example, the term unit
yielded interesting data. These include the of care, which has many meanings
Canada Health Promotion Survey; Statis- ranging from a physical space or a car-
tics Canada Violence Against Women Sur- ing episode to a period of time, may be
used in the database. Therefore, the re- vey has gained wide acceptance in the
searcher must become familiar with the research community as a cost-effective
vocabulary, data entry procedures, and approach to fully using previously col-
any associated documentation about lected data, secondary analysis of quali-
the system. tative data sets has not been as enthu-
The group of interest must be ade- siastically endorsed. In attempting a
quately represented in the database. If secondary analysis of qualitative data,
the researcher wishes to examine a several major challenges come to mind.
subpopulation of the data set, he or These concerns are identied and ad-
she must determine that sufcient in- dressed by Hinds and associates (1997,
formation on the subset exists. The na- pp. 409423) and summarized as follows.
ture of the sampling and measurement
in the database may seriously con- The rst issue to resolve is the appro-
strain a secondary analysis. priateness of qualitatively generated
Adequacy of method is a major con- data for re-analysis. Because the large
cern with an established database. The volumes of qualitative data generated in
researcher must accept the method a single investigation must be reduced
used to collect the original data and as part of data management, the re-
has no recourse to redesigning the ini- searcher selectively samples the data.
tial study. The result is that potentially important
Permission must be obtained to use the data may not be considered in the nal
database for research purposes if it is analysis if it does not contribute to the
not in the public domain. Subjects who emerging patterns or themes in the pri-
gave permission when the data were mary analysis. Hinds and associates
collected originally usually do not have (1997) suggest the t may best be deter-
to be asked again for permission unless mined by the extent of missing data.
the risk to participants condentiality Missing data are those that did not nd
is signicant. Human subjects commit- their way into the analysis process. For
tees will determine this. example, if an issue is explored in one in-
Many databases are not originally de- terview but not in previous or subse-
signed for research purposes, so the re- quent interviews, it may be that the topic
searcher must structure the data in a us- emerged spontaneously or that the re-
able format. This involves compiling a searcher decided to narrow the studys
workable data set from the database; scope and rene the interview ques-
checking the quality, accuracy, and reli- tions; therefore, the topic is not ad-
ability of information; eliminating re- dressed again. A secondary analysis of a
dundant and duplicate entries; complet- qualitative data set does not enable the
ing missing data that may be available researcher to answer his or her research
from earlier records in the database; question if the phenomenon of interest
and learning the meaning of any special is not uniformly addressed in the pri-
vocabulary or codes in the data. The mary study. The researcher must deter-
best way to correct redundant entries, mine this before proceeding further.
missing data, and other errors is to vi- Assessing the quality and appropriate-
sually inspect a printout of the data. ness of the qualitative data from the
primary study for secondary analysis
b. Using Qualitative Data Sets involves having direct access to it and
to members of the research team. Ac-
Although secondary analysis of quantita- cess can be negotiated to include taped
tive data such as a national census sur- interviews; hard copies; disk copies;
SURVEY DESIGNS 169
and copies of eld notes, memos, and the timing (either concurrent with the
interpretive notes. Negotiated access primary analysis or after it) is appro-
is crucial to determining if the data can priate for secondary analysis. Factors
be validly used for a secondary analy- to consider in doing so include whether
sis. Figure 5.1 illustrates a tool for as- the context in which the data were col-
sessing the appropriateness of a data lected has changed. If the phenomenon
set for secondary analysis. of interest has changed signicantly
The similarity between the question in since the primary study, a secondary
the primary research study and the analysis may yield very little new and
question for the secondary analysis relevant information. If done concur-
must be signicant. If the difference be- rently, the researcher has easier access
tween the two questions is considerable, to participants if conceptual clarica-
it is unlikely that the primary data set tion of secondary analysis is required
will yield an answer because it will lack but if premature sharing of ndings be-
appropriate depth and detail related to tween the two studies is done, one set
the secondary analysis question. may unduly inuence the other.
A pilot study using data from the pri-
mary study will help determine if the
data are capable of meeting the re- 4. Examples of Research Using
search objectives of the secondary Secondary Data Analysis
analysis. Usually three randomly se-
This section illustrates two examples of
lected interviews are sufcient to de-
secondary analysis, one using qualita-
termine if the study purpose can be
tively generated data and the other using
achieved or if it needs to be revised.
quantitatively generated data to investi-
Permission from participants in the
gate a nursing issue.
primary study to have their data rean-
alyzed by other researchers must be
sought either at the time of the initial a. Emancipatory Potential of
study consent or subsequently. Figure Storytelling in a Group
5.2 provides an example of a consent
form to use data for secondary analy- In this qualitative investigation, Banks-
sis. It is ideal if this can be sought with Wallace (1998) examined the health-
the initial study, but often researchers promoting function of storytelling in a
do not anticipate doing a secondary group of women through the use of sec-
analysis until after the primary study ondary analysis. All of the information was
is complete and additional phenomena made available through analyzing a subset
not represented in the study ndings of 115 stories derived from the transcripts
are noted in the data. of 28 women of African descent who par-
Researchers must be sensitized to the ticipated in the original study titled Sisters
context of the primary study. The pro- in Session (SIS): Focus Groups as a Re-
cess of qualitative data collection and search Tool and Intervention for Working
analysis is affected by the context in with Women of African Descent. The pur-
which it occurs; therefore, secondary poses of the original study were to in-
analysis needs to consider the context crease awareness of barriers to womens
and try as much as possible to get as participation in or conducting of research,
close to the experience of having been to identify research priorities for people of
there. African descent, and to explore the thera-
Researchers must determine whether peutic potential of focus groups. The sec-
primary data sets are still current and if ondary data analysis enabled the re-

SURVEY DESIGNS 171

searcher to answer a research question prospective study of insufcient milk

that was closely related to one of the pur- syndrome (IMS). Original study consent
poses of the initial study, namely, What is forms allowed for additional unspecied
the role of storytelling in promoting health analysis of the milk samples. Humenick et
among SIS participants? (Box 5.10). al. were interested in studying variables
that mediate the relationship between el-
evated BM [Na] levels and breastfeed-
b. Breast Milk Sodium as a Predictor ing outcomes. Although the number of
of Breastfeeding Patterns variables they could study were limited
to those present in the original data, the
The quantitative study by Humenick and secondary analysis enabled them to ex-
associates (1998) was designed to repli- plore the inuence of important psy-
cate and extend the work of Morton chosocial risk factors on breastfeeding
(1994) which reported that elevated BM outcomes. Box 5.11 provides a discus-
[Na] during early lactogenesis was pre- sion of the ndings of this secondary data
dictive of poor breastfeeding outcomes. analysis study.
Humenick and associates were able to
meet their research objectives by con-
ducting secondary analysis of 6-day post- G. ADVANTAGES AND
partum breast milk and maternal data LIMITATIONS OF SURVEY
collected from a subset of 41 mothers DESIGNS
that were part of the original study. The
original study consisted of 340 subjects Sample surveys are the major tools in stud-
selected for inclusion in a longitudinal, ies attempting to represent large popula-
SURVEY DESIGNS 173
EMANCIPATORY POTENTIAL OF STORYTELLING IN A GROUP

Therapeutic benefits of storytelling were noted (4) to provide a means of catharsis, (5) to pro-
in the nursing literature more than 60 years ago. vide a means for resisting oppression, and (6) to
Recently there has been a renewed interest in provide a vehicle for educating other partici-
the scientific value of storytelling. The current pants. (For a detailed description of each of these
project used a descriptive, secondary analysis functions with illustrative stories, interested
design to examine the health-promoting func- readers should check the original source).
tions of storytelling in a convenience sample of Storytelling was used as both a means of
28 African-American women. learning more about factors affecting the well-
Data were audiotaped in four focus groups being of women of African descent and as a tool
convened during a 6-week period. The six-part for improving the lives of study participants.
narrative analysis included (1) providing infor- The stories shared by women provided insight
mation about the historical setting for the study, into factors that constrained their choices or
(2) demarcating transcripts into individual sto- enabled them to move further along their jour-
ries, (3) analyzing the context in which the sto- ney. Listening to the women and designing a
ries occurred and the content of specific stories, study that centered their experiences resulted
(4) grouping stories according to themes and in a project that not only provided answers to
functions, (5) comparing story themes and func- the researchers questions but also allowed par-
tions across sessions, and (6) reviewing stories ticipants opportunities to learn and grow.
for conspicuous absences. Through storytelling, participants were better
Six major functions of storytelling in the able to make decisions related to living as a
group were identified: (1) to provide contextual woman of African descent. Group storytelling is
grounding, (2) to provide a means of bonding a tradition that continues to sustain communi-
with other participants, (3) to provide a means ties of African descent and other cultures
of validating and affirming womens experiences, throughout the world.
SOURCE: Summarized from Banks-Wallace, J. (1998). Emancipatory potential of storytelling in a group. Image: Jour-
nal of Nursing Scholarship, 30(1), 1721.
tions. They often attempt to deal simulta- were born in or the salary of your rst full-
neously with many variables and attempt time job), it may be very difcult to recall
to describe the complexity of human be- how well you got along with your mother
havior. Surveys also permit researchers to when you were 6 years old. Another fun-
construct new variables by combining a damental issue has to do with the connec-
number of characteristics; status integration between words (what people say they
tion or status crystallization are examples will do) and deeds (what they actually do).
of such constructed variables. There are therefore doubts about the ex-
Despite the many advantages of survey tent to which surveys reect ideal behav-
research, it is difcult to make clear causal ior as opposed to real behavior. To what
inferences from such data. Most of these extent, then, can we claim to reect reality
difculties derive from the nature of the with survey data?
data. Survey data are normally based on Table 5.3 summarizes the advantages
self-reports and frequently involve data and limitations of the rst two approaches
based on recollection. Although it is prob- to research design that we have consid-
ably not difcult to recall factual informa- ered so far, experiments and surveys.
tion (e.g., reporting the community you Causal inferences are clearest when exper-
BREAST MILK SODIUM AS A PREDICTOR OF BREASTFEEDING

PATTERNS
Findings be enhanced by combining it with the psy-
Consistent with the findings of the original chosocial variables of support density and self-
study, 80 percent of those with a BM [Na] of perception of breastfeeding by the mother.
16 mmol/L or lower at day 6 sustained a high Summary
level of breastfeeding at week 4, compared with In summary, this secondary analysis provided
only 50 percent of those with an elevated BM the following answers to the research ques-
[Na] (2 4.05, df 1, P .04). This differ- tions:
ence was even greater in a subgroup of moth-
ers predicted to be at risk for insufficient milk Elevated levels of BM [Na] at postpartum
supply on the basis of support density and self- day 6 do predict a significant decrease in
perception variables. Of the latter group, 75 breastfeeding sustainment at week 4.
percent with low BM [Na] sustained a high When Insufficient Milk Syndrome (IMS) risk
level of breastfeeding at 4 weeks postpartum, group was used as a mediating variable, high-
compared with only 22 percent with an ele- risk mothers showed a striking decrease in
vated BM [Na] (2 .65, df 1, P. .01). sustained breastfeeding at week 4 (75 per-
In contrast, among the low-risk mothers, cent compared with 22 percent) for low and
BM [Na] levels were not associated with any elevated BM [Na] groups, respectively.
difference in breast milk sustainment (89 per- For mothers who had been judged to be at low
cent and 82 percent sustainment for low- and risk for IMS, level of BM [Na] showed no
high-sodium groups, respectively). Thus, a nor- prediction of future breastfeeding patterns.
mal decrease in BM [Na] is predictive of Thus, BM [Na] appears to be a marker pre-
higher sustainment of breastfeeding. However, dicting poor breastfeeding outcomes only
the predictive validity of this marker appears to for mothers at high risk for perceived IMS.
SOURCE: Summarized from Humenick, S., Hill, P., Thompson, J., and Hart, A. (1998). Breast-milk sodium as a pre-
dictor of breastfeeding patterns. Canadian Journal of Nursing Research, 30(3), 6781.
imental and, to a lesser degree, panel data umn is there because such groups are typ-
are used. Generalizing about larger popu- ically selected in the most convenient
lations is the fort of survey designs. way. Probability procedures have there-
Although neither surveys nor experi- fore not been followed in the way that
ments rate high on validity, this does not would be necessary if one wished to make
mean that they are, by nature, invalid. The extrapolations to the general population.
negative sign simply means that this is a (See Chapter 15 for details on sampling.)
problem area in these designs. It is dif- The major limitation of research based
cult to demonstrate that respondents on secondary data is that the necessary in-
have not altered their behavior or their an- formation is often unavailable, incom-
swers on a questionnaire in response to plete, or inappropriate for the purposes of
the fact that they know they are being the secondary analysis. But with imagina-
studied. Both experimentalists and survey tion, excellent work can be done. Health
researchers take the problem of validity researchers are impressively adept, for
seriously and try to minimize distortions. example, at locating indirect indicators
The negative sign for group-administered in available data (e.g., using the number of
questionnaires in the generalization col- hits on the dietary or cholesterol educa-
SURVEY DESIGNS 175

Research Design Category General Validity Causal Inference Multivariate Probing
Experimental Designs
Pre-experimental
Experimental
Quasi-experimental
Survey Designs
Individual questionnaire
Group administered
Phone survey
Interview
Comparative analysis
Secondary data
Meta analysis
In each category, a means that this is an advantage of the technique; means this is a possible limitation; means
that in some conditions it is a limitation, but in others it is an advantage.
Causal inference refers to the ease with which inferences about causal relations among variables may be made.
tional page of the American Heart Founda- more geographically dispersed samples
tions Web site as an indirect measure of can be studied through the use of existing
the usefulness of that page as an educa- databases rather than limiting the sample
tional tool). And if the data are reported at to what a researchers funds can support
different points in time, a researcher who for an original study. Finally, the collect-
is interested in tracking and explaining ing and tracking of longitudinal data and
health trends may be able to build and test trends over time are made reasonably af-
elaborate causal models. So, whether they fordable with the use of secondary analy-
are attempting to understand the link be- sis of databases and archival materials.
tween diet and health status or between Whatever research design is selected, it
cardiac disease and utilization of health is important to understand the strengths
care resources, many researchers make and limitations of each of the approaches.
efforts to understand the world by using Such knowledge helps researchers select
secondary data. appropriate designs and try to deal with
Secondary data analysis is cost effec- the weak points of each design. When a de-
tive and, providing the data are relatively cision has been made to use a survey
complete, can lead to sound general state- design, Chapters 13, 14, and 15 contain
ments about the world. Many major re- information of particular relevance to the
search costs (e.g., personnel salaries and survey researcher; measurement, ques-
data collection costs) have already been tionnaire construction, and sampling meth-
assumed by the initial study. Larger and ods are all covered there.
E X E R C I S E S
1. Suppose you wished to understand the 3. Suppose you are interested in exploring
factors that inuence the grade perfor- gender differences in attitudes con-
mance of university nursing students. cerning the acceptability of permissive
What kind of design would you recom- parenting practices. What design would
mend for such a study? Outline the ra- you recommend to explore such a
tionale for your choice. question? Outline the rationale for your
choice.
2. Suppose you wished to investigate four

alternative explanations for the rela- 4. Suppose you wished to explore the fre-
tionship between social support and quency and the intensity with which
adolescents perceptions of stress. female undergraduates have experi-
What kind of design would you recom- enced sexual harassment in their lives.
mend? Outline the rationale for your What variables would you wish to
choice. measure in such a study?
Brentro, M., and Hegge, M. (2000). Nursing fac- F.A. Davis. This text and workbook provide
ulty: One generation away from extinction? a brief but useful discussion of the survey
Journal of Professional Nursing, 16(2) 97 as a data collection method.
103. A state-wide survey of nurses with Fowler, F.J. (1993). Survey Research Methods
graduate degrees was undertaken to un- (2nd ed.). Newbury Park, CA: Sage. This
derstand the career paths and intentions of text provides a comprehensive overview of
nurses in one midwestern state. survey design, the process, advantages,
Clark, D., Clark, P., Day, D., and Shea, D. (2000). disadvantages, and practical examples.
The relationship between health-care re- Humphris, D. (1999). Types of evidence. In
form and nurses interest in union repre- Hamer, S., and Collinson., G (Eds.). Achiev-
sentation: The role of workplace climate. ing Evidence-Based Practice: A Handbook for
Journal of Professional Nursing, 16(2) 9296. Practitioners. Edinburgh: Harcourt Publish-
This is an example of an exploratory sur- ers. This chapter explores the survey as a
vey design using a mailed instrument to method of providing evidence-based infor-
1500 registered nurses. mation for use in practice. In addition, it ex-
Fain, J.A. (1999). Reading, Understanding, and plores how to make judgments about the
Applying Nursing Research. Philadelphia: quality and usefulness of such evidence.
Chapter 6
Qualitative Research
Designs
CHAPTER OUTLINE
A. Entering the World of Qualitative 3. Grounded Theory Perspective
Research 4. Ethnographic Perspective
1. Dening Qualitative Research C. Scientific Adequacy of Qualitative
2. Core Activities in Qualitative Research Research
B. Qualitative Research Designs 1. Traditional Standards of Quality
1. Characteristics of Qualitative Research 2. Contemporary Standards of Quality
Designs D. Advantages and Limitations of
2. Phenomenological Research Perspective Qualitative Approaches
KEY TERMS
Audit trail Emic Participant

Bracketing Ethnography Phenomenological
Componential analysis Ethnonursing perspective
Confirmability Etic Qualitative research
Constant comparative Gatekeeper Saturation
method Grounded theory Simultaneity paradigm
Core variable Holistic ethnography Symbolic interactionism
Credibility Informant Taxonomy
Culture Intersubjectivity Themes
Cultural scene Key informant Theoretical memos
Cultural themes Lived experience Trustworthiness
Dependability Maxi ethnography Totality paradigm
Domain analysis Mini ethnography Transferability
177
T
he previous two chapters pre- perspective, one that attempts to tap into
sented introductions to experimen- clients perceptions of themselves and of
tal and survey approaches to nurs- how they see others and for a description
ing research. Both are predominantly on the process of how they learn to act,
quantitative (using numbers to summa- live, and die in a hospice setting.
rize the ndings of the research) and are Fortunately, some qualitative perspec-
associated with the positivist, scientic tives in nursing science pay attention to
approach that emphasizes careful mea- such matters. Qualitative approaches are
surement, sampling, rationality, objectiv- dened as those that use words rather
ity, prediction, control, and the logic of than numbers to describe ndings, as-
hypothesis testing. A vast number of re- sume a dynamic reality, and emphasize
search questions can be posed and seeing the world from the eyes of the parti-
answered using these approaches. So cipants being studied. The goal of qualita-
whether it is attempting to describe and tive approaches is understanding rather
explain the patterns surrounding the rela- than prediction and they emphasize the
tionship between gender and suicide or subjective dimensions of human experi-
trying to determine the effectiveness of a ence. They are holistic (i.e., studying the
prenatal education program in improving whole institution, group, or culture) rather
maternal and infant outcomes, a quantita- than reductionistic. Generally, they are as-
tive approach is likely to be used. When sociated with the interpretive approach,
well done, these studies have high levels which is discovery oriented, explanatory,
of credibility because they are using sci- descriptive, and inductive in nature (see
entic techniques that are fairly well un- Chapter 1).
derstood and respected in our culture. This chapter introduces the most fre-
The problem with limiting oneself to quently used qualitative approaches in
positivistic approaches is that there is a nursing research, including phenomenol-
whole range of human activity that is not ogy, grounded theory, and ethnography.
easily understood using positivistic ap- Most qualitative researchers do not sub-
proaches. For example, if you wanted to scribe to the broad classication of re-
know what it is like to be a terminally ill search designs proposed in this text. Al-
client on a hospice unit, one would be though some may refer to their designs
quite limited in getting a real understand- as case studies ( i.e., exploration of a sin-
ing of the hospice experience by conduct- gle entity bounded by time and activity
ing a survey on client perceptions and atti- using a variety of data collection proce-
tudes toward hospice care. A great deal of dures) or eld studies ( i.e., observations
information could be gleaned from such a of behavior in natural settings, in which
study, but could you really understand the participants are acting as they usually
experiences of the terminally ill? Could do), many simply describe their perspec-
you understand the complex relations be- tive as grounded theory, phenomenol-
tween the clients, nurses, family members, ogy, biography, and so on and go on to ex-
physicians, and other care attendants? A plain what they do
reality is ignored if one limits oneself to Data collection and analysis methods
survey or experimental approaches. In this specic to the approaches discussed in
example, can you make much sense out of this chapter are described in Chapter 7.
how the hospice unit works within the These include participant observation
health-care institution or the complexity of studies, in-depth interviewing, focus group
how a client experiences the transition interviews, observational studies, and eld
from active treatment into hospice care? experiments. Let us begin by dening qual-
These kinds of questions beg for a holistic itative research and comparing the core
QUALITATIVE RESEARCH DESIGNS 179
research activities required for doing a In the simultaneity paradigm, reality

qualitative investigation with those re- is characterized by a mutual process of
quired for a quantitative study. Common human and environment, and persons ex-
characteristics of qualitative designs are ist in open participation with the uni-
discussed as a means of establishing a verse and are more than and different
foundation for describing the three ap- from the summed parts studied in the to-
proaches discussed in this chapter. Exam- tality paradigm. The wholeness or uni-
ples of qualitative designs are provided tary nature of the human being is primary
from the research literature. The examples in this paradigm. Thus, nurse scientists
are useful models for research students study the universal lived experiences of
who are interested in designing a qualita- health or patterns of the whole. A study
tive study. in this paradigm is conducted not to con-
trol the human being and the world but to
gain insights into human living. The re-
searcher believes in the personal mean-
A. ENTERING THE WORLD OF
ing individuals give to their lived experi-
QUALITATIVE RESEARCH
ence. The simultaneity paradigm is best
suited for the qualitative research ap-
Several leading authors have written per-
proach and is associated with the inter-
sonal perspectives on qualitative re-
pretive tradition (see Chapter 1).
search (Creswell, 1994, 1998; Denzin and
Parse and associates (1985) offer a def-
Lincoln, 1994; Marshall and Rossman,
inition of qualitative research that is
1995; Munhall and Oiler, 1986; Parse et al.,
rooted in nursings simultaneity para-
1985; Silverman, 1997). Their work has
digm. They state:
been instrumental in advancing our un-
derstanding of this research tradition. It Qualitative research identies the char-
is a eld of inquiry that cuts across disci- acteristics and the signicance of hu-
plines, elds, and subject matter and has man experiences as described by sub-
much to offer nurses and others who are jects and interpreted by the researcher
interested in gaining a holistic under- at various levels of abstraction. In qual-
standing of human experiences. itative research the researchers inter-
pretations are intersubjective, that is,
given the researchers frame of refer-
1. Defining Qualitative Research ence, another person can come to a sim-
ilar interpretation. Qualitative data are
To understand the place of qualitative re-
processed through the creative abstrac-
search requires discussion of the two ma-
tions of the researcher as the subjects
jor paradigms that are reected in nursing
descriptions are studied to uncover the
science today. Parse (1987) has named
meaning of human experiences (p. 3).
these belief systems the totality paradigm
and the simultaneity paradigm. In the total- From this denition, we see that the
ity paradigm, reality exists independent of aim of qualitative research in nursing is
the knower, humans are viewed as the sum an understanding of what it means to be
of their parts, persons adapt to an external human. Nursing science needs to be open
environment through cause-and-effect re- to the variety of ways of knowing and
lationships, and health is viewed along a learning that are evolving from the two
continuum from illness to wellness. The to- paradigms. Each presents challenges to
tality paradigm is best suited for the quan- researchers who are working to ground
titative research approach and is associ- nursing practice in nursing science.
ated with the positivist tradition. A generic denition of qualitative re-
search is proposed by Denzin and Lin- They go on to discuss the qualitative

coln (1994, p. 2). They note: researcher as a bricoleur, which is a Jack
of all trades or do-it-yourself person.
It is multimethod in focus, involving an The multiple methodologies of qualita-
interpretive, naturalistic approach to its tive research are dened as a bricolage or
subject matter. This means that qualita- a pieced-together, close-knit set of prac-
tive researchers study things in their nat- tices that provide solutions to a problem
ural settings, attempting to make sense in a concrete situation. This metaphor is
of, or interpret phenomena in terms of helpful in concretely illustrating the com-
the meanings people bring to them . . . It plex and multifaceted nature of qualita-
involves the use of a variety of empirical tive research. Box 6.1 summarizes their
materials-case study, personal experi- description of bricoleurs.
ence, introspective, life story, interview, The eld of qualitative research is far
observational, historical, interactional, from a unied set of principles that are
and visual textsthat describe routine promoted by scholars with shared vi-
and problematic moments and mean- sions. Rather, Denzin and Lincoln discov-
ings in individuals lives. Accordingly, ered that the eld is dened primarily by a
qualitative researchers deploy a wide series of essential tensions, contradic-
range of interconnected methods, hop- tions, and hesitations. These tensions
ing always to get a better x on the sub- work back and forth among competing
ject matter at hand (p.2). denitions. Although there are differing
BOX 6.1 The Qualitative Researcher as Bricoleur
The multiple methodologies of qualitative re- sonal history, biography, gender, social class,
search may be viewed as a bricolage and the re- race, and ethnicity, and those of the people in
searcher as the bricoleur. The bricoleur is adept at the setting. The bricoleur knows that science is
performing a large number of diverse tasks, power because all research findings have polit-
ranging from interviewing to observing, to in- ical implications. There is no value-free science.
terpreting personal and historical documents, to The bricoleur also knows that researchers all tell
intensive self-reflection and introspection. The stories about the worlds they have studied.
bricoleur reads widely and is knowledgeable Thus, the narratives or stories scientists tell are
about the many interpretive paradigms (e.g., accounts couched and framed within specific
feminism, Marxism, cultural studies, construc- storytelling traditions, often defined as para-
tivism) that can be brought to any particular digms (e.g., positivism, postpositivism, and con-
problem. He or she may not, however, believe structivism).
that paradigms can be mingled or synthesized The product of the bricoleurs labor is a brico-
that is, paradigms are overarching philosophical lage, which is a complex, dense, reflexive, collage-
systems denoting particular ontologies, episte- like creation that represents the researchers im-
mologies, and methodologies and cannot be eas- ages, understandings, and interpretations of the
ily moved between. The researcher as bricoleur world or the phenomenon under analysis. The
theorist works between and within competing bricolage connects the parts to the whole, stress-
and overlapping perspectives and paradigms ing the meaningful relationships that operate in
The bricoleur understands that research is an the situations and the social worlds studied
interactive process shaped by his or her per- (Weinstein and Weinstein, 1991, p. 164).
SOURCE: Adapted from Denzin, N.K., and Lincoln, Y. (1994). Handbook of Qualitative Research. London: Sage.
views on how one may dene qualitative it with quantitative research (Brink and
research and its many perspectives, most Wood, 1998; Burns and Grove, 1997; Fain,
would agree that its essence includes a 1999; LoBionda-Wood and Haber, 1998).
commitment to the naturalistic, interpre- These distinctions are helpful in con-
tive approach to its subject matter. Denzin trasting assumptions and values that are
and Lincolns denition espouses this a profoundly different in each perspective.
priori perspective, which is grounded in Table 6.1 summarizes the distinctions re-
the philosophical assumptions of the in- ported by these authors. The table is not
terpretive worldview, and highlights the intended to be exhaustive; rather, it high-
importance of multiple sources of infor- lights the key differences for each re-
mation available to researchers. search tradition.
The work of Denzin and Lincoln (1994) Having dened qualitative research, it
has been expanded by Creswell (1998, p. is important to note that it is comple-
15). His denition emphasizes the commentary to quantitative research. Both
plex, holistic picture created by qualita- processes produce different kinds of
tive research. He writes: nursing knowledge that are valued by the
profession and both are needed to pro-
Qualitative research is an inquiry
mote excellence in practice. Qualitative
process of understanding based on dis-
and quantitative methods enable nurses
tinct methodological traditions of in-
to engage in the systematic collection
quiry that explore a social or human
and analysis of data on the organization,
problem. The researcher builds a com-
delivery, and outcomes of nursing care
plex, holistic picture, analyzes words, re-
for the purpose of enhancing the clients
ports detailed views of informants, and
health and understanding the meaning of
conducts the study in a natural setting.
health and the health-care experience
This denition incorporates the idea from the perspective of the client (Para-
that the researcher is an instrument of hoo, 1997). Both qualitative and quantita-
data collection who focuses on the mean- tive research contribute signicantly to
ing of participants experiences in social the knowledge base required for the pro-
contexts. The use of descriptive language vision of quality care and excellence in
that emerges from the complex narrative practice.
of participants is critical to Creswells per-
spective. His denition adds the dimen-
sion of distinct methodologies within the 2. Core Activities in Qualitative
traditions of inquiry. For Creswell, these Research
traditions include the historians biogra-
phy, the psychologists phenomenology, In many ways, the format for a qualitative
the sociologists grounded theory, the an- investigation follows the traditional re-
thropologists ethnography, and the politi- search approach of asking a question or
cal scientists case study. Although some presenting a problem, collecting data, an-
of these traditions are explored in this alyzing the data, and presenting ndings.
chapter, a full discussion is beyond the There are, however, important distinc-
scope of this text. For a complete descriptions in the manner in which a qualitative
tion of each tradition, interested readers investigation is designed and conducted
are directed to Creswells text in the Rec- as compared with a quantitative study.
ommended Readings at the end of this Figure 6.1 illustrates the formats for de-
chapter. signing a qualitative and a quantitative
A signicant number of researchers study. Streubert and Carpenter (1999)
dene qualitative research by comparing identify several core activities that are
Table 6.1 Qualitative and Quantitative Research Contrasted

Qualitative Quantitative
Multiple realities Single reality
Reality is socially constructed Reality is objective
Reality is context interrelated Reality is context free
Holistic Reductionistic
Strong philosophical perspective Strong theoretical base
Reasoning is inductive Reasoning is deductive and inductive
Discovery of meaning is the basis of knowledge Cause-and-effect relationships are the bases of
knowledge
Develops theory Tests theory
Theory developed during study Theory developed a priori
Meaning of concepts Measurement of variables
Process oriented Outcome oriented
Control unimportant Control important
Rich descriptions Precise measurement of variables
Basic element of analysis is words Basic element of analysis is numbers
Uniqueness Generalization
Trustworthiness of ndings Control of error
different in the two approaches. These data analyzed. Although some qualitative
are summarized next. investigators accept that a cursory re-
view may be conducted initially to help
focus the study, purists argue that the re-
a. Literature Review view should not occur until completion of
the study. The rationale for this is to pre-
A quantitative investigation involves an
vent the researcher from leading partici-
extensive review of the literature early on
pants in the direction of what has already
in the research process to establish the
been discovered about the phenomena of
need for the investigation. The researcher
interest. The purpose of the literature re-
must be familiar with what others in the
view in qualitative studies is to show how
eld have reported about the research
the current ndings t into what is al-
topic. This information is important so
ready known about the research topic.
the researcher can have condence that
this investigation is the next appropriate
one to advance the particular eld of b. Explicating Researchers Beliefs
knowledge. The literature review helps
the quantitative investigator rene the re- Before beginning most qualitative investi-
search question and situate the question gations, the researcher engages in a
within a larger theoretical framework. process known as bracketing. This does
In qualitative investigations, the litera- not occur in quantitative studies or in
ture review is usually conducted after the qualitative research done using Herme-
research has been conducted and the neutics (a type of phenomenology attrib-
Figure 6.1 Steps to design a research study: Quantitative and qualitative.
uted to Heidegger). Bracketing is a cogni- to the phenomena studied. To bracket

tive process used by researchers to set ones ideas about the research phenome-
aside ones biases and personal perspec- non, the investigator may keep a diary in
tives about the research topic. Its purpose which he or she writes personal thoughts
is to make known what the researcher be- and feelings about the topic. After these
lieves about the research topic so that the ideas are disclosed, they can be set aside.
researcher can then approach the topic Through bracketing, the researcher is
honestly. made aware of when data collection and
Bracketing is a critical step in most analyses reect personal beliefs rather
qualitative investigations because of the than those of the participants. For exam-
subjective nature of the investigation and ple, you may be interested in exploring
the close relationship of the investigator the phenomenon of meaning in life in el-
derly men. Perhaps you have worked with are usually conducted in controlled set-
elderly clients in a senior citizen complex tings where the researcher is detached
for many years. Based on this experience, from the subject. The researcher uses
you have come to believe that meaning in predetermined tools such as question-
life in this age group is explained by the naires or structured interviews to collect
presence of deep spiritual beliefs. If you the data.
do not acknowledge this personal per-
spective, you may lead participants in
their description of meaning in life in the d. Selection of Participants
direction of your own beliefs. By record-
ing ones beliefs in writing, the researcher The term participant or informant is
is able to see more clearly what is real for used to refer to the individuals who in-
the participants and the researcher. form the study in qualitative investiga-
tions. Participants are viewed as active
participants in the research process and
c. Setting for Data Collection equal partners with the investigator.
They want to help those interested to bet-
Qualitative investigators always collect ter understand their lives and the social
data in the eld. The eld is the natural contexts in which they live and create
world where participants live and experi- meaning. To do so, it is essential that par-
ence life. Qualitative researchers are more ticipants have rsthand experience with
likely than quantitative researchers to con- the phenomenon under investigation. For
front the constraints of collecting data in example, if you wished to study the
the everyday social world of participants. expert competencies of public health
Whereas they embed their ndings in this nurses in nursing vulnerable families, you
world, quantitative researchers abstract would have to select participants who,
from this world. The close social interac- indeed, were expert community health
tion that occurs in eld settings may either nurses with experience in visiting vulner-
enhance or limit data collection for the able families. This type of sampling is re-
qualitative investigator. The eld requires ferred to as purposive or purposeful. It im-
reciprocity in terms of decision making. plies a commitment to interviewing and
The participant and the researcher negoti- observing participants who have had ex-
ate which data will be shared, and this may perience with the phenomena of interest.
change through the course of the study. The purposeful selection of partici-
Participants must understand and be will- pants in qualitative investigations is a
ing to express their inner thoughts and strategically important decision in the
feelings about the phenomenon. For exam- study design. Regardless of the qualita-
ple, you are interested in studying the ex- tive approach selected, the researcher
pert competencies of public health nurses should establish clear criteria and ratio-
who visit societys most vulnerable fami- nale for selection purposes.
lies. You would need to negotiate with In quantitative investigations, individu-
nurses their stories about making a differ- als who participate in the study are usually
ence in the outcomes of families at risk. As referred to as subjects or respondents. This
participants reveal to you the essence of reects the position that subjects are
their work with families, your questions acted upon in such studies. Researcher
and probes for information may change. involvement is usually as an objective ob-
You will need to continually negotiate with server maintaining distance from the re-
the participants to elicit their stories. search subjects. Data are collected using
In quantitative investigations, studies structured instruments. Random selection
is often used, particularly in survey de- Qualitative data analysis is a cyclical

signs, and the number of subjects must be process of questioning and verifying the
large enough to establish signicance (see data ndings. The recursive process of
Chapter 15). Large sample sizes and ran- analysis begins immediately with the rst
dom selection are not required in qualita- data collection episode and continues
tive investigations because the goal is not throughout the study. Reading, reread-
generalization of ndings but rather the ing, intuiting, analyzing, synthesizing and
provision of rich descriptions of the phe- reporting on the data are required activi-
nomena by those who have experienced it. ties. Data collected from interviews or ob-
servations are continuously reviewed to
identify additional questions to be ex-
e. Saturation plored with participants. Data from early
analysis is routinely returned to partici-
Saturation is a term used in qualitative in-
pants for clarication or elaboration. As
vestigations to refer to a situation in data
the investigators review the data, they
collection in which the participants de-
must carefully look for meaning in the
scriptions become repetitive and conrm
data. At the conclusion of the study,
previously collected data. The process of
there is a protracted period of data im-
saturation is something that the re-
mersion during which the investigator re-
searcher looks for as an indication that
views all previous conclusions that have
data collection is complete. It means that
emerged, in the context of the whole
data collected from new participants con-
study. Data that are similar in meaning
rms previously collected data rather than
are clustered together into preliminary
adding new information. Saturation in
categories.
qualitative investigations is similar to ade-
In quantitative data analysis, the pur-
quate sample size determination in quanti-
pose is to answer the research question,
tative investigations. Saturation of data
test hypotheses, or both. Data analysis
gives the researcher condence that the
does not begin until after all data collec-
description of the phenomenon has been
tion is complete. Statistical procedures
captured. Similarly, random selection and
and computer programs such as Statisti-
statistical sampling that produce signi-
cal Package for the Social Sciences (SPSS)
cance give the quantitative researcher
are used to describe and analyze data.
condence that the results reect true re-
The research design and the type of data
sults that can be generalized to others.
collected determine the appropriate sta-
tistical tests to be applied. Data analysis
f. Data Analysis in quantitative investigations generally
requires less time than in qualitative data
Major differences exist in the data analysis analysis. The researcher must have a
processes of quantitative and qualitative thorough knowledge of statistics or ac-
investigations. In qualitative research, in- cess to a statistical consultant to ensure
vestigators must immerse themselves that appropriate statistical tests are ap-
fully in the data to bring order and mean- plied to the data. Inappropriate applica-
ing to the vast amounts of narrative that tion of statistics can lead to unjustied re-
result. This requires that one come to sults and inappropriate conclusions.
truly understand what the data are saying. Strategies of data analysis are discussed
It involves an extensive amount of time under each methodology because subtle
(i.e., weeks or months) spent in dialogue differences exist between them. Chapter
with the data using inductive reasoning 16 addresses aspects of data analysis in
(see Chapter 2). more detail.
B. QUALITATIVE RESEARCH differently in various social contexts. Re-

DESIGNS searchers must come to know these dif-
ferent realities and report them through
The use of qualitative methods to study use of extensive quotations and identi-
human phenomena has taken on impor- cation of themes that reect the de-
tant signicance in nursing. Indeed, some scriptions of the participants.
would say that a quiet revolution has taken 2. The researcher is committed to identi-
place in nursing research over the past fying a perspective to understanding
decade. This has been marked by an in- that will support the phenomena stud-
crease in the number of qualitative studies ied. Qualitative researchers address
reported in nursing journals and research questions about the phenomena of
textbooks, as well as an increase in the interest by using multiple methods or
number of qualitative research confer- perspectives to answer the question.
ences at both national and international Therefore, the discovery leads the
levels, and an increase in the number of choice of method rather than the me-
nurse researchers who are presenting re- thod leading the discovery. In many
ports of qualitative investigations in such qualitative investigations, a combina-
forums. Many nurses are now focusing on tion of methods, such as focus groups,
qualitative perspectives to enhance un- unstructured interviews, and observa-
derstanding of the human experience of tions, are required to fully understand
health and illness and subsequently to im- the phenomena of interest. Method and
prove practice. This chapter examines data collection strategies may change
three perspectives to qualitative research: as the study unfolds because the re-
phenomenology, grounded theory, and searcher is committed to using what-
ethnographic methods. Although these ever perspective best enables an under-
perspectives are quite different, they do standing of the phenomena of interest.
share a common essencethat of under- 3. The researcher is committed to the
standing and examining meaning in human participants points of view. Qualitative
experiences. researchers are co-participants in the
inquiry process. They seek to under-
stand the participants life experiences
1. Characteristics of Qualitative as constructed in their social worlds.
Research Designs Qualitative researchers become deeply
involved with participants immersing
Streubert and Carpenter (1999) identify
themselves in the eld to understand
six characteristics of qualitative investi-
the phenomena of interest from the par-
gations. Each warrant comment before
ticipants perspective. Extensive inter-
examining the three qualitative designs
views and observations are conducted
selected for discussion in this chapter.
to provide a view of reality that is im-
1. Multiple realities exist. There is not just portant to those studied rather than
one reality, one truth. Rather, all quali- what is important to the researcher.
tative researchers believe there are The researcher gains an insider per-
many realitiesthose of the researcher, spective. For example, Lee (1997) con-
the participants being studied, and the ducted a phenomenological study of the
readers of the research report. Through meaning of menopause as experienced
social interaction, participants come to by modern Korean women. To under-
know and understand phenomena in dif- stand this human health experience, in-
ferent ways and construct their realities. depth interviews were conducted with
Different people come to know things 65 women ages 40 to 60 years to reveal
the covert meaning of menopause. Par- 5. The researcher, as a participant-

ticipants were asked to reveal their observer, is considered to be an in-
thoughts, feelings, perceptions, and ex- sider. As the previous example illus-
pectations related to menopause and trated, the researcher is considered to
how they were coping with them. Four be the research instrument. This re-
patterns emerged from the lived experi- quires an acceptance of the investiga-
ence of the 65 women: (1) from suffering tor as part of the research study. The
to comfort, (2) from oppression to free- researcher plays a role as observer, in-
dom, (3) from being a good wife and terviewer, and interpreter of data;
mother to becoming a woman, and (4) therefore, qualitative investigators ac-
from a productive life to a transformed cept the subjective bias inherent in
life. qualitative investigations. The ac-
4. Inquiry is conducted in a way that knowledgment of the subjective na-
does not disturb the natural context of ture of qualitative research is funda-
the phenomena studied. Researchers mental to the conduct of qualitative
collect data in natural settings where inquiry.
people are acting as they usually do. 6. The research report is written in a lit-
There is considerable onus on investi- erary style rich with participants
gators to be as nonintrusive as possi- comments. The ndings of qualitative
ble. Researchers try to limit the dis- research focus on the experience of
tance between themselves and those the participants. The use of rst per-
being studied. Prolonged time is spent son ( i.e., I), metaphors, and a focus
in the eld to reduce the objective sep- on personal stories add to the richness
arateness that exists in quantitative of the report and to the understanding
studies. For example, Anderson (1996) of the meaning of the social interaction
spent extensive time in the eld to dis- of those studied. The language is very
cover how adolescents make decisions personal and emerges as the partici-
about substance abuse while in juvenile pants tell their stories.
detention. An ethnographic design was
used to learn, interpret, and describe Qualitative research is well suited to
perspectives of adolescents as they go many nursing investigations in which the
about the business of living their lives goal is to develop a deep understanding
and making sense out of their world of human experiences and the meanings
in detention. The researcher became participants attribute to these experi-
the data collection tool, learning how ences. Life processes such as maturing,
things worked in the closed institu- parenting, grieving, adapting to change,
tional setting of a detention facility by coping with stress, and dying are all phe-
living and working in the setting. Partic- nomena of central concern to nurses as
ipant observation, focus group inter- they work with individuals and family
views, and individual interviews were members experiencing these processes.
used to explore the dynamics among Qualitative exploration of these pro-
young women in juvenile detention as cesses provides nurses with critical in-
they discussed their resolutions for the sights and a deep understanding of par-
future. The presence of the investigator ticipants experiences. Such insights
in the natural setting over time (in this could not be gleaned from a reliance on
case, the detention center) usually en- strictly quantitative approaches.
ables the investigator to establish an in- Developing an understanding of the
clusive understanding of the particular full range of human experiences that one
phenomena of interest. encounters in the nursing world requires
a multitude of methods. Following is a elements related to the true meaning of

discussion of three frequently used qual- something that gives common under-
itative approaches in nursing science. standing to the phenomenon under study
Each is discussed in terms of its deni- (Spiegelberg, 1975; Fain, 1999). For exam-
tion, philosophical tenets, and applica- ple, in a phenomenological investigation of
tion of the method. Unique features of how nurses experience caring for dying
each perspective are highlighted and re- patients, the researchers reported that
search examples are provided. Table 6.2 four themes formed the essence of the ex-
presents a comparison of the three qual- perience of caring from the nurses per-
itative methods. spectives: knowing the patient, preserving
hope, easing the struggle, and providing
for privacy (Rittman et al., 1997). It is phe-
2. Phenomenological Research nomenologists role to uncover and con-
Perspective vey the true meaning or essence of the ex-
perience through the use of descriptive
The phenomenological perspective is a
language.
qualitative research method that de-
scribes the meaning of a lived experience
from the perspective of the participant. a. Philosophical Tenets
Lived experiences are the everyday
human experiences that are real to the in- Central tenets of phenomenology evolve
dividuals who experience them. Phenome- around the idea of social action. Phenom-
nology seeks to achieve a deep under- enology is predicated on the belief that
standing of the phenomenon being studied each person has a unique view of the
through a rigorous, systematic examina- world and each persons social reality is as
tion of it. Its purpose is to describe the valid and true as any other view (Munhall
essences of lived experiences. Essences are and Boyd, 1993). It focuses on the every-
Table 6.2 Comparison of Qualitative Methods

Method Focus Foundations Question
Phenomenology The meaning of the Philosophy What is essential for
lived experience social dancing in
caring for persons
with dementia in a
nursing home?
(Palo-Bengtsson &
Ekman, 1997)
Grounded theory To generate or discover Sociology symbolic What is the process
a theory about a interaction used by families to
basic social process manage the unpre-
dictability of heart
transplantation?
(Mishel & Murdaugh,
1987)
Ethnography Description and Cultural anthropology What are the mean-
interpretation of a ings of health to
cultural group or adolescents within
system their subculture?
(Rosenbaum & Carty,
1996)
day way in which people make sense of phenomenon. The phenomenological view
their being in the world. The philosophical enlarges the experience and tries to un-
basis for phenomenology lies in existen- derstand it in the complexity of its context.
tialism. Existentialism refers to the way Phenomenologists believe any experience
one views the world and stresses the per- is a valuable and valid source of knowl-
sonal here and now experience and the edge. They also believe that intuition (i.e.,
demands they place on the person as a developing ones consciousness through
free agent in a deterministic universe looking and listening) is important in
(Smith, 1997). Phenomenology is con- knowledge development (Smith, 1997).
cerned with the nature of being, the here
and now as it isthe lived experience.
Phenomenology has been impacted by b. Application of Method
philosophers such as Kant, Hegel, Lam-
For each perspective discussed, this sec-
bert, Husserl, and Heidegger (Schmitt,
tion addresses the following topics: de-
1972). Husserl (18591938), known as the
veloping the research question, the role
father of phenomenology, held unique be-
of researcher, selecting participants, data
liefs about what it means to be human. He
collection, data analysis, review of the lit-
introduced two ideas that are central to
erature, and reporting the ndings.
phenomenology: lifeworld and intersub-
jectivity. Lifeworld is the world of lived ex- ( i ) Developing the Research
perience, and intersubjectivity describes Question
how subjective awareness and under- The focus of a phenomenological investi-
standing can be reached in a common gation is to describe the meaning of the
world (Fain, 1999). Phenomenologists be- lived experience from the perspective of
lieve that whatever is known must appear the participants. Any experience that pre-
to consciousness; whatever does not ap- sents itself to consciousness can be a fo-
pear to consciousness cannot be known. cus of phenomenological investigation.
Consciousness then provides access to The nursing world, which confronts the
the world. To be conscious is to be aware experiences of human health and illness,
of the lived experience of some phenome- offers a rich source for phenomenologi-
non. Husserl believed consciousness and cal questions. Examples include being a
the lived experience were in relationship, patient, surviving breast cancer, receiv-
that is, consciousness was the means that ing a terminal diagnosis, giving birth, and
constituted the objects of experience. even experiencing laughter.
Everything that appears to ones con- The research question is usually quite
sciousness constitutes an object. This is broad. This is to allow the participants to
referred to as intentionality. Heidegger provide the answer. It usually takes the
(18891976) advanced the concept of be- form of What is the meaning of ones lived
ing in the world from the conception of experience? From this, the researcher
consciousness as intentionality. He be- may develop a central overarching ques-
lieved the nature of being human was tion and several subquestions that follow
concern or awareness of ones own being from the central question. Phenomenolog-
in the world. ical researchers avoid using hypothesis
All phenomenologists believe in multi- statements so as not to sway the partici-
ple realities that are constructed by indi- pants toward a desired answer.
viduals within the social context of their Suggested steps for formulating the
lives. Reality is constructed from human question have been developed by a num-
experiences that emerge from the relation- ber of nurse researchers (Munhall and
ship between human consciousness and Oiler, 1986; Omery, 1983; Parse et al.,
1985). The steps may vary depending on for data collection and therefore must be
the phenomenological perspective fol- an effective communicator and an empa-
lowed. Moustaka (1994, p. 99) suggests thetic listener. It is critical that the re-
four subquestions that may be useful in searcher establishes a good rapport with
explicating the meaning of the lived ex- the participants and listens attentively
perience. They focus on what is impor- with of all her or his senses to the partici-
tant about the experience to the participants stories. The meaning of the lived
pant and how the experience impacts on experience can only be revealed when the
what it means to be human: investigator understands and interprets
correctly the participants stories.
1. What are the possible structural mean- Researchers usually explicate their be-
ings of the experience? liefs about the research topic through the
2. What are the underlying themes and process of bracketing, which was previ-
context that account for the experi- ously discussed. This process is impor-
ence? tant in phenomenology in which the re-
3. What are the universal structures that searcher develops a close relationship
precipitate feelings and thoughts about with the participants and engages in pro-
the experience? longed dialogue. Through bracketing, the
4. What are the structural themes that fa- researcher identies personal biases,
cilitate a description of the experience? making it easier to attend to issues intro-
duced by the participants instead of lead-
Holldorrsdottir and Hamrin (1997) con-
ing the participant in the direction of is-
ducted a phenomenological study to ex-
sues deemed relevant to the researcher.
plore caring and uncaring encounters
An example of a researchers bracket-
with nurses and other health-care profes-
ing her perspective on the research phe-
sionals from the perspective of cancer pa-
nomenon is provided by Parse (1993) in
tients. The research question was, What
her phenomenological investigation of
is the essential structure of caring and un-
laughter. Parse, who was interested in un-
caring encounters with nurses and other
covering the structural denition of the
health professionals from the cancer pa-
experience of laughing, bracketed her
tients perspectives? Participants were
personal biases and perspective on laugh-
asked three questions that followed from
ing by articulating her views in writing.
the research question:
She explicitly describes her perspective
1. Can you tell me of your personal expe- on the lived experience of laughing. By do-
rience of a caring nurse or health-care ing so, she provides the reader with an
professional whom you have encoun- opportunity to assess how her perspec-
tered during your illness experience? tive may inuence the research ndings.
2. Describe what the nurse or health-care It is to be noted, however, that some re-
professional did and try to analyze searchers believe that it is impossible to
why you felt that he or she was caring. set aside ones beliefs and personal bi-
3. Could you describe how you felt dur- ases. This is particularly true of Her-
ing the encounter? meneutic phenomenology, in which the
researchers self-understandings must be
The same questions were posed for
identied, but they claim it is unrealistic
the uncaring experiences.
if not impossibleto set them aside.
( ii ) Role of the Researcher
The phenomenological researcher enters ( iii ) Sample
into the experience of the research par- A phenomenological investigation re-
ticipant. The researcher is the instrument quires a purposive sample composed of
individuals who have experienced the exist in the process of analyzing phenome-
phenomenon of interest. It is essential that nological data. Regardless of perspective,
participants are able and willing to talk data collection and analysis occur simulta-
about their experience of the phenomenon neously and the researcher immerses her-
and describe their inner feelings associ- or himself in the data. This requires the re-
ated with it. The sample size may vary, but searcher to dwell with the data by read-
usually a small number of participants are ing and rereading the transcripts in their
selected because of the extensive docu- entirety to get a sense of the lived experi-
mentation that results from the partici- ence as a whole. Depending on the tech-
pants narratives. One cannot easily pre- nique chosen, the data analysis process
dict the required sample size. It must be may require returning to participants to
large enough to produce a rich and com- get further clarication or more descrip-
prehensive description of the phenome- tions of the phenomenon. The researcher
non under investigation. This occurs when synthesizes the descriptions of partici-
the data are saturated; that is, when the in- pants and reduces the data into a smaller
clusion of new participants does not lead and smaller number of categories to arrive
to new information but rather conrms the at a consistent description of the meaning
previously collected data. Sample sizes of the lived experience for all participants.
typically range from 5 to 15 participants, The steps usually involve:
but this may vary considerably depending
on how quickly saturation is reached. 1. Reading and rereading the transcripts
in their entirety.
( iv) Data Collection
2. Identifying key statements of partici-
A variety of methods are used for data col-
pants that have a bearing on the phe-
lection, including in-depth interviews,
nomenon.
written descriptions of specic experi-
3. Identifying and naming the themes in
ences in diaries or journals, and observa-
the identied statements; it may be
tion. Interviews are usually tape recorded
helpful to write these in the margins of
and transcribed verbatim and last varying
each transcript.
lengths of time ranging from 1 hour to sev-
4. Identifying the essential structure of
eral hours. It is often necessary to con-
the phenomenon in the transcribed
duct several interviews with the same
data. This involves taking all of the
participants to ensure that the phenome-
themes and trying to identify the com-
non is fully described and reects the
monness or essential structure of the
meaning of the lived experience for the
phenomenon in each of the dialogues.
participants. Saturation of data helps de-
5. Identifying the essential structure of the
termine the number of interviews re-
phenomenon. This involves comparing
quired.
the different dialogues to nd similari-
(v) Data Analysis ties and differences in the dialogues to
A number of techniques are available for construct the overriding theme as well
analyzing data in phenomenological stud- as the essential structure of the phe-
ies. These include the methods of Colaizzi nomenon.
(1978), Georgi (1985), Patterson and Zde- 6. Comparing the essential structure with
rad (1976), Streubert (1991), Van Manen the data. After the essential structure
(1990), and Van Kaam (1959). For detailed of the phenomenon has been identi-
information on a specic method, re- ed, it is compared with the transcripts
search students are advised to consult the to make sure it ts the data. The re-
original source. Although each method searcher looks for themes in the tran-
presents unique aspects, commonalities scripts that are not accounted for in
the description. A decision must be away from the report with a better under-
made as to whether they should be in- standing of what it is like to experience the
cluded or excluded and why. phenomenon that was investigated.
7. Sharing the analytic description of the
phenomenon with the participants for
verication. Participants should be c. Example of Phenomenological
able to recognize their own experi- Research
ences in the description.
An account of the ndings from a phe-
This way of doing phenomenology has
nomenological study is provided in Box
been referred to as the Vancouver School
6.3. The following comments are offered
of Phenomenology. It is adapted from Co-
about the study in Box 6.3:
laizzis method (Fig. 6.2). Box 6.2 demon-
A phenomenological perspective is par-
strates Halldorsdottir and Hamrins (1997)
ticularly appropriate for this study that
adaptation of Colazzis framework to iden-
sought to gain an understanding of the
tify themes from cancer patients perspec-
process of developing a motherchild re-
tives of caring and uncaring encounters
lationship in the preadoptive period from
with nurses and health-care professionals.
the perspective of women who became
(vi ) Review of the Literature mothers through international adoption.
In phenomenology, the literature review Because this process is not well under-
is conducted after the data collection and stood and there is little investigation of it
analysis are complete. The purpose of the in the nursing literature, the phenomeno-
literature review is to place the study nd- logical perspective is particularly appro-
ings within the context of what is known priate for exploring peoples preadoptive
about the phenomenon. Parse (1993) experiences from their own perspective.
used the review of the literature to relate The women in this study recently lived the
her ndings on the lived experience of experience and are a rich source of data.
laughing to other investigators views of An open-ended interview was used
laughing. A cursory review may be con- that included a semistructured guide to
ducted to verify the need for the study; focus discussion on the motherchild re-
however, the concern with examining the lationship rather than the adoption
literature too early in the process is that process. A questionnaire or structured
it may bias the researcher in the investi- interview would have imposed the re-
gation of the phenomenon. searchers perspectives on the women
and would have been inappropriate. The
(vii ) Reporting the Findings early interviews were used to inform
Phenomenological ndings report the es- questions for the later interviews. The re-
sential structure or essence of the experi- spondents talked freely and asked im-
ence. The researcher describes the unify- promptu questions reecting what they
ing meaning of the experience that is thought was important rather than what
recognized by participants as their experi- the researcher thought was important.
ence. For example, Parse (1993) described Data were analyzed inductively using a
the experience of laughter as a buoyant method developed by Strauss and Corbin
immersion in the presence of unantici- (1990). Themes that reected the experi-
pated glimpsings prompting harmonious ence of the participants in the preadop-
integrity which surfaces anew through tive stage were identied. The literature
contemplative visioning. The reader of a was reviewed to compare the study nd-
phenomenological report should come ings with what was already known about

BOX 6.2 Theme Clusters That 3. Grounded Theory Perspective

Constitute the Essential
Structure of Caring and Grounded theory is an inductive, quali-
Uncaring Experiences tative research method that seeks to un-
derstand and explain human behavior.
Essential Structure of a Caring Encounter
The goal of this method is the develop-
Embodied in Three Themes ment of theory that explains underlying
1. The nurse or health-care professional is social and psychological processes. It is
perceived as caring and as an indispens- different from phenomenology in that the
able companion on the cancer trajec- researcher is not seeking to describe the
tory. phenomenon under investigation but
2. The result is mutual trust and caring con- rather to develop a theory about the
nection. dominant processes in the social scene
3. The perceived effect of the caring en- under investigation (Stern, 1980). In the
counter is a sense of solidarity, empow- social situation studied, the researcher is
erment, well being, and healing. interested in understanding how people
Essential Structure of an Uncaring interact, take action, or engage in the
Encounter Embodied in Three Themes process in response to the phenomenon.
1. The nurse or health-care professional is For example, in their grounded theory in-
perceived as uncaring and as an unfor- vestigation of family adjustment to heart
tunate hindrance to the perception of transplantation, Mishel and Murdaugh
well being and healing. (1987) studied the processes used by
2. The result is a sense of mistrust and dis-
family members of heart transplant re-
connection.
3. The perceived effect of the uncaring en-
cipients to manage the unpredictability
counter is a sense of uneasiness, dis- evoked by the need for and receipt of a
couragement, and a sense of being bro- heart transplantation.
ken down. Grounded theory is useful in areas in
which little research has been conducted
SOURCE: Adapted from Halldorsdottir, S., and or in gaining a new perspective in areas
Hamrin, E. (1997). Caring and uncaring encounters
within nursing and health care from the cancer pa-
that have an extensive research back-
tients perspective. Cancer Nursing, 20(2), 120121. ground. It is particularly appropriate in ex-
ploratory and descriptive studies (Glaser
and Strauss, 1967) and is now among the
most inuential and widely used modes of
qualitative research when the researchers
goal is generating theory (Strauss and
the process of motherchild relation- Corbin, 1997). It is making an important
ships. Womens narratives demonstrated contribution to knowledge development in
the strength, commitment, and resilience nursing because of its ability to produce
inherent in the journey to create a family. middle-range theory. Middle-range theory
Qualitative investigations such as this is the most useful to nursing because it can
one contribute to the eld of nursings be empirically tested (see Chapter 2). It is
knowledge base on the experience of the level of theory that ts between the
waiting for the adoptive child and can in- broad, grand theories and the narrow
crease nurses sensitivity to and under- practice or partial theories. Glaser and
standing of the needs of preadoptive Strauss (1967) made the bold claim that af-
mothers. The study can also educate and ter a grounded theory study has been con-
inform other women who are waiting to ducted in an area, little additional research
adopt. is done because of the density of the the-
ANTICIPATING THE ADOPTIVE CHILD: A PHENOMENOLOGICAL

STUDY
This phenomenological pilot study explored the to their babies a world away, similar to a preg-
experiences of three women who became moth- nant belly to rub and smooth.
ers through international adoptions. The study Honoring the childs origins and learning
sought to gain an understanding of the womens about their childs land and culture were ways of
experiences of developing a motherchild rela- paying tribute to the country that was giving
tionship in the preadoptive period. them their child. They saw the childs time in
Findings provided a description of seven their country of origin and with their caretakers
themes that describe the process of developing a as worthy beginnings.
motherchild relationship in the preadoptive pe- Investing personally went beyond money,
riod. The themes include taking control, creating time, and resources of the mothers. They in-
a family, celebrating the pictures, honoring the vested themselves and pursued adoption with
childs origins, investing personally, and bonding. blind determination. After they began, stopping
Taking control encompasses the mothers the pursuit to adopt seemed no longer an option.
process of beginning to develop a motherchild Bonding was the term used by mothers to de-
relationship. Mothers took control of their lives scribe what they did during the preadoption pe-
and their environments in order to make a riod. It was a significant theme running through-
place in the world for a family purposely cre- out their experience of developing a relationship
ated through adoption. All made conscious de- with their adopted child. One mother summa-
cisions not to pursue pregnancy, seeing adop- rized:
tion as a natural, proactive choice that put them The bonding starts with all the preparation you
in control of their reproductive health and de- make in your home, just psychologically, all the
sire to be parents. anticipation, having the pictures. What is that
Creating a family was the second theme that like? Maybe it is like parents who first feel their
emerged. All mothers stated clearly that their child kicking. The things that we did to bond,
goal was to create a family, be a parent, and many of them are the same as the couple who
have a child. One mother summed it up nicely: are pregnant. I mean you paint a room, you get
Its a feeling, like you have been on this road all a crib, you have a baby shower. Many of those
your life and everything that came before was things are the same, and then some of the other
bringing you to this point. These were the kids things are different.
meant to be in this house, in this home, to be our The narratives of the women describe themes
children. I have no doubt about it. This is the way in the process of preparing to welcome a child.
and the time they had to come to us. These were different from those that emerge
Anticipating, the third theme, was charged during pregnancy, yet there are parallels between
with emotion, preparation, and expectancy. the preadoptive process and pregnancy.
They prepared for their children by sharing their The stories of these women demonstrate the
anticipation and the adoption process with fam- strength, commitment, and resilience inherent in
ily, friends, acquaintances, and even strangers their journeys to create families. Future research
who were willing to listen. may focus on providing adoptive mothers with
Celebrating the pictures was a significant validation of both their journey and the pro-
component of developing the relationship. All cesses inherent in it, credibility for this alterna-
mothers described the pictures of their child as tive but natural path to motherhood, and recog-
something to hold onto, share with others, and nition of the beauty of the adoptive mothers
help them prepare for the arrival of their child. development of a deep, loving, lifelong mother
The photographs became tangible connections child relationship.
SOURCE: Summarized from Solchany, J.E. (1998). Anticipating the adopted child: Womens preadoptive experi-
ences. Canadian Journal of Nursing Research, 30(3), 123129.
ory that emerges. Because the generated tionism focuses on the nature of the dy-
theory emerges from the data, it is said to namic social interaction occurring among
be grounded in the data and useful to prac- persons and seeks to understand it.
tice.
b. Application of Method
a. Philosophical Tenets
( i ) Developing the Research
Grounded theory emerged from the disci- Question
pline of sociology and has its theoretical A grounded theory investigation is con-
foundations in symbolic interactionism, a ducted to identify concepts and proposi-
theory about human behavior and human tions about the relationships between
group life. Symbolic interaction focuses on them, as they relate to a basic social psy-
the meaning of events to people in every- chological phenomenon. Unlike most
day settings. It is this meaning created by methods, the researcher approaches the
the people in the situation that guides be- study without a specic well-dened ques-
havior, action, and the consequences of tion in mind. Instead, the question initially
that action (Chenitz and Swanson, 1986). is broad and general in nature. It serves to
At the core of symbolic interaction is an focus the study. It identies the phenome-
examination of how people dene events non to be investigated in a particular so-
or reality and how those beliefs shape cial context. The question emerges, is re-
their actions. George Herbert Mead (1934) ned, and may change several times as the
and Herbert Blumer (1969) advanced the data are collected and analyzed. The basic
symbolic interaction tradition. Mead de- social process is usually a gerund (a noun
scribed the process whereby a person ending in ing) indicating change across
develops a sense of self through social time as a social reality is realized (Lo-
interaction with others. Blumer believed Bionda-Wood, 1998). For example, Green-
symbolic interaction rested on three prem- berg (1998) in a grounded theory study of
ises: (1) humans react to things on the ba- humor within the nurseclient relation-
sis of the meanings that the things have for ship posed the question, What is the
them; (2) meanings are derived from the experience of humor in the nurseclient
social situations that one has with others; relationship from the perspectives of the
and (3) meanings are altered and shaped professional nurse and the client involved
through an interpretive approach used by in the nurseclient relationship? Thera-
individuals in dealing with things they en- peutic playing was the basic social process
counter (Blumer, cited in Chenitz and or core category that emerged. It was used
Swanson, 1986). by study participants to integrate hu-
For social life to emerge, group mean- mor within the relationship. The question
ings must be shared among the group. may not be explicitly stated in the re-
Communication provides the mechanism search study but can be implied from the
for such sharing. Meaning of the event studys purpose or goal statements. Other
emerges through both verbal and nonver- examples of questions appropriate for
bal behavior; therefore, the interaction grounded theory studies include (1) how
becomes the focus of the investigation to do family members who have survived an
discover the true meaning. Through in- adolescent suicide experience healing?
teraction, human beings are changed and (Kalischuk, 1998); (2) what is the process
inuence each other. Through interaction used by nurses to deal with the problem of
with others, humans create meaning being stymied in their person-related in-
about life and themselves. Social interac- teractions in the presence of technology?
(Alliex, 1998); or (3) What is the process scribe the theory that has emerged from
of living with bromyalgia syndrome for the process.
women? (Hughes, 1998).
( iii ) Sample
( ii ) Role of the Researcher The researcher selects participants who
In grounded theory research, the investi- are experiencing the social process under
gator is intimately involved with the in- investigation. For example, in their study
vestigation. The researcher aims to de- of caregiving to persons with cancer and
scribe social behavior as it takes place in AIDS, Stetz and Brown (1997) recruited 26
natural settings. This means the re- family caregivers of persons with cancer
searcher must not only study the behav- or AIDS from newspaper advertisements
ior but the social setting that inuences and announcements at home care agen-
the interaction. The researcher plays an cies and public health departments. The
integral role in the investigation as a par- sample size is not predetermined as in
ticipant in the world and an observer of statistical sampling used in quantitative
that world. Chenitz and Swanson (1986, research; rather, it is determined by theo-
p. 7) described the role of the researcher retical sampling. This is a process where-
as follows: by the researcher collects, codes, and an-
alyzes the data simultaneously and then
The researcher needs to be able to un- decides what additional data are needed
derstand behavior as the participants to enhance the theory. Similar to phe-
understand it, learn about their world, nomenology, data collection continues
learn their interpretation of self in the until saturation is reached. In this situa-
interaction and share their denitions. tion, saturation refers to the inability of
In order to accomplish this the re- new data to add new codes or expand ex-
searcher must take the role of other isting codes. At this point, the researcher
and understand the world from the par- concludes that the data are saturated and
ticipants perspective. The researcher, the theory has emerged. As a rough esti-
therefore, must be both a participant in mate, saturation typically requires about
the world and an observer of partici- 20 to 30 interviews and several visits to
pants in the world. Finally, in order for the eld, although it varies depending on
the knowledge to be understood and ac- the quality of the data that emerges and
cepted by the researchers discipline, the phenomenon and social process be-
the researcher, as observer, must trans- ing investigated.
late the meaning derived from the re-
searcher as participant into the lan- ( iv) Data Collection
guage of the research discipline. Astute observation of participants in the
social group and skilled interviewing are
The researcher needs to be able to col- the primary sources of data collection
lect and analyze data, recognize and avoid available to the grounded theorist. Inter-
bias, and think abstractly. The investiga- views are usually audiotaped and tran-
tor must set aside his or her own per- scribed verbatim and eld notes of ob-
spectives and theoretical notions so that servations are transcribed as soon as
the theory can emerge from the data. This possible on return from the eld. Docu-
is critical to the method; the theory must ments such as journals and diaries may
be grounded in the data. In addition, the also be used, but they are not typical of
researcher needs to possess keen inter- grounded theory. The researcher partici-
personal and observation skills, powerful pates in the social group and observes
analytical abilities, and the art of accu- and records information relevant to the
rately crafting a good narrative to de- study purpose. Interviewing may be in-
formal or formal using a semistructured pared with other pieces for similarities
interview guide or topical outline. Such and differences in the lives of those inter-
guides are not rigidly adhered to rather viewed. Figure 6.3 illustrates the uid
the questioning emerges from the data and cyclical nature of data analysis in
and becomes more focused as the study grounded theory research. It requires de-
progresses. Questions are broad, open- tailed record keeping of interviews, eld
ended, and general in nature. For exam- notes, and memos. It follows a standard
ple, Stetz and Brown (1997) used the format that has been described by a num-
Caregiver Outreach Interview Guide to ber of experts in the eld (Beck, 1993;
collect qualitative data on the caregiving Creswell, 1998; Glaser, 1978; Stern, 1980;
experience. The guide contained open- Strauss and Corbin, 1990; Streubert and
ended questions about the overall expe- Carpenter, 1995; Chinitz and Swanson,
rience, as well as specic questions on 1986). The steps are briey summarized
the caregivers perceptions of their role, below using the process outlined by Stern
and the demands of that role on their (1980) and Streubert and Carpenter (1995,
lives. Sample questions included: pp. 155161).
1. Tell me about your family members ill- 1. Concept formation: coding. The re-
ness. searcher reads the transcripts and
2. How was the decision made that you looks for an underlying pattern in the
would be a caregiver? data. Coding occurs at three levels.
3. Sometimes persons caring for their fam- Level I coding involves studying the
ily member encounter situations that data line by line and identifying key
are a concern for them. What things processes in the data. Level I codes are
about this experience are of concern for called substantive codes because they
you? codify the substance of the data and use
4. How has caring for a family member the words of participants (Stern, 1980).
changed your life? Level II codes assign data to cate-
gories according to obvious t. A cate-
Researchers need to be careful that
gory is a unit of information composed
the questions they pose do not limit them
of events, happenings, and instances
from hearing what the participants have
(Strauss and Corbin, 1990). They
to say or restrict the participants from
emerge from condensing level I codes
having their own perspectives. One use-
by comparing each level 1 code with
ful technique is to ask participants to
all other level I codes. Each category is
share stories of their experiences in life.
then compared with every other cate-
This can provide rich detail and add
gory to ensure that they are mutually
depth and texture to the data (Carla
exclusive.
Mueller, personal communication, 2000).
Level III codes identify the core vari-
(v) Data Analysis able or the basic social psychological
A unique feature of grounded theory re- process (BSP). A core variable is one
search is that data collection, coding, and that focuses the theory and accounts
analysis occur simultaneously from the for most of the variation in a pattern of
beginning of the study. The process is behavior that is both relevant and prob-
systematic but uid, allowing the re- lematic for the participants involved. It
searcher to respond to new information is central to the other categories (Beck,
that emerges from the data. The constant 1993). BSPs usually represent the title
comparative method is used when each given to the themes that emerge from
piece of information is coded and com- the data. They are processes that occur
Figure 6.3 Grounded theory: Data generation and analysis.

over time and involve changes over see if there is a higher order or umbrella
time. For example, Redesigning the category under which a number of the
Dream was the BSP that emerged from existing categories can be merged. It is
Mishel and Murdaughs (1987) study of similar to clustering items in factor
family adjustment to heart transplanta- analysis (see Chapter 18). The re-
tion. Table 6.3 illustrates examples of searcher identies theoretical links
BSPs discovered through grounded among the categories and collapses
theory research. them to form more general categories.
2. Concept development. Three steps are For example, in their study of family
involved in the development of the core adjustment to heart transplantation,
variable: reduction, selective sampling Mishel and Murdaugh perceived the
of the literature, and selective sampling three categories of freeing self, symbio-
of data. Reduction is a vital step in iden- sis, and trading places all mean to pro-
tifying the core variable or BSP. In remote the welfare and comfort of the pa-
duction, each category is compared to tient while they waited for a heart.
Table 6.3 Nurse Researchers at Work Selected Examples of Grounded Theory Studies
Basic Social Psychological
Title Process Stages
The Invisible Burden: Bearing the invisible burden 1. Dealing with traumatic life
Womens Lives with 2. Yearning for afrmation
Fibromyalgia 3. Searching for relief
4. Adjusting to limitations
HIV-Positive Women: Double Defensive mothering 1. Preventing
Binds and Defensive 2. Predicting
Mothering 3. Protecting
Transforming: The Process of Transforming 1. Battling
Recovery from Schizophrenia 2. Connecting
3. Determining
4. Committing
Handling the Hurt: Women Who Handling the hurt 1. Numbing out
Use Cocaine and Heroin and 2. Giving up
Lose Child Custody 3. Running away
4. Cleaning up
5. Dealing with feelings
Inner Strength in Women: Learning to live a new life 1. Dwelling in a different place
A Grounded Theory 2. Healing in the present
3. Giving and allowing
nurturance
4. Connecting with the future
5. Being spiritual
Integration: The Experience of Integration 1. Having diabetes
Living with Insulin Dependent 2. The turning point
(Type 1) Diabetes Mellitus 3. The science of one
Loss and Bereavement: Personal work 1. Reconciling
HIV/AIDS Family Caregiving 2. Making life-and-death
Experiences decisions
3. Letting go
Thus, it was possible to reduce these retical codes to sensitize researchers to

three categories to form the broader the subtleties of their data and the pos-
category of immersion (a process sibilities for integration of data. Three
whereby family members pledge them- examples of Glasers coding families
selves to the welfare of the patient). (and their constituent members) in-
Selective sampling of the literature clude (1) the consequences family (i.e.,
occurs simultaneously with, or fol- the six Cs: causes, contexts, contingen-
lows, data analysis. The literature re- cies, consequences, covariances, and
view is conducted to determine what conditions); (2) the interactive family
is known about the concepts that are (i.e., mutual effects, reciprocity, mutual
emerging from the data. The literature trajectory, mutual interdependence, in-
is considered data and used to ll in teraction of effects, and covariance);
the gaps in the emerging theory and and (3) the strategy family (i.e., strate-
add completeness and clarity to the gies, tactics, mechanisms, manipula-
theoretical description. tions, maneuvering, dealing with han-
Selective sampling of the data occurs dling techniques, ploys, means, and
after the main concepts have emerged. goals).
At this point, additional data are col-
Stetz and Brown (1997) used the above
lected in a selective manner to develop
three coding families to connect the cate-
the hypothesis statements further, iden-
gories and subcategories of taking care
tify the properties of the main variables,
that emerged from the data. Box 6.4 illus-
and ensure saturation of the categories.
trates the strategies, consequences, and
The researcher generates hypotheses
interactions that constitute the process
about categories and their relationships
of taking care.
and interrelationships and then tests
Theoretical memos are the ideas the re-
these hypotheses with selective data
searcher holds about codes and the rela-
that either support or fail to support the
tionships as they strike the researcher dur-
hypotheses. Researchers stop collect-
ing analysis. Memos vary in length from
ing new data when they are satised
one line to several pages. Their purpose is
that they are not hearing anything new
to help the researcher put the fractured
about the category or the emerging hy-
data back together. Glaser (1978) identies
pothesis. At this point, one core category
three roles for memos: (1) to raise data to
variable emerges that can explain the
a conceptual level, (2) to develop the prop-
relationship between all of the others.
erties of each category, and (3) to generate
This core variable best explains how a
hypotheses about relationships between
problem is processed (Chenitz and
the categories. Memos are sorted into a
Swanson, 1986).
theoretical outline to assist the researcher
3. Concept modication and integration.
in writing up the grounded theory that was
This phase of data analysis involves two
discovered. Glasers rule of sorting memos
processes: theoretical coding and mem-
is to begin sorting categories and their
oing. Theoretical coding forms theoret-
properties only as they relate to the core
ical linkages or connections between
category or BSP. Memos that do not t are
the data categories. Whereas the sub-
saved until such time as a new focus of the
stantive coding that occurred at level I
study is considered. Sorted memos be-
fractured the data into pieces, theoreti-
come the basis of the research report.
cal coding helps weave the fractured
pieces back together again, by intro- (v) Review of the Literature
ducing level II code (Beck, 1993). Glaser Grounded theory research contrasts with
(1978) has identied 18 families of theo- quantitative investigations in that there is
no review of the literature in the substan- are provided to show how the theory is
tive area of the study before data collec- embedded in the data. A review of the lit-
tion. Indeed, the dictum in grounded the- erature and numerical results are not in-
ory research as stated by Glaser, cited in cluded in a grounded theory report.
Beck (1993, p. 7) is: There is a need not to
review the literature in the substantive
area under study. The rationale for this c. Example of Grounded Theory
dictum is to avoid biasing the researchers Research
efforts to create concepts and hypotheses
An account of a grounded theory study is
from the data that truly t the data. The re-
provided in Box 6.5. The following com-
view of the literature is reserved until the
ments are offered about the study in Box
theory begins to emerge from the study
6.5:
data. At this point, the researcher inte-
A grounded theory perspective is par-
grates the literature with the emergent
ticularly appropriate for this study, which
theory during saturation, sorting memos,
sought to develop a theory of the process
and the writing of the nal report.
engaged in by families to manage the un-
(vi ) Reporting the Findings predictability of cardiac transplantation.
The ndings of grounded theory research Redesigning the Dream is the core
include a substantive theory that de- variable or basic social psychological
scribes the phenomenon under investiga- process that accounts for the pattern of
tion supported by study data. Rich narra- behavior among participants as they ad-
tives of the study process and ndings just to the reality of the treatment envi-
A GROUNDED THEORY STUDY: FAMILY ADJUSTMENT TO

HEART TRANSPLANTATION
The processes family members of heart trans- constant comparative method. Data collection
plant recipients use to mange the unpredictabil- and analysis continued until theoretical satura-
ity evoked by the need for and receipt of a heart tion was achieved.
transplant were explored in a grounded theory Redesigning the Dream was the main
study. The sample consisted of 20 family mem- theme emerging from the data that explained
bers of patients who received heart transplants. family adjustment to heart transplantation.
The subjects were labeled as partners to desig- Through this process, family members learned to
nate the significant other relationship to the pa- readjust their cognitive thinking and behaviors in
tient. The partners were participants in three a manner that enabled them to cope with the sit-
separate support groups, each of 12 weeks du- uation of heart transplantation in their loved one.
ration. The groups consisted of 14 wives, five Their initial dream that life would return to nor-
mothers, and one sister. mal after the patient received the transplant had
Data were collected from the three support to be revised to reflect the reality of the med-
groups over a period of 2.5 years. The investi- icaltechnological treatment environment. The
gators collected and audio taped all data from theory that emerged contained three major
the 12 weekly group sessions that lasted ap- concepts: immersion, passage, and negotiation.
proximately 1.5 hours each week. The investi- These concepts parallel the stages of waiting for
gators recorded content discussed as well as a doctor, hospitalization, and recovery. This
the participants reactions, group dynamics, and study makes an important contribution to nurs-
the investigators thoughts about the processes ing in the area of middle-range theory related to
that occurred during the group meetings. A to- the psychosocial needs of patients and families
tal of 36 transcripts were analyzed using the awaiting cardiac transplantation.
SOURCE: Summarized from Mishel, M.H., and Murdaugh, C.L. (1986). Family adjustment to heart transplantation:
Redesigning the dream. Nursing Research, 36(6), 332338.
ronment. This core variable focuses the 4. Ethnographic Perspective

development of the theory in the study.
The theory contains three main con- Ethnography is a qualitative research
cepts: immersion, passage, and negotia- method that attempts to understand hu-
tion and a number of subcategories. man behavior in the cultural context in
Data were collected over a 2.5-year which it is embedded. It provides a de-
period and analyzed using the constant scription and interpretation of a cultural
comparative method previously de- group that is gleaned from the re-
scribed. This study shows the promise searchers prolonged examination of the
that grounded theory yields for accu- observable and learned patterns of behav-
rately depicting the participants real ior, language, customs, interactions, and
world and the processes that partici- ways of life of the culture-sharing group
pants use to deal with problems in that (Creswell, 1998). Its aim is to understand
world. Theories such as Redesigning the the way in which people live from the emic
Dream, which are grounded in data, hold (insider or natives) point of view, and how
promise for a deeper understanding of they derive meaning from their experi-
nursing knowledge. ences (Spradley, 1980). The emic view is in
contrast to the etic view, which is the view the systematic documentation, descrip-
of a researcher or an outsider. tion, explanation, and interpretation of the
Ethnography is the work of describing a comprehensive meanings of both obscure
culture (Spradley, 1980). Leininger (1970), and obvious nursing phenomena related
a nurse anthropologist, denes culture as to care, health, prevention of illness, and
a way of life belonging to a designated recovery from illness or injury. Ethnogra-
group of people. It includes all the ways a phy is a viable research method for nurses
group of people solve problems, as re- because it provides a means to study cul-
ected in their language, dress, food, tra- tural variations in health and illness and
ditions, and customs. It is a pattern of liv- nurses and their clients as subcultures of
ing that guides the groups thoughts, society.
actions, and sentiments. The purpose of Two types of ethnography are useful
ethnographic research is to make explicit for answering nursing questions: mini and
what is implicit within a culture by study- maxi ethnographies. A mini ethnography
ing various cultural characteristics. The is dened as a small-scale ethnography fo-
ethnographers role is to paint a portrait of cused on a narrow area of inquiry. An ex-
the culture-sharing group. ample is the study by Holroyd and associ-
Within anthropology, a number of eth- ates (1997) that explored the cultural
nographic interpretations have emerged practices of women in Hong Kong during
over the years, including holistic, semiotic, the postpartum period (see Box 6.7). In
behavioristic, cognitive, and ethnonursing this study, the researchers interviewed
perspectives. The proliferation of schools seven women about their self-care prac-
of ethnography has led to a lack of ortho- tices within the family home during the
doxy in method; therefore, it is important month after the birth of their rst child.
for investigators to identify which school The number of participants in the study is
they espouse when discussing their meth- small and the focus is limited to postpar-
ods. Holistic ethnography, which denes tum practices in the rst month after de-
culture as a way of life and studies it as an livery. The small sample size and the lim-
integrated whole (Fain, 1999), and eth- ited focus on one aspect of cultural
nonursing are the two perspectives most practices over a 1-month period qualify
frequently found in nursing investigations. this study as a mini ethnography. This is
A discussion of all the subtypes of ethnog- in contrast to a maxi ethnography, which
raphy is beyond the scope of this chapter; is dened as a large and comprehensive
however, ethnonursing requires some study of general and particular features of
comment because of its important contri- a designated culture (Leininger, 1985, p.
bution to nursing knowledge and its popu- 35). Most maxi ethnographies have a
larity among nurse researchers. broad focus of inquiry, extend over a long
The ethnonursing tradition developed time period (e.g., several years), and are
by Leininger (1978, 1985, p. 38) is dened published in book form. Mini ethnogra-
as the study and analysis of the local or in- phies are well suited to investigations by
digenous peoples viewpoints, beliefs, and nursing students who wish to study cul-
practices about nursing care phenomena tural aspects of health situations but do
and processes of designated cultures. not have the time to engage in extensive
The goal of ethnonursing is to discover eld work required in maxi ethnographic
new nursing knowledge as perceived or ex- studies. Mini ethnography enables the
perienced by nurses and consumers of student to experience the challenges and
nursing and health services (Leininger, richness of doing work in the eld within
1985, p. 35). The focus in ethnonursing is a reasonable time frame.
a. Philosophical Tenets studied. This is best expressed by

Spradley (1980, p. 5), who states:
Ethnography has its roots in cultural an-
The essential core of ethnography is
thropologythe study of the origin of
this concern with the meaning of ac-
people, their past ways of living, and their
tions and events to the people we seek
strategies for surviving through time. The
to understand. Some of these meanings
philosophical underpinnings of anthro-
are directly expressed in language;
pology are still evolving, but there is con-
many are taken for granted and com-
sensus that culture is the central concept
municated only indirectly through word
of interest and ethnography remains cen-
and action. But in every society people
tral to the study of culture within anthro-
make constant use of these complex
pology. Anthropologists seek to discover
meaning systems to organize their be-
the interrelationships between all the
havior, to understand themselves and
parts that make up a culture so that a pic-
others, and to make sense out of the
ture of the wholeness of the culture
world in which they live. These systems
emerges. These parts include the mater-
of meaning constitute their culture;
ial culture consisting of manmade objects
ethnography always implies a theory of
associated with a given group; the ideas,
culture.
beliefs, and knowledge expressed in lan-
guage; social networks; social and politi-
cal institutions; and the ideals that the b. Application of Method
group holds as desirable goals (Burns
Implementation of the ethnographic
and Grove, 1997).
method involves three phases of re-
Ethnography involves learning from
search activity: preeld work, eldwork,
people rather than studying people. This
and posteld work (Fain, 1999). Within
important distinction means that the re-
each stage, specic steps must be fol-
searcher must become a student and the
lowed. These have been described by au-
participants of the culture become the
thors such as Atkinson and Hammmers-
teachers. By observing and studying
ley (1994), Leininger (1985), and Spradley
what people do (cultural behavior), what
(1980) and are listed in Box 6.6.
people say and know (cultural knowl-
edge), and the things people use and ( i ) Developing the Research
make (cultural artifacts), the ethnogra- Question
pher comes to learn about the culture by Formulating the research question re-
making cultural inferences (Spradley, quires that decisions have already been
1980). The ethnographer attempts to dis- made about the phenomenon to be inves-
cover the insiders view (emic) of the tigated, the scope of the study, and the site
world. To do this, the ethnographer must and cultural group. The phenomenon of
set aside the idea of nave realismthe concern usually focuses on some pattern
belief that all people dene the real world of behavior, lifestyle, custom, or belief of
in a similar way. Ethnography requires the culture-sharing group. The researcher
the researcher to adopt an attitude of selects a group of people to study and a
conscious ignorance about the culture site where an intact group has developed
under investigation (Spradley, 1980). shared values, beliefs, and assumptions.
In seeking to describe a culture at a Table 6.4 highlights phenomena nurse re-
particular moment in time, ethnogra- searchers have chosen to study using
phers strive to learn the meanings of life ethnographic methods. In addition to read-
experiences and events to the people ing Table 6.4, research students who are
POSTPARTUM PRACTICES IN CHINESE WOMEN

An ethnographic study was conducted to deter- 2. Dirt and prohibitions. Bathing and hair wash-
mine the cultural practices of Chinese women in ing were not permitted because they would
Hong Kong during the postpartum period. Seven open the pores and allow the wind in, lead-
multiparous women were interviewed and asked ing to harmful effects in later life. Preventa-
to reflect on their self-care practices within the tive rituals were thought to restore youth
family home during the month after the birth of and insulate against the crushing effects of
their first child. The interviews were translated old age.
from Chinese to English and then transcribed. 3. Rest and appeasing the placenta god. Bed
Nonverbal behaviors observed during the inter- rest was essential after delivery to appease
views were described in field notes immediately the placenta god.
after the interviews. The data were coded and 4. Housework: variations and variety. Women
then analyzed using content analysis, constant made modifications in doing housework. In
comparative techniques, and a code-mapping many situations, family members (e.g., mother
category described by Morse (1991). Categories and father-in-law) did this work. These acts
and themes were identified that described the demonstrated care and respect by family
daily enactment of the postpartum practices of members.
the Chinese women and to what extent other 5. Poisonous sexual intercourse. Sexual inter-
people influenced these. course was delayed for at least 1 month af-
Results indicated that this group of Chinese ter childbirth. It was implied that intercourse
women demonstrated customary beliefs and was unclean and had the potential to pollute
behaviors. Most women perceived that follow- the mother. In general, women resumed
ing traditional Chinese practice would speed sexual activity 3 months after delivery, more
their recovery after childbirth. They believed aptly described by the Chinese lunar calen-
that if the cultural rules were violated, ill effects dar as 100 days.
would result in their older age as a result of an 6. Competing loyalties. Some women expressed
imbalance of the Ying and Yang in their body. In not really wanting to follow traditional cultural
addition, significant others acted as potent re- practices but did so because of family pres-
inforcers of cultural mores, often stirring salient sure. Mothers-in-law were seen as potent re-
associations with past memories. inforcers of tradition and the act of checking
According to the Chinese belief system, the on their daughter-in-law was present in nearly
postpartum period lasts for 30 days after deliv- every dialogue describing a customary prac-
ery, during which time the mother must per- tice. The older people still reinforce beliefs
form a variety of avoidance rituals and remain at that they once practiced themselves, yet the
home. This is called doing the month. These reality of modern-day Hong Kong and
specific practices are regarded as not only cura- womens changing perception of their place in
tive but also preventative. To a new mother, the society pose conflicting demands on the young
degree of importance attached to these customs women.
serves to indicate her importance in her hus-
This study highlights the importance of cul-
bands household, with increased significance if
turally sensitive and congruent nursing practice.
the woman has produced a son.
It suggests that nurses have unique opportuni-
Analysis of categories and themes revealed
ties to support and provide culturally sensitive
five major postpartum practice areas that Chi-
interventions such as hotcold dietary guide-
nese women adhere to:
lines, hygiene suggestions, and activity and
1. Good food and bad blood. Food was used as leisure recommendations that can help women
a tonic to restore the balance of the good and their families restore their optimal health
blood in the body and was also seen as aid- without violating cultural norms, standards, or
ing in the expulsion of bad blood. beliefs.
SOURCE: Summarized from Holroyd, E., Katie, F., Chun, L., and Ha, S. (1997). Doing the month: An exploration
of postpartum practices in Chinese women. Health Care for Women International, 18, 301313.
Table 6.4 Phenomena Investigated in Ethnographic Studies

Culture-Sharing Group Site Phenomenon Researchers
Women in western Western Africa and a Practice, beliefs, and Morris (1996)
Kenya and Somali Somali section of attitudes toward
refugees in the San Diego female circumcision
United States
Health professionals Telemedicine sites in Diffusion of telemedi- Ibbotson (1999)
with the National Scotland cine across Scotland
Health Service of
Scotland
Adolescents Juvenile detention Decision making about Anderson (1996)
facility substance abuse
Adolescent peer Local community Meanings and experi- Rosenbaum and Carty
subculture youth center ences of care and (1996)
health within peer
subculture
Hutterites Dariusleut and Concepts of health and Brundt, et al. (1997)
Lehrerleut groups in empowerment in
Alberta, Canada health promotion
composing ethnographic questions may Spradley (1980, p. 31) suggests that an

nd it helpful to consider the ve areas ethnography is usually conducted with a
identied by Fain (1999) as appropriate for single general problem in mind: to dis-
ethnographic investigation: cover the cultural knowledge people are
using to organize their behavior and inter-
1. Description of a cultural group that pret their experience. However, many
has not yet been described. (e.g., a ethnographers design their studies to ad-
rst-generation group of immigrants to dress more limited problems. Nursing in-
a new country). This would entail a vestigators tend to focus on one or more
maxi ethnographic study that would aspects of life in a community that impact
be long term and involve complex ob- the health experience of its members.
servations and interviews with multi- Nurses are interested in asking questions
ple informants. about how ones cultural knowledge, val-
2. An aspect of culture that has not yet ues, norms, and standards inuence ones
been described among a given group health experience (LoBionda-Wood, 1998).
of people (e.g., Native American chil- Questions may be asked about health and
drens beliefs about death and dying). illness beliefs, self-care practices, lifestyle
3. A theoretical problem that requires in- decisions, life events, health values, and at-
vestigation (e.g., health beliefs in con- titudes as ones cultural group impacts
ict with cultural practices). them. Often the question in an ethno-
4. A practical problem needs some solu- graphic report is implied in the purpose
tions identied (e.g., high rates of car- statement of the study rather than explic-
diovascular disease among Hutterite itly stated. For example, Holyrod and asso-
community members). ciates (1997) crafted their study to answer
5. A society previously studied needs the question: What are the self-care prac-
to be reexamined to see the changes tices of Hong Kong Chinese women within
that have occurred (e.g., Hong Kongs the family home during the rst month af-
change of sovereignty). ter the birth of their rst child? This ques-
tion is implied from the purpose statement non of interest. Informant rather than sub-
of the study. ject is the term used to refer to partici-
pants in an ethnographic investigation.
( ii ) Role of the Researcher
The term informant implies that the per-
The role of ethnographic researchers is
son has special attributes such as knowl-
to be interpreters of the experience of
edge or experience that are important to
others. Researchers enter that world for
the investigation. Creswell (1998) sug-
an extended period of time, asking ques-
gests it is best to select a group of infor-
tions, observing, participating in the re-
mants that the researcher can gain access
spondents experiences, and collecting
to and one to which the researcher is a
whatever data are available about the
stranger. Unless a compelling case can be
study question. Ethnographers observe
made to study people in ones own back-
behavior but go beyond it to inquire
yard, it should be avoided because of the
about the meaning of that behavior. Simi-
difculties that can arise. Studying those
larly, ethnographers see artifacts and
well known to the ethnographer can cre-
record emotional states but go beyond
ate expectations for data collection that
these to discover the meanings people at-
may compromise the value of the data
tribute to these objects and feelings. It is
and cause individuals to withhold infor-
the researchers role to make inferences
mation or slant it in a way to please the in-
from their observations and then to test
vestigator (Glesne and Peshkin, cited in
these inferences over and over again with
Creswell, 1998). Students are often at-
their population of interest until they are
tracted to do ethnographic investigations
condent that they have an adequate de-
of this type because of the easy access to
scription of the culture (Spradley, 1980).
participants. It is a procedure, however,
Ethnographers attempt to learn, inter-
that should be discouraged because of
pret, and transcribe perspectives of indi-
the difculties inherent in making known
viduals and groups of people as they go
to others what is only implicitly under-
about the business of living their lives
stood by the investigator.
and making sense out of their world
Access to the cultural group is usually
(Aamodt, 1991). As the primary data col-
gained through a gatekeeper, an individ-
lection tool, the researchers are integral
ual with special group status who can lead
to the study.
the researcher to other key informants.
To understand the study phenomenon
Key informants are the group members
from the participants perspective (emic
who are most knowledgeable about the
view), the ethnographer must explicate
study phenomenon. They are representa-
his or her own beliefs and set aside or
tive of the culture and have potential to
bracket personal biases. For example, in
yield substantive data about the study
their study of the health-care beliefs in the
phenomenon. Key informants are willing
subculture of adolescence, Rosenbaum
to teach the ethnographer about the study
and Carty identied Leiningers (1991)
phenomenon and refer him or her to gen-
theory of culture care diversity and uni-
eral informants who also have knowledge
versality as the framework that guided
of the phenomenon under study. Leininger
their study. This perspective denes care
(1985) suggests that depending on the
in relation to assisting, supporting, or en-
study scope and purpose and the phe-
abling experiences that lead to an im-
nomenon under investigation, about 10 to
proved human condition or lifestyle.
15 key informants and 30 to 60 general in-
( iii ) Sample formants constitute a desirable number of
The investigator selects a cultural group informants for ethnonursing or ethno-
that has experience with the phenome- graphic studies. In Rosenbaum and Cartys
study of the subculture of adolescence, 27 data on the content of teenagers deci-

key informants were interviewed. They sions; the meaning of substance abuse in
were effective key informants because their lives; and the interactional dynamics
they were knowledgeable about their own among young women in juvenile detention
subculture and they could teach the inves- as they discussed their resolution for the
tigators about their meanings and experi- future related to substance abuse.
ences with health and care. In addition, 44 In this example, the critical rst step
general informants were interviewed to re- was to establish a working relationship
ect on the meanings expressed by the key with the teens and staff. This required the
informants. investigator to learn how things worked in
the closed setting of a juvenile detention
( iv) Data Collection facility (JDF). Daily routines, such as meal
Participant observation, formal and in- and bed times, attendance at classes, and
formal interviews, and collection of arti- events such as Sunday morning church
facts and documents are all means of services and Thursday evening co-ed
gathering data in ethnographic research. dances, all provided opportunities to learn
Of these methods, participant observa- the folkways at JDF. As Anderson com-
tion and interviews are the most popular mented, Participant observation during
forms. Through the process of eldwork, these activities helped me to learn what it
the researcher spends prolonged periods means to live and work behind locked
of time in the natural settings of the par- gates.
ticipants and observes members of the Data collection began with focus groups
culture in their everyday lives. Field designed to obtain perceptions on a de-
notes are written and provide descrip- ned area of interest in a permissive, non-
tions of the researchers observations threatening environment in which the
and experiences in the eld. They be- teens had an opportunity to present their
come valuable sources of data as the own perspective about substance abuse
study unfolds. The steps for conducting a decisions. Each focus group session lasted
participant observation study, writing 1 to 1.5 hours. One-hour participant obser-
eld notes, and doing in-depth interviews vation sessions were carried out on the
are described in detail in Chapter 7. dormitory units where the informants
During the eldwork phase, investiga- lived both before and after the focus
tors may decide that additional informa- groups. Observations were recorded in
tion is required about what they are ob- eld notes along with notations regarding
serving in the eld. For example, collection nonverbal behaviors during the focus
of basic demographic data may be accom- group discussions. During the focus
plished through census taking; mapping groups, talk about decisions regarding
may be used to identify the location of peo- past and future substance use became a
ple and understand how they interact with dynamic process. The girls made resolu-
their physical environment; genealogies tions and plans, offering them to each
may be collected to study the kin relation- other, and in their interactions they
ships of informants; and life histories may shaped and reshaped their decisions. Gen-
be collected to show how general cultural eral questions were used such as, How do
patterns are integrated into a persons life teenagers usually solve their problems?
(Fain, 1999). and How do teenagers decide to use
Anderson (1996) provides a useful ex- drugs or alcohol? Questions that re-
ample of ethnographic data collection. She peated some of the discussion during the
used participant observation, in-depth in- focus groups were asked in individual in-
terviews, and small focus groups to collect terviews to discover the individual partici-
pants perspectives. Focus groups and in- 1. Domain analysis. In the domain analy-
terviews were audiotaped and transcribed sis, the researcher is moving from ob-
verbatim by a linguistic anthropologist serving a social situation (a set of be-
who specialized in the slang of inner city haviors carried out by people in a social
teenagers. Three months were required to situation) to discovering a cultural
collect the data. scene. A cultural scene is an ethno-
graphic term used to refer to the culture
(v) Data Analysis under study (Spradley, 1980). Cultural
Analysis of data in ethnographic inquiry domains are categories of meaning that
follows a cyclical pattern whereby the re- include other smaller categories. The
searcher goes back and forth collecting rst step in a domain analysis is to se-
ethnographic data, recording the data, lect a situation to observe. For example,
analyzing the data, and returning to the the researcher may observe a pediatric
eld to collect more data. It is a process unit. The category, people on the unit, is
of questiondiscovery (Spradley, 1980). the rst domain to be analyzed. The re-
Unlike quantitative researchers, who go searcher asks the question, Who are
into the eld with specic questions in the people on the unit? She refers to
mind, ethnographers analyze the eld her eld notes and easily identies the
data compiled from participant observa- various categories of people (e.g., chil-
tion and interview to discover more ques- dren, parents, family members, doctors,
tions that bring the researcher back to nurses, ward clerks, cleaning staff ).
the eld for more data collection, more Spradley suggests that it is important to
eld notes, and more analysis, and the cy- identify the semantic relationships in the
cle continues. observations made in your particular
After each session in the eld, the re- cultural scene. He identies a list of uni-
searcher analyzes the data in order to versal semantic relationships. For ex-
know what to look for in the next period ample, x is a kind of y; x is the result of
of participant observation. The analysis y; x is a part of y. Children are a kind of
proceeds through four levels as the re- patient on the unit. Furthermore, an-
searcher learns, describes, and inter- other analysis can be done to explore
prets the meaning of cultural symbols on the types of children admitted to the pe-
the informants language. The experi- diatric unit. By creating the domain
enced researcher conducts the four lev- analysis, the researcher identies addi-
els of analysis simultaneously through- tional questions and makes more fo-
out the research project; however, it is cused observations that lead to explor-
recommended that the research student ing the roles and relationships of other
learn to do each in sequence before mov- members in the cultural scene.
ing on to higher levels. The suggested se- 2. Taxonomic analysis. A taxonomic an-
quence includes domain analysis, taxo- alysis is a more in-depth analysis of the
nomic analysis, componential analysis, domains. The researcher is searching
and theme analysis (Spradley, 1980). for larger categories to which the do-
Throughout the analysis, the researcher main may belong. A taxonomy is a set of
is looking for patterns in the data. The categories organized on the basis of a
goal is to discover the cultural patterns single semantic relationship. The major
people are using to organize their behav- difference between the two is that the
ior, to make and use objects, to arrange taxonomy shows more of the relation-
space, and to make sense out of their ex- ships among the things inside the cul-
perience. These patterns make up the tural domain (Spradley, 1980). You are
culture (Spradley, 1980). moving from the general to the particu-
lar. For example, children are a type of advertisements), bills, magazines, per-
patient on the pediatric unit. One way sonal letters, and so on. A person from
the children can be categorized further a culture without mail would nd it
is based on their treatment require- more difcult to classify the mail than
ments: medical patients, surgical pa- you would because he or she would not
tients, mental health patients, and so see the differences. This person would
on. These can be further broken down see all the paper as mail. Although it is
based on the specic diagnosis (e.g., true that they are all kinds of mail, they
surgical patients may be those under- each have a unique cluster of attributes
going tonsillectomies, appendectomies, that convey the cultural meaning of the
hernia repairs, and so on) assigned to mail to you. For example, whereas bills
them in the particular culture under are usually impersonal printed forms,
study. After this analysis is complete, personal letters are often hand addres-
the ethnographer looks for relation- sed and often use your rst name. All
ships among the parts or relationships these different bits of mail have compo-
to the whole (Streubert and Carpenter, nents of meaning (attributes) attached
1999). Based on these new categories, to them that make them meaningful.
additional questions and observations You act on the implicit cultural mean-
are made. For example, are surgically ings without even thinking about them.
classied children treated differently You know to discard the junk mail, pay
than children with mental health prob- your bills, and save your personal let-
lems? Are there different types of ters to respond to them.
nurses who care for children with dif- The attributes for all the cultural do-
ferent diagnoses? mains are represented in charts known
3. Componential analysis. Cultural mean- in ethnography as paradigms. A sample
ing comes not only from patterns based paradigm for the domain of mail is pro-
on similarities but also from patterns vided in Table 6.5. It highlights the dif-
based on contrasts. Componential ferences (attributes) for the three cul-
analysis looks for all the contrasts tural categories of e-mail directed
among the cultural categories in the do- toward (1) everybody in the company,
mains. Spradley denes it as the sys- (2) members of your department, (3)
tematic search for the attributes (com- you personally.
ponents of meaning) associated with To complete a componential analy-
cultural categories (p. 131). He uses sis, ethnographers should carry out
the domain of mail to exemplify com- the eight steps outlined by Spradley
ponential analysis. Think about all the (1980, pp. 133139). These include:
different kinds of mail you receive. You Select a domain for analysis (e.g.,
get junk mail (i.e., yers, notices, and people on a pediatric unit).

Take an inventory of all contrasts One way to search for themes is to

previously listed (e.g., some people look for similarities and differences
are patients, some people are pro- across domains. This focuses your at-
fessional staff, some are support tention on the cultural scene as a
staff ). whole. Another strategy to identify
Prepare a paradigm worksheet. themes is to compare different cultural
Identify dimensions of contrast that scenes. A third strategy is to make a
have binary values (e.g., pediatric schematic diagram of the cultural
patient, yes or no). scene, identifying relationships among
Combine closely related dimen- the domains. Spradley suggests a list of
sions of contrast into ones that have universal themes that may be used as a
multiple values (e.g., nurses and basis for scrutinizing your data. The list
doctors are professional staff who is not intended to be exhaustive; rather,
provide care to patients). it simply suggests possible themes that
Prepare contrast questions for miss- you may nd in studying your cultural
ing attributes. scene. The universal themes include:
Conduct selective observations to Social conict. In every social situa-
discover missing information. tion, conicts emerge; identifying
Prepare a complete paradigm. these conicts helps researchers in-
The nal paradigm may be used as a terpret the culture. For example,
chart in your ethnography. It enables conict may arise in a pediatric unit
you to present a great deal of informa- between nurses and physicians con-
tion in a concise manner. cerning treatment plans.
4. Theme analysis. Cultural themes are Cultural contradictions. Every soci-
recurrent patterns in the data that are ety has contradictory messages.
used to connect domains. Spradley What inherent contradictions have
denes them as any principle re- people learned to live with and how?
current in a number of domains, ex- Informal techniques of social con-
plicit or tacit, and serving as a relation- trol. What formal and informal
ship among subsystems of cultural means are used in the group you are
meaning. . . . They usually take the form observing to get people to conform
of an assertion such as men are supe- to the values and norms that make
rior to women (p. 141). Themes are as- social life possible? For example,
sertions that apply to numerous situa- nurses may get physicians who con-
tions and have a high degree of tinually forget to reorder their
generality. To identify themes, an eth- clients drugs to conform to the hos-
nographer must immerse him- or her- pital policy of reordering clients
self in the data. Spradley suggests that medications every 5 days by phon-
if you are not able to live in another so- ing those who refuse to do so late at
ciety for a year or two (as many nurses night to request the reorder.
are not able to do), you can still im- Managing impersonal social rela-
merse yourself in the data by blocking tions. How do people in the cultural
several days to immerse yourself in the scene you are observing deal with
cultural setting and look for themes and people they know well and those
then take several days to review your they do not know well?
eld notes in an intensive manner. This Acquiring and maintaining status.
type of immersion often reveals pat- What cultural symbols and icons
terns and themes that relate to the do- convey status and prestige, and how
mains previously identied. are they acquired and maintained?
Solving problems. A persons cul- drawn such as maps of places, activi-

tural knowledge is often designed to ties, and interactions.
solve problems. Ethnographers seek 4. List all the major and minor themes.
to discover the cultural problems 5. Take an inventory of all the examples
that exist and nd out about the of descriptions of events or experi-
knowledge that is used to solve them. ences (many examples will be in-
For example, within the nursing cul- cluded in the nal report because they
ture, much professional knowledge is bring the analysis to life and make it
focused on helping clients solve meaningful to the reader).
problems related to their health or ill- 6. Identify organizing domains (there may
ness states. Much of what nurses be one or more domains that tie to-
know deals with promoting the gether many facets of a cultural scene).
health of an individual or family or 7. Complete a table of contents.
community. 8. Complete an inventory of miscella-
The nal step of thematic analysis is to neous data (e.g., memos, newspaper
write a summary overview of the cultural clippings, photographs, and other arti-
scene. The goal is to condense everything facts related to the cultural scene).
you know down to essentials and deal pri- 9. Suggest areas for future study (Sprad-
marily with the relationships among the ley, 1980).
parts of the culture. Although ethnogra- These steps organize all of your data,
phers examine small details of a cultural inuence your thinking about the cultural
scene, at the same time they seek to un- scene, and prepare you to write the nal
derstand the broader cultural landscape. report.
Through the analysis process and writing If you are conducting a maxi-ethno-
of the ethnographic report, the researcher graphic study, the nal report may take
is hoping to provide an in-depth analysis the form of a book or monograph that de-
of selected domains, an overview of the scribes a particular culture. If you are
cultural scene, and a description that con- writing a mini-ethnographic study (as
veys a sense of the whole. many nurses who use the ethnographic
(vi ) Reporting on the Findings method do), you will probably present
Ethnographic studies yield a great deal of your ndings in a scholarly journal. It is
data. They include voluminous eld important to know who your audience
notes; transcripts of interviews; artifacts; will be. One of the best ways to know how
and miscellaneous information such as to present an ethnographic report is to
photographs, charts, and maps. The re- look at good examples of what others
searcher must compile all of these items have done.
in a way that meaningfully describes the
observed culture. Spradley suggests that
this be done in two steps: rst, writing a
c. Example of Ethnograhic Research
cultural inventory and then proceeding to An account of an ethnographic study is
write the descriptive report. The cultural provided in Box 6.7. The following com-
inventory organizes the data for the re- ments are offered about the study in
searcher. After it is completed, the re- Box 6.7.
searcher is ready to write the nal report. Emphasis in this ethnographic study is
It includes nine steps: placed on obtaining the emic perspective
1. Make a list of cultural domains. of Chinese women regarding the self-care
2. Make a list of analyzed domains. practices in which they engaged in the
3. Collect all the sketch maps you have family home during the month after the
BOX 6.7 Phases of Ethnographic to observe, describe, document, and ana-

Research lyze the lifestyles or patterns of a culture
or subculture (Leininger, 1991).
Preeld The main method of data collection in
ethnography is participant observation.
Identify the phenomenon to be studied
The act of participating gives opportuni-
and establish the scope of the study.
Select the culture and site to be studied. ties for the researcher to have conversa-
Identify significant variables within the tions with or interview people as he or she
culture to be studied. interacts with them. These are referred to
Review the literature and gather infor- as informal interviews. As in this case, for-
mation on the culture and the specific mal interviews can also be arranged. In
phenomenon to be studied. this study, the researchers carried out
Prepare interview guides, instruments, formal interviews and observations dur-
data collection equipment, and so on.
ing the interviews that were recorded in
Obtain ethical approval.
eld notes. The interviews used open-
Fieldwork ended questions and then semistructured
Gain entrance to the cultural group. questions to explore both anticipated and
Identify key informants. unanticipated information and were tape
Gain cultural immersion in group. recorded. In addition, a demographic pro-
Gather data (i.e., participant observation, le was used to collect data.
interviews, artifacts, and documents).
The sample size of seven was appropri-
Identify major themes.
ate for this mini-ethnography. The study
Collect additional data that is more fo-
was short term and limited in scope. It ad-
cused and selective.
Continue with participant observation dressed the phenomenon of doing the
and interviewing, raising more sensitive month in a subunit of the Hong Kong Chi-
issues. nese culture (postpartum women).
Double check data and probe for mean- This ethnographic study sheds light on
ings. an important area in nursingthe need
Refine themes. for culturally congruent nursing care to
Determine how representative findings Chinese people in Hong Kong. It is likely
are of the culture as a whole. that the advent of Hong Kongs recent
Posteld change of sovereignty will exacerbate this
Finalize analysis and findings. concern and nurses will need to deliver
Develop a cultural inventory. care to a large proportion of clients who
Write an ethnography. hold more traditional Chinese beliefs with
minimal Western inuence.
birth of their rst child. This was achieved C. SCIENTIFIC ADEQUACY OF

by getting personal accounts from the QUALITATIVE RESEARCH
women themselves of the complex issues
that contemporary Chinese women in In quantitative research, the canons of va-
Hong Kong face in trying to follow tradi- lidity and reliability are critical to the con-
tional Chinese practices in a changing dence we put in research ndings. Simi-
world. larly, in qualitative research, it is just as
An ethnographic design is appropriate important to assess the rigor of qualitative
for this study that explored postpartum studies to know how much faith one can
practices in Chinese women. This per- have in the ndings. Multiple approaches
spective is considered a systematic way exist for dening and assessing the trust-
worthiness of qualitative studies and there 2. How applicable are these ndings to
is a gulf of distance among those engaged another setting or group of people?
in this discourse (Creswell, 1998). Trust- 3. How can we be reasonably sure that
worthiness relates to the process of estab- the ndings would be replicated if the
lishing the validity and reliability of quali- study were conducted with the same
tative research. Many researchers argue participants in the same context?
that the use of positivist language is no 4. How can we be sure that the ndings
longer appropriate or adequate to a dis- are reective of the subjects and the in-
cussion of rigor in qualitative research quiry itself rather than the product of
(Creswell, 1998; Lather, 1993; Webb, 1993; the researchers biases or prejudices?
Denzin and Lincoln, 1994; Sandelowski,
When appropriately addressed, these
1993). They reject the notion of objectivity,
questions establish the trustworthiness of
reliability, validity, and generalizability as
a study, its applicability, its consistency,
applied to qualitative methods. They be-
and its neutrality. These terms are associ-
lieve it is as inappropriate as using the
ated with terms from the positivist para-
rules of one religion to judge the behavior
digm of internal validity, external validity,
of people from another religion: certain be-
reliability, and objectivity. Lincoln and
haviors can be seen as sinful when judged
Guba are among the many researchers
through the eyes of one religion but not an-
who believe it is inappropriate to rely on
other (Parahoo, 1997). Lincoln (1995) has
positivists concepts of validity and relia-
proposed a new approach to quality; her
bility to evaluate qualitative research that
views are presented after the traditional
operates within a different paradigm from
approach discussed next.
the positivist perspective. They identify
four alternative concepts that more accu-
rately reect the assumptions of the quali-
1. Traditional Standards of tative paradigm: credibility, dependability,
Quality transferability, and conrmability (Lincoln
and Guba, 1985, p. 300). Taken together,
In their classic statement, Lincoln and
these criteria establish the trustworthi-
Guba (1985) shed some light on the issue
ness of qualitative ndings. Table 6.6 illus-
of assessing truth in qualitative research.
trates the concepts of validity and reliabil-
They outlined four questions that any in-
ity in qualitative research. A discussion of
vestigation (quantitative or qualitative)
threats to validity and reliability in quanti-
into the human condition should con-
tative and qualitative studies is found in
sider if the trustworthiness (quality) of
Chapter 13.
the project is to be properly assessed:
Credibility refers to the accuracy of
1. How truthful are the particular nd- the description of the phenomenon un-
ings of the study? By what criteria can der investigation. The portrayal of the re-
we judge them? ality must be faithfully represented and
Table 6.6 Concepts of Rigor in Quantitative and Qualitative Research

Quantitative Research Qualitative Research
Validity Credibility (authenticity)
Generalizability (external validity) Transferability (ttingness)
Reliability Dependability (auditability)
Objectivity Conrmability
plausible to those who have experienced searcher is interested in determining the

that reality. Credibility can be enhanced extent to which another investigator with
by prolonged time in the eld collecting similar methodological training, rapport
data and repeatedly observing and inter- with participants, and knowledge of the
acting with participants. Triangulation of eld would make the same observations.
data sources (e.g., different persons, This is determined by an audit trail, in
times, and places) methods (e.g., obser- which a researcher does an audit of the re-
vations, interviews, documents), data search process, documenting all the raw
type (e.g., qualitative, quantitative, re- data generated as well as methods and
cordings) investigators (e.g., researcher sources of data generation and analysis
A, B), and theories is a popular means of decisions. By reading the audit trail, an-
establishing credibility. Triangulation other researcher should arrive at compa-
was dened in Chapter 1 as the use of rable conclusions given the same setting.
multiple referents to draw conclusions Transferability is concerned with the
about what constitutes the truth. Trian- generalizability or ttingness of study
gulation leads to trustworthiness of the ndings to other settings, populations,
data by enabling the researcher to look and contexts. In assessing transferability
for patterns of convergence as well as of ndings, the research consumer hopes
counter patterns or negative cases in the to show that the results are not context
data. Essentially, triangulation is sup- bound. To enable the research consumer
posed to support a nding by showing to do this, the researcher must provide
that independent measures of it agree sufcient detail in the study report so
with or, at least, do not contradict it that readers can assess the appropriate-
(Miles and Huberman, 1994). Other tech- ness of the ndings to other settings. The
niques suggested to gain accurate and lack of transferability of ndings to other
true impressions of the phenomenon be- settings is often viewed as a weakness of
ing studied include conducting member qualitative methods.
checks. This involves having participants Conrmability refers to the objectiv-
involved in the data analysisthat is, the ity of the data. Meanings emerging from
participants or subjects who were ob- the data have to be tested for their plau-
served actually read the analysis and sibility, their sturdiness, and their con-
then rene it to be congruent with their rmability so that two independent
experience. Collaborative research also researchers would agree about the mean-
improves the credibility and believability ings emerging from the data. Otherwise
of results. This is accomplished by hav- we are left with interesting stories about
ing other investigators verify the effec- what happened, but they are of unknown
tiveness of data collection procedures, truth and usefulness (Miles and Huber-
the comprehensibility of descriptions, man, 1994). The basic issue here is one of
the inclusivity of samples, and the logic of neutrality. In other words, do the conclu-
arguments. Credibility is enhanced by an sions depend on the subjects and condi-
accurate description of the setting, par- tions of the inquiry rather than on the re-
ticipants, and events observed. searcher (Lincoln and Guba, 1981; 1985)?
Dependability refers to both the stabil- To assess conrmability, the audit trail is
ity and the trackability of changes in the used. The researcher must be explicit
data over time and conditions. The issue about how personal biases, assumptions,
of dependability in qualitative investiga- and values may have come into play in
tions reects the reality that situations the study. The researcher considers the
constantly change and peoples realities plausibility of alternative conclusions
differ. In assessing dependability, the re- and rival hypotheses. To a certain extent,
the researcher is trying to limit the inu- quiry t and make sense and are they true
ence of the Hawthorne effect (see Chap- to the understanding of ordinary people
ter 4), which is an inevitable part of doing in the everyday world?
research. Box 6.8 contains some useful Drawing conclusions from qualitative
questions that can help qualitative re- research data requires careful assessment
searchers assess the trustworthiness of and involves the processes of forming pat-
their ndings. These are not rules to be terns in the data, looking for contrasts,
applied stify but simply test questions clarifying relationships, posing rigorous
that may help researchers focus on the test questions (such as those posed in Box
central question: Do the results of the in- 6.8), and collecting respondent and peer
BOX 6.8 Questions to Be Asked of Qualitative Research to

Assess Rigor
Test for Credibility (Validity) Test for Dependability (Reliability and

Auditability
Do the findings make sense to the people
studied and do they consider them accurate? Are the research questions clear and are the
How context rich and meaningful are the de- features of the study design congruent with
scriptions? them?
Does the account ring true and enable a Did data collected across the full range of ap-
vicarious presence for the reader? propriate settings, times, and respondents
Did triangulation among complimentary meth- suggest the research questions?
ods and data sources produce converging con- Do multiple field workers have comparable
clusions? If not, is there an adequate explana- data collection protocols and do their ac-
tion for this? counts converge in areas where they might
Was negative evidence sought? Found? What be expected to?
happened then? Were data quality checks made (e.g., for bias,
Are the data well linked to the categories of deceit, and informant knowledgability)?
the emerging theory? Were any forms of peer or colleague review
in place?
Test for Transferability (Generalizability and
Fittingness) Were coding checks made, and did they
show adequate agreement?
Are the characteristics of the original sample
Test for Conrmability (Objectivity)
of persons, settings, processes, and so on
fully described to permit adequate compar- Are the studys methods and procedures de-
isons with other samples? scribed in sufficient detail so that an audit
Is the sampling theoretically diverse enough trail can be followed?
to encourage broader applicability? Can the actual sequence of how data were col-
Does the researcher describe the scope and lected, analyzed, condensed, transformed, and
the boundaries of reasonable generalization displayed be followed for the drawing of con-
from the study? clusions?
Do many readers report the findings to be Has the researcher been explicit about per-
consistent with their own experiences? sonal biases, assumptions, and values and
Are the findings congruent with, connected how these may have come into play in the
to, or confirmatory of prior theory? study?
Have the findings been replicated in other Were competing hypotheses considered?
studies to assess their robustness? If not, Are study data retained and made available
could replication be done easily? for reanalysis by others?
SOURCE: Adapted from Miles, M., and Huberman, A. (1994). Qualitative Data Analysis. London: Sage.
feedback. In their investigation of family 2. Contemporary Standards of

adjustment to heart transplantation, Mi- Quality
shel and Murdaugh (1987) demonstrated
how the trustworthiness of their data was More recently, Lincoln and Guba (1994)
assessed using the traditional criteria out- have noted that although the traditional
lined by Lincoln and Guba (1985) for de- criteria (referred to previously) have
termining rigor in qualitative investiga- been well received, their parallelism to
tions. Box 6.9 illustrates the procedures positivist criteria make them suspect.
followed by the researchers to test the sci- Lincoln and Guba continue to call for ad-
entic rigor of their study ndings. ditional critique of the issue of quality in
TESTING SCIENTIFIC RIGOR OF QUALITATIVE RESEARCH: FAMILY

ADJUSTMENT TO HEART TRANSPLANTATION
Redesigning the Dream was identified as the In relation to transferability of findings, the
integrative theme in the substantive theory that sample covered all adult age groups, education
emerged from interviews with 20 family mem- levels, and adult role activities. Therefore, the
bers of heart transplant recipients. The trust- subjects, content, and range of data were ade-
worthiness of the study findings was assessed quate to provide the basis for assessing the rel-
using the criteria for determining rigor in qual- evancy to related contexts and care providers.
itative investigations. Because the partners provided the data on pa-
Six methods were used to assess credibility. tients, the generalizability of the findings is lim-
These included: ited to the perspectives of the partners.
Dependability of the findings was deter-
1. Prolonged engagements at the site mined by an audit trail. To do this, a research
2. Persistent observations during lengthy data assistant reexamined a portion of the tran-
collection periods scribed data from the group sessions. The re-
3. Peer debriefing and exposure of the investi- search assistant was able to draw comparable
gators thinking to a jury of peers by sched- conclusions given the same data, perspective,
uling discussions with the transplant team and context.
members who worked closely with the pa- Confirmability was established by gathering
tients and their families data from the heart transplant team members
4. Triangulation was accomplished by compar- in addition to the subjects. In addition, alterna-
ing the emerging perspective with lay articles tive explanations for the data were explored
on transplant recipients, professional articles with the team members. The researchers also
on life after transplantation, literature on reflected on detailed notes depicting the initial
spouses experiences living with a chroni- directions in the organizing and processing of
cally ill partner, and television documen- the data, enabling changes in the early formula-
taries. tions. Through the means identified by Lincoln
5. Investigators findings were compared with and Guba (1985), the researchers of this inves-
recorded information from consultants who tigation can have confidence in the trustworthi-
worked with family members and patients. ness of their findings and conclude that their
6. Member checks to verify or modify concepts description of Redesigning the Dream re-
that emerged were conducted during the flects the true state of human experience for
group meetings. family members of heart transplant recipients.
SOURCE: Adapted from Mishel, M.H., and Murdaugh, C.L. (1987). Family adjustment to heart transplantation: Re-
designing the dream. Nursing Research, 36(6), 332338.
qualitative research. Although many re- BOX 6.10 Lincolns Standards

searchers continue to observe the four of Quality
traditional criteria (Miles and Huberman,
1994; Streubert and Carpenter, 1999), 1. Diverse inquiry communities exist with
some argue there is no longer a need to unique traditions of research with their
defend the value and rigor of qualitative own standards of rigor. These serve to
research with criteria that parallel those either exclude or legitimize researchers.
in quantitative research. This has led to 2. Positionality: A balanced and honest re-
an effort to determine more appropriate flection of the participants and the au-
thors stance.
methods of determining quality in quali-
3. Community: All research takes place in a
tative research (Creswell, 1998; Sande-
community and serves that communitys
lowski, 1993). Lincoln (1995) traces her purposes.
own thinking on the issue of quality and 4. Voice: Qualitative research gives multi-
the scientic adequacy of qualitative re- ple or alternative voices to participants
search. She has moved beyond the tradi- that must be heard in the text.
tional parallel criteria of the classic work 5. Critical subjectivity: The researcher
by Lincoln and Guba (1985) to her cur- should have a heightened self-awareness
rent perspective. It includes a commit- before, during, and after the research
ment to emergent relations with respon- encounter, leading to personal transfor-
dents, a commitment to a set of stances, mations.
6. Reciprocity: The relationship between
and a commitment to a vision of research
the researcher and participant should be
that promotes justice (Lincoln, 1995).
one of mutual respect, trust, and intense
Flowing from these commitments are sharing.
eight standards found in Box 6.10 for as- 7. Sacredness of the relationship between
sessing the quality of qualitative re- researcher and participant: There is an
search. Adherence to these standards inviolate relationship between the re-
helps the researcher answer the all im- searcher and the participant that is both
portant question Did we get it right? egalitarian and collaborative.
And, as Sandelowski noted (1993): 8. Sharing privileges: Qualitative researchers
share the benefits of their research with
. . . we can preserve or kill the spirit of the participants. This may take the form
qualitative research; we can soften our of authorship, rights to publication, and
notion of rigor to include the playful- royalties.
ness, imagination, and technique we
associate with more artistic endeav- SOURCE: Adapted from Lincoln, Y. (1995).
Emerging criteria for quality in qualitative and in-
ours, or we can further harden it by the terpretive research. Qualitative Inquiry, 1, 275289.
uncritical application of rules. The
choice is ours: rigor or rigor mortis.
examples of qualitative research per-

spectives used to describe and explore
D. ADVANTAGES AND
phenomena of concern to nursing. These
LIMITATIONS OF
methods share similar characteristics
QUALITATIVE APPROACHES
that include:
Qualitative research approaches have 1. Attention to the social context in
made important contributions to knowl- which events occur and have meaning.
edge development in nursing during the 2. Emphasis on understanding the social
past decade. Phenomenology, grounded world from the participants point of
theory, and ethnographic methods are view.
3. The perspective is primarily inductive. Researchers are likely to be subjective in

4. Data collection techniques include deciding which participants to select and
participant observation, interviewing, which data to accept or reject. The small
and printed documents. and unrepresentative samples and the
5. Data collection tools and procedures analytic procedures that rely on subjec-
are subject to change in the eld. Con- tive judgments make qualitative methods
cern is primarily with discovery and appear to be weak in terms of reliability
description of phenomena. and generalizability.
6. Hypotheses are developed during the Qualitative researchers have devised
research rather than a priori. their own terminology such as truth value
7. Analysis is presented in narrative form. and have established their own proce-
dures to assess the scientic rigor of their
Qualitative methods are congruent work. Rigor in qualitative research is less
with the eld of nursings philosophy of about adherence to rules than about -
holistic, humanistic caring. Such methods delity to the spirit and standards of quali-
enable a holistic understanding of the pat- tative work (Sandelowski, 1993).
terns, characteristics, and meanings of the Some researchers believe it is appro-
phenomenon under study. They attempt priate to integrate quantitative and quali-
to focus on the experiences of people and tative data within a single study or a clus-
seek to describe the uniqueness of partic- ter of studies on a particular phenomenon.
ipants human health experiences. Simi- Not everyone agrees, however, that meth-
larly, nurses hope to focus on the client as ods can be mixed in the same study.
an individual requiring holistic care that Creswell (1994) suggests that there are
encompasses biological, psychological, three positions on the subject of integra-
social, cultural, and spiritual dimensions. tion: the purists, who believe methods and
Because qualitative methods focus on the paradigms should not be mixed; the situa-
whole of the human experience and the tionalists, who believe certain methods
meanings ascribed to them by clients, are appropriate for certain situations; and
they provide nurses with deep insights the pragmatists, who support integrating
into the experiences of clients that would methods in a single study.
not be possible using quantitative meth- Table 6.7 summarizes the advantages
ods exclusively. and limitations of the qualitative designs
The major strength of qualitative re- discussed in this chapter. There are ob-
search is the validity of the data it pro- vious strengths in the areas of validity
duces. Because researchers collect data and in the ability to probe in depth the
primarily by means of in-depth inter- human experience.
views and participant observation in nat- The authors of this text believe there
ural settings, the participants true real- are advantages to mixing methods as we
ity is more likely to be reected in the work toward the common goal for re-
rich descriptions that result than would search of understanding of the world in
be reected in data collected in contrived which we live. Both quantitative and qual-
settings using quantitative instruments. itative methods are complementary
The major limitation of qualitative re- processes. The Oxford Dictionary denes
search is its perceived lack of objectivity complementary as to make complete
and generalizability. It is argued that be- that is, one perspective completes or
cause qualitative researchers become makes up for the lack or deciency in the
the research tools, they become so inti- other perspective. Hence, by mixing per-
mately involved with the data collection spectives to address an area of inquiry,
process that they cannot be objective. the weakness of a single perspective may

Pre-experimental
Experimental
Quasi-experimental
Survey Designs
Group administered
Phone survey
Interview
Secondary data
Meta analysis
Qualitative Designs*
Phenomenology
Grounded Theory
Ethnographic
In each category, a means that this is an advantage of the technique; a means this is a possible limitation; means
that in some conditions a limitation, others an advantage.
*Note: this table is intended for comparisons with previously considered designs and uses criteria quantitative
researchers typically use. See body of text for alternative criteria used by qualitative researchers.
be overcome or reduced. Neither perfor higher costs, and researcher training
spective on its own can establish the in both quantitative and qualitative meth-
truth about phenomena of interest to ods. Many believe the benets to knowl-
nurses. Rather, each serves a unique pur- edge development that result from inte-
pose; taken together, they can provide a grating methods outweigh the hassles of
rich repertoire for building nursing knowl- blending methods and encourage nurse
edge. The successful blending of methods researchers to make the effort to consider
in a single investigation requires resolu- integration if it is appropriate to the study
tion of epistemological biases, budgeting question.
E X E R C I S E S
1. Suppose you are interested in explor- search question? Does the researcher
ing the meaning of healthy lifestyle address issues related to scientic
choices for adolescent girls. What type rigor? Comment on how the credibility,
of qualitative design would you chose transferability, dependability, and con-
to study this phenomenon? How would rmability of ndings are handled by
you go about selecting your sample? the investigator. What suggestions
What setting would you select to con- would you make for enhancing the
duct the study? Explain your choices. rigor of the study?
2. You are working with cancer patients 5. Read an article that uses only quanti-
who are waiting to receive bone mar- tative data to respond to the research
row transplants. You are interested in question. Discuss how the addition of
understanding the processes that qualitative data may enhance the va-
their family and they use to cope with lidity of the ndings. What sugges-
the ordeal of receiving the bone mar- tions, if any, would you recommend to
row transplant. Which research design the studys designer to strengthen the
would you select to investigate this study?
phenomenon? Explain your choice and
describe the steps you would follow in
conducting the study.
6. Identify a lived experience from a phe-
nomenological study. List your as-
sumptions and beliefs about the expe-
3. Formulate a research question from the rience. What means would you use as
practice setting that is appropriate for a researcher to deal with your precon-
a mini-ethnographic investigation. Sug- ceived notions and ideas about the
gest the type and size of sample that phenomenon under study? Elaborate
might be appropriate for this investiga- on the process.
tion. Discuss how you came to your de-
cision on sampling. Describe the setting
for data collection and how you would
go about the process of collecting data 7. Review a qualitative nursing research
to answer your research question. article that uses a grounded theory
perspective. Pay particular attention
to the process of data analysis. Com-
ment on how the researcher identied
4. Review a recent issue of a qualitative themes or common patterns in the
nursing research journal such as Qual- data. Describe the core variable. What
itative Health Research or the Western steps were involved in identifying the
Journal of Nursing Research. Select an basic social psychological process?
article of interest to you and assess the
scientic adequacy of the design. Is the
design appropriate to answer the re-
Creswell, J.H. (1998). Qualitative Inquiry and amples and is appropriate for beginning
Research Design. London: Sage. Provides an researchers as well as practicing re-
excellent overview of ve different tradi- searchers who are dealing with qualitative
tions of qualitative inquirybiography, data analysis issues.
grounded theory, ethnography, phenome- Nelms, T. (2000). The practices of mothering
nology, and case studies. in caregiving an adult son with AIDS. Ad-
Denzin, N.K., and Lincoln, Y.S. (1994). Hand- vances in Nursing Science, 22(3), 4657. A
book of Qualitative Research. London: Sage. phenomenological investigation of mother-
This comprehensive handbook synthesizes ing and how it informs our understanding
the existing literature on qualitative meth- of what it means to be women, mothers,
ods and helps to shape the future direction nurses, and human beings.
of qualitative research. Written from the Spradley, J.P. (1980). Participant Observation.
perspective of sociology and education, it is New York: Harcourt Brace College Publish-
a springboard for new thoughts on this reers. An excellent guide to conducting par-
search tradition. It is a key addition for the ticipant observation studies.
library of serious qualitative researchers. Streubert, H.J., and Carpenter, D.R. (1999).
Leenerts, M., and Magilvy, J. (2000). Investing Qualitative Research in Nursing (2nd ed).
in self-care: A midrange theory of self-care Philadelphia: J.B. Lippincott. This clearly
grounded in the lived experience of low in- written text presents the essentials of qual-
come HIV-positive white women. Advances itative research as it relates to nursing.
in Nursing Science, 22(3), 5875. This is a Zerwekh, J. (2000). Caring on the ragged edge:
good example of how midrange theory Nursing persons who are disenfranchised.
emerges in a grounded theory project. Advances in Nursing Science, 22(4), 4761.
Miles, M.B., and Huberman, A.M. (1994). Qual- The voices of seven nurses in their fearless
itative Data Analysis. London: Sage. This is caring for the disenfranchised are heard in
a practical sourcebook that describes a this phenomenological study. The words of
rich variety of approaches for analyzing the nurses reveal the meaning of their work
qualitative data. It contains numerous ex- and the life experiences of their clients.
Chapter 7
Field Study Approaches
CHAPTER OUTLINE
A. Participant Observation Studies D. Field Experiments

1. Rationale for Participant Observation E. Naturalistic Observational Studies
Studies 1. Rationale for Naturalistic Observational
2. Steps in Conducting a Participant Studies
Observation Study 2. Steps in Conducting a Naturalistic
B. In-depth Interviews Observational Study
C. Focus Group Interviews F. Analyzing Qualitative Data
1. Evolution 1. Using Computers to Analyze Data
2. Rationale for Focus Groups 2. Content Analysis
3. How a Focus Group Works G. Conclusion
4. Selecting Focus Group Members
1. Advantages of Field Studies
5. Recording Data
2. Limitations of Field Studies
6. Analyzing the Data
3. Summary
7. Example of a Focus Group Study
8. Advantages and Disadvantages of Focus
Groups
KEY TERMS
Analytic files Grounded theory Nodes

Coefficient of reliability In-depth interviews NUD*IST
Content analysis Index system Participant observation
Field experiment Master field file study
Field notes Master table Proxemics
Field studies Naturalistic observational Tally sheet
Focus group study
224
FIELD STUDY APPROACHES 225
C
hapter 6 presented a general intro- which can enrich ndings from any of the
duction to three popular qualitative types of eld studies covered or even func-
approaches to nursing research: tion as a study method on its own.
phenomenology, grounded theory, and The information in Chapters 6 and 7
ethnography. This chapter discusses eld together is adequate to provide an under-
studies, which are the data collection and standing of the process and the steps in-
analysis methods most commonly associ- volved in doing a qualitative investigation,
ated with these qualitative approaches. as well as give you a feel for the experience.
However, quantitative eld research, in the If you elect to design a study using one of
form of eld experiments and naturalistic the methods discussed, you should seek
observational studies, can also be con- additional sources of guidance for analyz-
ducted in unaltered social environments ing data. Sources by Coffey and Atkinson
and is also discussed in this chapter. (1996), Hanna (1997), Kidd and Parshall
Field studies are among some of the (2000), Maynard-Tucker (2000), Miles and
most admired work in nursing; they in- Huberman (1994), and Silverman (1993),
clude investigations in which the re- referenced at the end of this chapter in the
searcher observes and records the behav- Recommending Readings, are useful re-
ior of individuals or groups in their natural sources for beginning researchers. It is
settings. Part of the appeal of such studies also suggested that you nd a mentor,
is that humans are being directly observed someone experienced in the appropriate
in everyday situations; as a consequence, method of data analysis to assist you to de-
these studies are more convincing be- velop the necessary skills and judgment re-
cause they ring true to the reader. Field quired.
studies tend to pay more attention to the
qualitative dimensions of the phenomenon
being studied. So whether researchers are A. PARTICIPANT
observing who is more likely to exhibit OBSERVATION STUDIES
supportive behavior in crisis situations;
monitoring trends in the portrayal of A participant observation study ordinar-
nurses in popular literature; or compar- ily involves an intensive examination of a
ing baccalaureate, associate degree, and particular culture, community, organiza-
diploma-prepared nurses on caring behav- tion, or group. It is the method used in
iors, their studies are likely to focus on the ethnographic and ethnonursing studies,
context in which the phenomena occur which are discussed in Chapter 6. Nor-
and to quote from cases to illustrate the mally, such a study is based on a careful
generalizations emerging from the studies. and complete exploration of one case and
The three main types of qualitative eld involves having the researcher join the
studies explored in this chapter are par- group for an extended period. For exam-
ticipant observation studies, in-depth in- ple, in studying a preliterate society by
terviews, and focus group interviews. Two living with the group, the anthropologist
types of quantitative eld studies, includ- adopts an emic perspective (see Chapter
ing eld experiments and naturalistic ob- 6). Because the researcher is living with
servational studies, are also considered in the group and is involved in the daily
this chapter. Finally, this chapter closes lives of its members, such studies are
with a closer look at the analysis of quali- called participant observation studies.
tative data. Here, readers are introduced The researcher is a participant in the
to the use of computers in analysis of qual- lives of the groups members, sharing
itative data. The chapter then briey ex- their joys and pains. Such studies fre-
plores the technique of content analysis, quently take much time to complete,
which is not surprising given the need to lives of the people they are studying, at-
learn a new language and to absorb the tempting to provide accurate descrip-
intricacies of the culture. tions and explanations of their activities.
Classical anthropology led the way in The important characteristics of these
developing participant observation tech- activities emerge during the course of
niques; the names of Bronislaw Mali- eld observations. Surveys and experi-
nowski, Margaret Mead, and Oscar Lewis ments begin with specic variables and
come immediately to mind. Nurses associ- proceed to analyze them; participant ob-
ated with this tradition include Mary servers begin with the more ambitious
Breckinridge, Madeleine Leininger, Toni goal of understanding how a whole group
Tripp-Reimer, and Jan Morse. Working in functions.
this qualitative tradition, researchers have The rationale of such studies is that
studied everything from health-care sys- only through sharing in the daily lives of a
tems and exotic cultures to prisons (Gre- group can a researcher fully understand
sham M. Sykes, 1968), hobos (Nels Ander- the behavior that is manifested. Re-
son, 1929), and the world of punk rockers searchers doing participant observation
(Baron, 1989). Erving Goffmans study on attempt to minimize the effect of precon-
mental institutions, Asylums (1962), Wil- ceived ideas by trying to see the world
liam F. Whytes analysis of a street gang, from the points of view of the members of
Street Corner Society (1955), and Becker the group they are studying. By spending
and associatess Boys in White (1961) are a good deal of time within the community,
classic participant observation studies. the researcher gains rst-hand knowledge
of social behavior as it unfolds over time.
The fundamental point is that the conclu-
1. Rationale for Participant sions of a participant observation study
Observation Studies are grounded in the datathat is, based
on direct and careful observations of
The method of participant observation everyday life within the group.
can best be understood by comparing it To the extent that we wish to study
with different research approaches. We commonsense understandingsthe so-
have seen in Chapter 4 that the experi- cial construction of realitythen it is
ment examines causal relations among a clear that we have to try to understand
limited range of variables and that it typ- the world from the point of view of the
ically identies treatment effects of one humans involved in these constructions.
or two variables on a dependent variable. A researcher doing experiments on the
Surveys (see Chapter 5) focus on mea- people involved or conducting a survey
suring a large number of variables and ex- of them may fail to document adequately
ploring the statistical relationships the social construction created by them.
among them. In contrast, participation Participant observation studies try to
observation studies try to understand in- capture both the subjective and the ob-
stitutions, groups, communities, and jective complexity of human behavior,
even whole cultures or subgroups within trying to penetrate the inner lives of the
a culture. These studies are holistic be- people being analyzed. Nurse scientists
cause they attempt to understand the challenge themselves when they attempt
whole group. Typically, researchers us- to understand the subjective dimensions
ing this approach do not begin with a lim- of human activity. Doing participant ob-
ited number of variables like survey re- servation studies is one of the ways in
searchers or experimentalists; instead, which nurse scientists attempt to meet
they immerse themselves in the everyday this challenge.
2. Steps in Conducting a sons whose permission will be required to

Participant Observation Study gain entry. A brief written statement that
outlines the goals and methods of the pro-
Each participant observation study is dif- posed study should then follow this initial
ferent, depending on the research ques- contact. The preparation of this document
tion, design selected, study sample, con- assists the researcher in dening the
text, and setting (see Chapter 6). However, problem more clearly (in participant ob-
the following steps are necessary in most servation, the precise problems studied
studies: gaining entry to the group, estab- emerge through the research process and
lishing a working rapport, creating eld are not necessarily present from the be-
notes, analyzing observations, and writing ginning of the project). The letter outlining
the report. the project should:
Establish the legitimacy of the project.
Step 1: Gaining Entry into the Group This may be achieved by using the sta-
tionery of the sponsoring organiza-
Gaining entry is sometimes a simple mat-
tion, and mentioning the sponsors of
ter but is more difcult at other times. For
the project.
starters, the type of group has to be taken
into account: if the researcher wishes to
Indicate that the project has been
approved by the institutional ethics re-
enter a formal organization (e.g., hospi-
view board.
tal, religious group, prison, bureaucracy,
business organization, or school) the en-
Indicate the goals and methods of the
study. This provides the leadership of
try methods will differ from those used
the target organization with a simple
when attempting to study an informal or-
statement that can be used to commu-
ganization (e.g., community response to
nicate the project to others in the or-
a crisis, street clinic for gays and les-
ganization.
bians, homeless people, street gangs, or
the business lunches of young execu-
Indicate how the goals of the organiza-
tion may be enhanced through partici-
tives). But no matter what the researcher
pation in the project.
studies, entry should be viewed as an on-
going and reciprocal relationship be-
Specify the length of time that re-
searchers will be on site. This provides
tween the researcher and the population
the organizations leaders with impor-
being studied. The researcher has to ne-
tant information as to how long the re-
gotiate a relationship with each person in
searchers will be present.
the study population.
In entering a formal organization, it is
Indicate the amount of time that will be
spent interviewing various members
probably best to begin by nding out as
of the organization.
much as possible about the organization.
Who are the key actors? Are there any crit-
Indicate the extent to which anonymity
and condentiality will be possible for
ical issues (e.g., strike threatened, new
various participants.
methods introduced for care delivery, re-
quired technology and equipment becom-
Indicate the form in which feedback on
the project will be provided to the or-
ing increasingly expensive) currently fac-
ganization. This may take the form of
ing the organization? What are the best
informal verbal reports, seminars, or
times of the week or month to approach
formal reports.
the leadership of the organization?
Having determined that the intended Research reports typically mask the
study would be appropriate, informal con- location of the site and the names of
tact should be established with the per- participants. However, the researcher
should assume that published reports is writing about how Native Americans
circulate among those studied. When use the health-care system (this will
studying a group, it may be difcult to dis- most likely be accepted because the
guise the identity of some of the individ- nurse will be seen by most residents
uals behind the pseudonymsespecially many times during the day as they visit
those whose role is occupied by only one the health center).
person. One must be careful, therefore, in Second, it is usual for a social gap to
promising anonymity before starting a exist between the researcher and the
study because it may not be possible to subjects being studied. Complete ac-
mask everyones identity adequately in ceptance by a community is not a
the nal report even though steps are requirement for successful eldwork.
taken to disguise identities. These steps Indeed, there are advantages to re-
may include using pseudonyms or cti- maining somewhat of an outsider in
tious names of organizations; modifying that the researcher can adopt the role
descriptive data to mask the identity of of a naive learner.
the organization; and when using direct Third, it is helpful to enlist the aid of lo-
quotes, stating the reference as one sub- cal sponsors to legitimize the re-
ject stated . . . searchers presence in the community.
In studying informal organizations, it is For example, if a member of the tribal
usually not necessary to get an ofcials council approves of the presence of the
permission to enter the group; one sim- researcher, it is likely that the tribe will
ply has to establish cordial relations with cooperate in the study even though the
the people being studied. For example, to members may consider the researcher
do observations among homeless people, an exotic addition to their community.
one would simply need to establish a The development of a trusting rela-
solid working relation with them. To keep tionship between the researcher and
things simple, one would probably just the community is key to a successful
say that a book (or an article) is being project.
written about the lives of the homeless,
that the researcher would be talking to Although most participant observers
the homeless in the area, and that the are open about their studies, sometimes
identities of individuals will not be re- covert entry is used to become part of a
vealed in the publication. group. Although arguments have been
Gaining entry to the group is a critical advanced (see the discussion of ethics in
step in participant observation studies. Chapter 10) that all research should be
There are no hard and fast rules to follow done openly, with the knowledge and
because much depends on the personal- consent of those being studied, there
ity and skill of the investigator in gaining have been times when this was not feasi-
access and minimizing the effects of in- ble. Festinger and associates (1956), for
trusion on the group observed. Three example, wished to study a group pre-
general guidelines are proposed for dicting the end of the world. However, be-
those planning a participant observation cause the group initially avoided public-
study: ity, observers joined the group covertly
to conduct the study. It should be noted
First, it is probably wise to keep ones that it is highly unlikely that a similar pro-
explanations simple. For example, a ject would ever receive the approval of an
nurse studying health-care utilization ethics review board today.
patterns in a Native American tribal The following are some of the advan-
community may simply state that she tages of covert entry:
Entry is gained in circumstances in tance; whereas the researcher is in the

which it might have been denied. group temporarily, the regular partici-
Those being observed are less likely to pants are there longer, often for life. The
alter their behavior to please a re- commitment of the researcher in the eyes
searcher (as in overt entry studies), of the group is therefore suspect.
hence increasing the validity of the ob- There are times when gaining entry is
servations. extremely difcult, taking weeks or even
If the people studied think they are months to accomplish (Morse, 1992;
dealing with a regular member of the Morse and Field, 1995; Streubert and Car-
group, they may be more open in shar- penter, 1999). In the case of nurse re-
ing their thoughts and feelings than searcher Evaneshko (cited in Leininger,
they would with someone known to be 1985), who was studying diabetes in an
a researcher. American Indian tribe, she found that it
took months to achieve success in over-
The following are among the disadvan- coming the fear locals who spoke no Eng-
tages of covert entry: lish had of outsiders. She explains the
process of seeking approval from innu-
There may be ethical reservations
merable subcommittees, the tribal coun-
about doing research on individuals
cil, the Department of Indian Affairs, and
without having their consent. Indeed,
the local health authorities, as well as the
as research continues and the ob-
Navaho people themselves.
server develops a rapport with the
group, considerable guilt may arise on
the part of the researcher because of Step 2: Establishing Rapport
the deception involved; today it would
be rare to be able to get an ethics re- One challenge that participant observa-
view board to approve a project in- tion researchers face is to develop good re-
volving covert entry. lationships with the people being studied.
The researcher must play the role of a Maintaining these relationships is some-
regular member and therefore may not times strained because the subjects may
have the freedom or time to roam wish to capture the researcher to their
about, ask questions, and collect data. points of view. The researcher must be
By becoming more intimately in- careful to not become overly attached to
volved, it may be more difcult for the one faction because this would make it
researcher to remain objective. more difcult to get information from
The researcher may have a difcult members of other factions. The researcher
time nding a way of leaving the group may also be blamed for difculties that the
gracefully. group encounters. Furthermore, the re-
searchers objectivity may be strained if an
When entry has been made without de-
unequal amount of time is spent with each
ception, the observer has the advantage
side.
of being able to ask questions, move
about, and explore issues to a greater de-
gree than might be appropriate in every- Step 3: Creating Field Notes
day social relations. Moreover, the ob-
server need not expend effort to disguise Another challenge of participant obser-
the fact that research is being carried out. vation is the creation and recording of de-
The role of the observer is a marginal tailed and comprehensive eld notes.
one: the researcher is in the group but Field notes attempt to capture, with as
not really part of it. There is always dis- much accuracy as possible, descriptions
and interpretations of individuals, inter- In order to make sense of eld experi-

actions, and events. The exact time and ences, the researcher interprets the data
location of observations should be continuously. For example, if the observer
recorded, along with other descriptions sees a person explaining something to an-
to help the researcher recall the events other person, this may be thought to indi-
(e.g., weather conditions, other signi- cate the dominance of the person doing
cant events going on in the community or the explaining. Such dominance may be re-
in the world that day). Emphasis should ected in a variety of ways, and all of these
be placed on reecting exactly what peo- may be thought to reect dominance and
ple say and how they say it and describ- submission behaviors. Grounded theories,
ing the reactions of others to what is said. Corbin and Strauss argue, are built out
As a practical suggestion, Strauss and as- of conceptualizations of behavior, not out
sociates (1964), in their seminal presen- of the actual incidents themselves; obser-
tation, distinguished types of quotations: vations are generalized, and behaviors
are interpreted as reections of concepts
Words recalled verbatim enclosed
(Strauss and Corbin, 1997).
within . . .
In deciding which situations, or per-
When there is less certainty about the
sons, are to be observed, Corbin and
exact wording . . .
Strauss suggest that researchers should
When meaning is clear but wording not
sample according to concepts and their
exact: no quotation marks
properties and not try to locate represen-
Field notes should also distinguish tative individuals. They argue that in
clearly between descriptions of events grounded theory representativeness of
and people and interpretations of them. A concepts, not of persons, is crucial (1990,
suggestion for doing this is to organize p. 9). Thus, if you want to study peer sup-
eld notes so that there is a wide column port in urban adolescent girls, you would
for descriptions and a narrow column for go to where adolescent girls hang out and
interpretations. Note, too, that there are watch how they interact and support each
two kinds of interpretations: (1) the sub- other. You may visit schools, shopping
jects own interpretations of their behav- malls, video arcades, youth centers, and
ior and the behavior of others and (2) the local spots frequented by youths. While at
interpretation that the observer places these locales, the observer would note
on these same activities. events that occur and begin classifying
The emphasis, particularly in the ini- them according to some general princi-
tial stages of research, should be on de- ples. What techniques do girls use to initi-
scriptive accuracy. Field research is a dy- ate contact with peers? What is the nature
namic process in which ideas develop, of the interaction? How are friendships
are rened, and are then tested and mod- negotiated? How is support offered and re-
ied throughout the observational pe- ceived? Perhaps the researcher would
riod. The writing of theoretical memos is then try to observe some of the most and
encouraged to help the observer system- least popular girls to note differences in
atically develop explanations (see Chap- how they interact in offering and accept-
ter 6). These tentative interpretations ing support.
emerge and are tested during the obser- Field notes should be made as soon af-
vational period. Indeed, participant ob- ter eld observations as possible, prefer-
servers do well to concurrently carry out ably on the same day. At the very latest,
both observations and theoretical reec- notes should be made the morning after
tion about these observations (Strauss an evenings observations. Expect to
and Corbin, 1997). spend as much time writing up eld notes
as you spent in the eld. Three hours of particular people you interviewed. On
observation is usually followed by 3 hours the other hand, it may well be the case
of writing up the observations. The use of that not every part of the master le will
a laptop computer is most helpful in com- make it into any of the subles you cre-
piling eld notes. It is also possible to tape ate. In summary, your les will typically
eld interviews and then have these in- include:
terviews transcribed. Some researchers
even prefer to dictate their observations Master eld le. This is the original
and then have them transcribed. Field complete le of the eld notes. Pages
notes tend to be lengthy. The eld notes are numbered and each entry is dated.
for When Prophecy Fails, combined with This represents the raw data to be an-
transcriptions of recorded information, alyzed.
came to well over 1000 typewritten pages Analytic les. These les each deal
(Festinger, 1956). Boys in White produced with a particular type of observation
about 5000 single-spaced typewritten or relationship (e.g., descriptions of in-
pages of eld notes and interview mater- teractions between status unequals,
ial (Becker, 1961). When writing the re- interactions across gender lines, inter-
port, the challenge will be to condense actions of strangers).
the eld notes into a coherent document Background or history le. This sub-
of reasonable length. le contains information drawn from
the master eld le as well as from
other sources.
Step 4: Analyzing the Observations Key character les. Individual les
may be established on key players in
It is recommended that eld notes be en- the organization or group being ob-
tered into a word processor. Then a print- served. These les attempt to reect
out of the complete journal of these the personality, mannerisms, and typi-
notes, called the master eld le, should cal behaviors of central characters.
be placed in a binder and retained.
Before you can begin analyzing the In addition to qualitative analysis of
data, you should make several additional eld data, there will be occasions when
copies of the master eld le to use as some quantitative analysis is appropriate.
raw material for building up analytic les. Where the analysis involves few variables
Analytic les are les relating to a spe- and not too many cases, a quantitative
cic topic or relationship explored in the analysis can be done by hand. Chapter 16
study. For example, you may wish to provides suggestions for how to proceed
build up a suble on interactions be- without the use of a computer and gives
tween status unequals. You would cut tips for how the data set should be pre-
out all the eld notes pertaining to such pared if computer processing is antici-
interactions. These will frequently be or- pated. At the end of this chapter, we dis-
dered along some dimension ( in this cuss the use of computers in analyzing
case, perhaps along the degree of status qualitative data. Content analysis is a
difference between the two persons in- method for analyzing qualitative data but
teracting). Each piece cut from the copy because it involves counting and the use
of the master le should have its original of numbers, some qualitative researchers
(master copy) page number on it. You do not consider it a qualitative analysis
may also want les related to the technique. It is discussed at the end of this
methodology of the study, the history of chapter as a means of quantitatively ana-
the institution or area being studied, or lyzing qualitative eld data.
Step 5: Writing the Report do with how the respondents organize

their world, such as: What kinds of ac-
The challenge in preparing a manuscript tivities generally go on at the meetings
is to condense the eld notes into a re- you attend?
port that is readable and that increases 3. Contrast questions, which urge the re-
the readers understanding of the subject spondent to indicate what is meant by
of the study. Suggestions for writing nal terminology which is peculiar to the re-
reports are presented in Chapter 19. spondents group. For example, What
do you mean when you say someone is
a super nurse?
B. IN-DEPTH INTERVIEWS
In-depth interviews tend to be done by
a senior member of a project. Because
In-depth interviews are often used along
sensitive information is sought, the for-
with participant observations and pro-
mat must be exible and the interviewer
vide a method of collecting respondents
needs to be free to pursue leads sug-
perceptions of their world. In-depth in-
gested by the respondent. Typically,
terviews identify areas to be covered and
highly experienced researchers do such
may not use standardized questions. Typ-
interviewing.
ically, such interviews record peoples
In-depth interviews are often used when
descriptions and explanations of events
high-status people are involved. In these
in their world. These are then used to
cases, it is thought that people of high sta-
communicate this world to the reader of
tus (e.g., leaders of minority people or
the research report. Quotations are se-
church groups, politicians, physicians,
lected from the interviews to illustrate the
and corporate executives) would not re-
points the author wishes to make.
spond well to having a questionnaire read
Spradley provides an excellent guide
to them, as in the case of a structured in-
to in-depth interviewing in his book, The
terview. Therefore, a more open-ended
Ethnographic Interview (1979). Three key
strategy is adopted. Here the interviewer
elements are present in ethnographic in-
makes the interview more conversational,
terviews: explicit purpose, ethnographic
following up in detail on interesting points
explanations, and ethnographic ques-
that the respondent makes.
tions. The explicit purpose simply means
Box 7.1 features the family section of
that both the participant and the inter-
Weitzs study of how AIDS patients deal
viewer are aware that the discussion has
with the stigma of their disease. In her
a research purpose. Even though it may
study of persons with AIDs (PWAs), Weitz
be quite conversational, it should not be
(1990) conducted 23 multiple semistruc-
confused with a casual conversation.
tured interviews, ranging in time from 2
Ethnographic explanations are those that
to 5 hours. Although Weitz began with a
the researcher tries out on the respon-
set list of questions, she probed into un-
dent to see if they make sense. Respon-
expected topics that seemed to warrant
dents are encouraged to use colloquial
additional attention. Weitz expanded the
language and to become the teacher as
topics covered in subsequent interviews
they familiarize the researcher with their
to deal with new themes that were emerg-
worlds. Ethnographic questions include:
ing in the research, as a researcher using
1. Descriptive questions, such as: What grounded theory would have done. Note
do you do when you go to a nursing so- that the report of her study uses a series
ciety meeting? of illustrations drawn from her inter-
2. Structural questions, which have to views; these drawings attempt to sum-

ROSE WEITZ: LIVING WITH THE STIGMA OF AIDS

deserved AIDS as punishment for his sins, re- Weve gotten closer . . . Theres the verbal I love
ported that initially his family wouldnt come in you, theres the letters. One of the nicest things
the room unless they had gloves and a mask and thats ever happened to me . . . is my father sent
they wouldnt touch me . . . [And] for a time I me a personal card. In the inside he wrote God
couldnt go over to somebodys house for dinner. bless you. I love you son. . . It meant the world
And they still use paper plates [when I eat there]. to me.
Even PWAs who feel such precautions are
necessary still miss the experience of physical Another man described how, despite their
warmth and intimacy. They report feeling stig- disapproval of his lifestyle, his fundamentalist
matized, isolated, and contaminated. Christian family had provided him with housing,
Although PWAs fear that their families will money, and emotional support once they
reject them once their illness becomes known, learned of his illness. As he described it, in his
they also hope that news of their illness will family, when little brother needed help . . . that
bring their families closer together. A 38-year- took priority over . . . They were right there.
old store manager who had never had a partic- Diagnosis can also bring families together by
ularly close relationship with his family de- ending previous sources of conflict. Whether to
scribed his fantasy that something like thisan preserve their own health, protect others from
experience where you come this close to death or infection, or because they simply lose interest in
the reality of deathis when you realize whats re- sex once diagnosed with a deadly, sexually trans-
ally important and not whos right and whos mitted disease, PWAs may cease all sexual ac-
wrong. tivity. For health reasons, PWAs may also stop
For the lucky ones, this fantasy material- smoking and drinking. As a result, families that
izes. The oldest man I interviewed, a 57-year- previously had disapproved of PWAs lifestyles
old lawyer, had always considered his father a may stop considering them sick or sinful, even
cold and selfish man, and had never been on if the PWAs continue to consider themselves
good terms with him. This situation changed, gay. Consequently, some PWAs achieve a new
at least partially, when he became ill. As he de- acceptance from relatives who attach less stigma
scribed it: to AIDS than to their former behaviors.
SOURCE: Rose Weitz (1990). Living with the Stigma of AIDS. Qualitative Sociology, 13(1), pp. 2629. Adapted with
permission.
marize the experiences of the 23 men in in arriving at conclusions and would

her study population. spend much more time in exploring the
High-quality journalism and studies theoretical and methodological issues in
based on in-depth interviews are quite such a study.
similar. Although nurse researchers typi-
cally are more concerned with how rep-
resentative their respondents are of some C. FOCUS GROUP INTERVIEWS
social category, they share with journal-
ists a concern for accurate description. A Compared with in-depth interviews, an
report written by a journalist and one by approach that is becoming popular again
a nurse researcher on the plight of the in academic research is the use of focus
homeless, for example, might not be that groups. A focus group typically consists
different. However, one would expect that of 6 to 12 individuals who are asked to
the researcher would display greater care discuss topics suggested by a facilitator.
The idea is for the researcher to observe course materials to the students. Although
the interactions among focus group mem- questionnaires may be used as part of the
bers and detect their attitudes, opinions, evaluation, there is the danger that key is-
and solutions to problems posed by the sues may not be anticipated by the re-
facilitator. On some occasions, even the searcher and may therefore not come to
topic to be discussed is left up to the fo- light. With a focus group approach, re-
cus group, but, the facilitator typically spondents are likely to mention fears they
provides some structure. had about doing a course that requires a
certain level of computer literacy. Or it
may be that students found they spend a
1. Evolution lot of time exploring the materials rather
supercially and they were overwhelmed
The use of focus groups in academic re- by the volume and found it hard to decide
search dates back to the 1940s, when, ac- what was critical to know for examinations
cording to Bruce L. Berg (2001), they versus what was supplementary material.
were used to assess the effectiveness of In short, the focus group interviews are
radio programs designed to boost morale likely to bring to the surface a number of is-
among military personnel. During these sues that would prove helpful in improving
sessions, members of the focus group the method of delivering a distance course
were asked to press either a red button to using multimedia technology.
indicate something they did not like or a Although important information may
green one if they were responding posi- be revealed in one-on-one interviews, the
tively to the program. Later the partici- rationale of focus groups is that they pro-
pants were asked to indicate the reasons vide a dynamic in which participants
for their responses. But the technique learn from one another and develop ideas
was not used much after the war years. together. The researcher is able to detect
Since the 1970s, market researchers whether there seems to be wide agree-
have found it useful to use focus groups in ment on a point or whether a view being
assessing the effectiveness of product expressed is limited to one or two indi-
marketing strategies. But because the viduals. Similar to open-ended questions
technique was largely seen as a tool of the in a questionnaire, the responses may
corporate world, academic researchers well lead into important issues initially
shied away from using focus group tech- unanticipated by the researcher.
niques for another decade. It was only in
the 1980s that the technique grew in pop-
ularity among academic researchers. To- 3. How a Focus Group Works
day it represents an important approach
that one encounters frequently in the lit- Focus groups appear to work best with
erature. from 6 to 12 participants. The informa-
tion collected represents the work of the
group and may be somewhat different
2. Rationale for Focus Groups from what would be obtained in separate
interviews with participants. Some issues
There are many appealing aspects of focus that would be raised in individual inter-
group interviews. Respondents may be views may never be expressed in a group
asked, for example, to discuss their expe- setting; on the other hand, as group mem-
riences taking a distance education nurs- bers talk about issues, they may come up
ing program that is presented using multi- with suggestions and solutions that may
media CD-ROM technology to present not emerge from separate interviews.
4. Selecting Focus Group notes, the researcher may use the tech-
Members niques associated with content analysis.
Normally the results are presented in
Standard sampling techniques (see Chap- terms of quotations from the transcript of
ter 15) may be used along with a variety of the discussions. Care should be taken to
research designs. In the example used reect the qualitative dimensions of the
previously on testing the effectiveness of participants interactions.
a multimedia CD-ROM presentation of
course materials, the researcher would
probably wish to compare the CD-ROM 7. Example of a Focus Group
participants (experimental condition) Study
with students enrolled in the same course
that is being delivered using traditional Box 7.2 presents an example of a study
methods (control condition). One way in using the focus group technique. Stewart
which the differences between the two and associates (1996) reported on the
groups could be measured is through the causes and cessation of smoking among
use of focus groups, questionnaires, test disadvantaged women. The study in-
scores, and individual in-depth inter- volved women from 10 Canadian
views. By using multiple approaches provinces and used a variety of tech-
(known as triangulation), one should be in niques to gather data. Besides phone in-
a good position to assess the effective- terviews and in-depth personal inter-
ness of the new course delivery method. views, the research team also used focus
groups. Note that the reporting style for
the focus group information is similar to
5. Recording Data that used in in-depth interviews. The gen-
eralizations that emerge from the study
Focus group discussions are typically are illustrated by using many direct quo-
recorded either manually by having tations from the participants. One gets a
someone take notes on what is said or by good sense of the frustrations and dif-
tape recording the proceedings. If taped, culties disadvantaged women have in
the discussion may then be transcribed. quitting smoking. What emerges from the
It is useful to record the names of the study is the crucial role that social sup-
speakers so that the discussion may be port plays in smoking cessation (Stewart
tracked precisely during the analysis et al., 1996).
stage. Simply writing down the name of
the speakers in the order in which they
speak is helpful. It is also helpful for the 8. Advantages and Disadvantages
facilitator to note any comments that will of Focus Groups
help in the evaluation of the discussion.
Focus groups present researchers with
Qualitative notes on attitudes and inter-
opportunities to observe how a small
actions are useful additions that may
group responds to an issue posed by the
prove helpful in interpreting the results.
facilitator. In a sense, focus groups are
similar to an open-ended question on an
6. Analyzing the Data in-depth interview. The difference, how-
ever, is that researchers do not probe for
The transcript of the discussion becomes meaning; rather, the group is left to pur-
the data in a focus group interview. Be- sue the topic. The researchers allow the
cause the data constitute the transcript group to spontaneously deal with the is-
of the discussion, plus the accompanying sues rather than having the facilitator di-
SMOKING AMONG DISADVANTAGED WOMEN: CAUSES AND

CESSATION
Miriam J. Stewart, Angela Gillis, Gerry Brosky, Grace Johnston, Susan Kirkland,
Gillian Leigh, Vena Persaud, Irving Rootman, Susan Jackson, and Betty Anne
Pawliw-Fry
Research teams at two Canadian health-pro- At each focus group, the facilitator was as-
motion research centers engaged in several se- sisted by an observer/recorder. During the
quential data collection activities during the six introductory portion of both the group inter-
phases of the project. These included: views and the individual interviews, the facilita-
tor/interviewer clarified the objectives of the
Phase 1: A comprehensive literature review study, provided a brief overview of the proce-
Phase 2: A secondary data analysis of the On- dure, and assured participants that all com-
tario Health Survey and the Atlantic provinces ments were confidential and that their names
Hears Health Surveys would not be used in the report.
Phase 3: Telephone interviews with repre- The focus group guide included eight ques-
sentatives of 13 women-centered cessation tions focusing on the reasons why these women
programs in Manitoba, Ontario, Quebec, began to smoke and continued to smoke; two
Nova Scotia, and Prince Edward Island questions about the perceived impact of non-
Phase 4: Focus group interviews with 254 dis- smoking messages in the media, in public places,
advantaged women and individual interviews and so on; eight questions on their opinions and
with 134 disadvantaged women in 10 provinces experiences with smoking cessation or curtail-
Phase 5: Telephone interviews with repre- ment; and six questions regarding strategies,
sentatives of 29 nontraditional support agen- services, and support that would help them
cies in 10 provinces stop or reduce smoking. The questions were
Phase 6: Telephone interviews with 22 disad- exploratory, open ended, and accompanied by
vantaged women using these nontraditional probes.
support services The research team conducted content analy-
This paper focuses on the Atlantic compo- sis of themes and subthemes in the qualitative
nent of the fourth and sixth phases, which in- data. The researchers identified themes and cat-
volved interviews with disadvantaged women. egories and attempted to achieve consensus.
The three specific objectives of the consul-
Findings: Factors Inuencing Smoking
tations with disadvantaged women were to (1)
record their recollections and reflections about A summary of findings is presented with exem-
their experiences with smoking; (2) elicit their plar quotations to illustrate some themes and
opinions and beliefs about smoking cessation; subthemes. . . . Disadvantaged women began
and (3) explore the psychosocial factors associ- smoking to project an image. The participants
ated with smoking and smoking cessation. A stories about their first smoking experiences as
total of 126 women participated in the Atlantic adolescents focused on their attempts to es-
region interviews: 84 in the nine focus groups tablish an identity consistent with those of their
and 42 in individual interviews. Disadvantaged peers and usually at variance with the expecta-
women smokers over age 20 years who were tions of parents or other members of society.
poor, unemployed, or single parents; had low The picture changed when the participants
levels of formal education; and lived in rural and talked about being adult smokers. In the rea-
isolated communities were included as partici- sons they gave for smoking and the feelings they
pants. Consistent with qualitative methodolo- expressed about being smokers, the women re-
gies, the sampling was purposive. vealed a deep ambivalence about a behavior
SMOKING AMONG DISADVANTAGED WOMEN: CAUSES AND

CESSATION
they knew was dangerous in the long term but selves, and they smoked as a break from the
that was compelling for the immediate sense of monotonous, burdensome routine of their
relief it brought to their daily lives. They spoke days. They smoked as an antidote to boredom
of the hold that smoking had on their lives. or inactivity and to relieve their sense of isola-
Addiction was a pervasive theme. Disadvan- tion, loneliness, and limited social opportuni-
taged women experienced the psychological as- ties. These women also smoked for social and
pects of addiction and the physical manifesta- recreational reasons because others were
tions of addiction and perceived that they smoking, and they used cigarettes as a reward
smoked more after a cessation attempt. and for pleasure.
Nonsmokers who have never smoked, or even Findings: Factors Inuencing Cessation
ex-smokers who are faithful at itI dont think Participants were unanimous in their dislike of
they fully understand how hard it is and how smoking. All participants stated that they would
much cigarettes play on your mind when you are like to be ex-smokers, and almost all had at-
not smoking. Ive never been a drug addict, but it tempted to stop at least once. However, par-
has to be along the same line. ticipants had different motivations for changing
Im sorry I ever picked up that first cigarette their smoking behavior. The level of impact of
again . . . I said just this one wont hurt; Ill never antismoking messages in the media was related
go back to smoking what I used to. But youre ly- to the womans Stage of Change (Prochaska,
ing to yourself. You go back, and youre smoking Velicer, Guadagnoli, Rossi, and DiClemente,
twice as much. 1991).
. . . Participants frequently mentioned lack of
Because of the pressing nature of the par- social support as a key barrier to smoking ces-
ticipants life circumstances, many were caught sation. Lack of support from partners and im-
in a daily struggle for survival. Consequently, mediate family posed the biggest problem. Lack
the long-term benefits of quitting had little rel- of support from friends and acquaintances was
evance for them. Disadvantaged women con- also cited as a barrier:
tinued to smoke to cope with the moment.
Many stressors and few resources with which I used to have temper fits and everything . . . I
to respond to stressful situations characterized was a contrary person. My husband said, Youre
their lives. Coping mechanisms predominated going to have to take up smoking again. I cant
in all the explanations women gave for their live with you.
smoking behavior. They smoked to cope with Friends? They all begged me not to quit. My girl-
the stress, chaos, and crises in their lives. friend said to me, Youre some friend, telling me
Let people in government try to live on $800 a not to smoke. I said, Go ahead, it doesnt
month with two kids. See how far they get with bother me if you smoke. They said, No, we
it. Rent, heat, lights and all that. If my kid gets dont want you with us if you dont smoke . . . I
sick, I cant afford to go out and get medicine for had a hard time.
him. But because I smoke [they say], Dont
smoke any more. Thats four dollars youre Participants unanimously acknowledged the
spending . . . Its [smokings] about the only positive role played by social support during
thing that I can afford that does relax me . . . But cessation attempts.
I have quit in the past . . . and I didnt find I had My father, hes a reformed smoker, hes my hero.
a lot more money to do anything constructive Every time I quit smoking, his praise was good. He
with. I still couldnt go out and buy my own home. never put me down for smoking, but whenever I
Most women had multiple role demands would go and tell him I hadnt had a cigarette in
that offered little space and time for them- a few days his praise made me feel better.
Summarized from Stewart H.J., Gillis A., Brosky G., et al. (1996). Smoking among disadvantaged women: Causes and
cessation. Canadian Journal of Nursing Research, 28(1), 4160.
rect the discussion. In this way, it is the bally aggressive individuals in the group
group that determines the direction of (Berg, 2001).
the discussion. However, a good facilita- In summary, focus group interviews
tor is always prepared with questions in are important qualitative tools for ad-
the event that the group gets too far off vancing our understanding of human be-
the topic or has difculty getting started. havior. It is a cost effective method par-
A further advantage is that of cost. If the ticularly for evaluating the responses of
researcher organizes the focus groups people to new products, assessing the ef-
and runs them, costs should be relatively fectiveness of programs and for sounding
modest, both in terms of time spent and out people for ideas on improving prod-
money expended. However, if numerous ucts or programs.
sessions are done or if recording equip-
ment and transcriptions of dialogue are
done, the costs increase accordingly. But D. FIELD EXPERIMENTS
generally the cost per inference made
should be relatively low compared with Recall that, in a laboratory experiment,
that of other research methods. the researcher introduces a treatment
Focus groups have some distinct disad- variable and measures the response. The
vantages. Although sampling techniques same occurs in a eld experiment; how-
may be used in establishing focus groups, ever, the researchers intervention oc-
the number of groups exposed to the re- curs in a natural setting rather than a lab-
search questions will generally be small; oratory, and the study is usually simple,
therefore, the extent to which generaliza- limited, and quickly completed. Suppose,
tions may be made is limited. One should for example, that a nurse researcher
expect fewer original ideas from the focus greets a stranger while walking along a
group members than from individual in- street and then records the type of re-
terviews (Jackson, 1999), and the ideas ex- sponse ( if any) that results. In this case,
pressed in the focus group may be some- the researcher is intervening in a natural
what more extreme than those reported environment and is interested in record-
from surveys of the same respondents ing the response to a mild form of non-
(Sussman et al., 1991; Jackson, 1999). conformity (greeting a stranger). The
Compared with participant observa- kinds of observations that the researcher
tion studies, focus groups are, in fact, con- can make here are quite limited; how-
trived creations of a researcher; there- ever, as in most observational studies,
fore, they lack the advantage of the fairly accurate measures may be recor-
natural setting observations. Participants ded concerning the subjects age, gender,
know that they are being studied, so they dress, and type of response.
may alter their presentations of them- In Chapter 4, we pointed out that labo-
selves to t the situation. In this way, fo- ratory experiments attempt to maintain as
cus groups are more similar to experi- much control as possible over the treat-
ments or surveysthey lack spontaneity. ment variable (or variables) and over the
One persistent difculty in focus group conditions under which the experiment
work is to limit the inuence of strong- takes place. Because observations are be-
willed individuals to dominate and con- ing made in a natural environment in eld
trol the directions that group discussion experiments, the researcher cannot exer-
take. One must be aware that although fo- cise as much control. But what is given up
cus group data represent group data, it in control may be compensated for by the
almost certainly is skewed toward the fact that the subjects are probably not
opinions and attitudes of the more ver- aware that they are part of an experiment
and hence react normally. Both eld and partner sat right next to a student of the
laboratory experiments attempt to under- opposite gender who was working in a li-
stand the relationship between a treat- brary where many other seats were avail-
ment variable and a measurable outcome, able. This process was repeated a num-
exercising as much control over condi- ber of times. The responses of the target
tions as possible. A researcher conducting students varied, but many erected a bar-
a eld experiment simply takes conditions rier with books to mark off territory; only
and events as they occur naturally, inter- rarely did the subjects ee.
venes in some way, and observes the re- In a related study, a female research
sponse to the intervention. student invaded a group of male students
The following example illustrates a eld who were standing in a public area chat-
experiment. Notice that an environment is ting between classes. The researcher was
set up under controlled conditions and not known by the male students. Once
systematic observations are made of the again, this intrusion was repeated with a
people who pass through the space. This number of all-male groups. Typically these
experiment investigates proxemics, the violations produced a moment of silence:
norms surrounding personal space and the men looked at one another in bewil-
the conditions under which such space derment, and then they turned and ed! So
will or will not be violated. It is a particu- far, no group of male students has volun-
larly important concept in the area of teered to see if the same phenomenon oc-
mental health and illness and counseling. curs with female target groups.
The researcher positions two partners Box 7.3 lists more student studies that
(i.e., people working with the researcher) were completed within the time con-
facing one another, apparently discussing straints of a one-term course. In each of
some issue, in a narrow corridor. A space the studies, students were encouraged to
about 18 inches wide is left between the pay particular attention to the nonverbal
wall and the back of one of the partners; dimensions of their observations. They
the other partner stands against the op- were encouraged to carefully describe
posite wall. As people pass through the how various target subjects responded to
corridor, an observer records information the situation created by the researchers.
about the subjects: perhaps their age, The project reports used qualitative de-
whether they are alone or with others, and scriptions to add depth to the report. Typ-
whether they cut between the investiga- ically the reports also tested a number of
tors or squeeze through the 18-inch space hypotheses related to differences in re-
along the wall. If the subjects cut through, sponse to the situation, such as if elderly
do they acknowledge this by saying, Ex- adults are more likely to be helpful than
cuse me or bowing their heads slightly as young adults.
a nonverbal apology? After a number of The major advantage of these types of
observations of this sort, the distance be- studies is that behavior is observed in
tween the researchers could be increased natural settings and is therefore not con-
or decreased. As a control, the research taminated by the articiality inevitable in
partners could be replaced by large trash- laboratory experiments. Such studies can
cans with 18 inches between them (Jack- be relatively inexpensive to do and, be-
son, 1999). cause conditions can be altered system-
Over the past few years, a number of atically by the researcher, some control
the authors students in a rst-level, one- can be maintained over the experimental
semester methods course have com- conditions. One of the disadvantages of
pleted eld experiments as part of the these studies is that only a limited num-
course requirements. In one, a research ber of variables can be measured. The
STUDENT FIELD EXPERIMENTS*
Project Student or Students Brief

Racial Discrimination Alicia Van De Sande (1995) This project compared speed of
in the Service Industry response in various local stores
with a young female African
American student to that of a
young white female student.
Nonverbal Mary E. Gillis (1996) Observed the impact of

Communication at researchers encroaching upon
Public Telephones: the personal space of target
The Effects of Personal subjects using a public tele-
Space Invasion phone; nonverbal responses
noted.
Proxemic Violations in Corine MacDonald (1996) This study compared nonverbal

the Library responses to different levels of
personal space violations in a
library reading room.
Greeting Behavior Maureen Keough (1996) This study compared the nonver-
bal responses of target subjects
with a greeting from a univer-
sity-aged woamn stranger to
that of a 38-year-old female
stranger.
Cutting Through Kelly Hannay (1997); Students set up different gender

Behavior: Who Cuts Chris Parsons (1997) combinations of interacting
Between Interacting pairs and observed who would
Pairs? or would not cut between them.
*These projects did not include recording information that would allow for the identification of individual subjects, in-
volved minimal risk to the subjects, and could not be carried out practically without the waiver of informed consent.
samples used are usually not representa- recording everyday behavior that occurs
tive and one cannot, therefore, generalize in public places. Because the behavior
the ndings. occurs in public and the subjects are usu-
ally not known by the observer, the issue
of anonymity does not arise. Given the
E. NATURALISTIC absence of free and informed consent by
OBSERVATIONAL STUDIES the subjects, one might question whether
such observations are within legitimate
A naturalistic observational study is one ethical boundaries. Our view is that such
in which the people being studied are studies are ethically acceptable if the ob-
unaware that they are being observed. servation process poses no danger to the
Typically these studies simply involve observer or to the observed, no attempt
is made to encumber those observed by the social environment in any way. The re-
reporting their behavior to authorities, searcher observes and records behavior
the observation is not staged, and the that occurs in natural settings. Similar to
anonymity of the individuals or group ob- eld experiments, naturalistic observa-
served is protected. So whether you are tional studies have high levels of validity
observing the age and gender of those (measuring what they claim to measure).
who wear or do not wear their seat belts However, with both eld experiments and
when driving a car or observe nonverbal naturalistic observational studies, sam-
behaviors in a public meeting, the re- pling is almost always haphazard (based
search appears to be within legitimate simply on who happens to walk by), mak-
ethical boundaries. ing it problematic to generalize experi-
In discussing the role of research mental ndings to people other than
ethics boards in approving naturalistic those actually observed. Nonetheless,
observational studies, the Tri-Council naturalistic observational studies can re-
Policy Statement Ethical Conduct for Re- veal much about human behavior.
search Involving Humans (1998) notes Returning to the area of proxemics, a
that naturalistic observation that does naturalistic observational study could
not allow for the identication of the sub- possibly observe pairs of individuals in
jects, and that is not staged, should nor- conversation (standing position) and note
mally be regarded as of minimal risk. the average toe-to-toe distance that they
The document presents the case that maintain over a 30-second observational
some or all of the elements of informed period. The researcher would use a re-
consent could be waived when: cording form to note information such as
the type of interacting pair (male/male,
The research involves no more than
male/female, female/female) and age and
minimal risks to the subjects.
status differences (teacher/student, par-
The waiver is unlikely to adversely ef-
ent/child, employer/employee, peer/peer)
fect the rights and welfare of the sub-
between the two. Such studies involve few
jects.
variables, so it is possible to analyze them
The research could not practicably be
without the help of a computer.
carried out without the waiver alter-
Many daily activities can be observed
ation.
and recorded without making subjects
Whenever possible and appropriate,
aware that their behavior is being moni-
the subjects will be provided with ad-
tored. Examples of this type of study are
ditional pertinent information after
included in Box 7.4.
participation.
The waivered or altered consent does
not involve a therapeutic intervention. 2. Steps in Conducting a
Naturalistic Observational
An additional discussion of research Study
ethics is presented in Chapter 10.
Suppose, for example, that you were do-
1. Rationale for Naturalistic ing the use of seat belt study and you
Observational Studies wished to relate seat belt use to the inde-
pendent variables of the drivers gender,
The goal of naturalistic observational age, and whether he or she is alone or
studies is to observe and to record be- with others. Let us go through the steps
havior that occurs in natural settings. Un- that would be involved in such a study,
like eld experiments, naturalistic obser- examining in greater detail how each step
vational studies do not attempt to alter would be carried out.
STUDENT NATURALISTIC OBSERVATIONAL STUDIES*
Dressing Appropriately Tracey MacMillan (1995) This project recorded weather

for the Winter appropriateness of outdoor
clothing worn by teenagers
during winter conditions.
Male and Female Food Tracey Pye (1995) Observed the differences in
Preferences food selections of males and
females in a university
cafeteria.
Parking Violations Steven Hawley (1995) This study compared recorded

differences (age category,
gender, and group status)
between people who parked
illegally and those who did
not.
Gender and Smoking Robyn MacDonald (1995) This study recorded the
smoking behavior (yes/no) of
subjects in various public
settings in a small commu-
nity. Gender, age category,
and number in group were
related to whether or not the
target subject was smoking.
The Effects Professors Jeanne Doiron (1995) A number of class situations

Gender on Class were observed and a com-
Participation parison made on the over- or
underrepresentation of
speaking up in class by
gender of student as related
to the gender of the
professor.
Seat Belt Compliance Lori Kiley (1995) Motorists were observed as to

whether or not they were
wearing their seat belts. Age
categories, gender or driver,
and number in the car were
noted.
Speeding in a Small Gordon Barker (1995) Using a radar device borrowed

Town from the local RCMP detach-
ments, student researchers
noted the speed, gender of
driver, age category, and
number in the car on a small
town thoroughfare.
BOX 7.4 Student Researchers at Work (Continued )
STUDENT NATURALISTIC OBSERVATIONAL STUDIES*
Who Purchases Stephanie Power (1996) Patrons purchases going

Healthy Foods at the through the checkout at a
Supermarket local supermarket were
observed for the amount of
junk food included in the cart.
Bank Machine Behavior Kirk D. Bailey (1996) Observers noted whether

users of automatic teller
machines checked the
printed information provided
at the machine and counted
the cash before leaving the
machine.
Nonverbal Cues Used Stacey Desmond (1996) Interacting pairs were

to Terminate a One-on- observed in a mall and the
One Conversation. nonverbal termination cues
were noted and related to age
category and gender combi-
nations.
Drinking Patterns Anne Simpson (1996) Type of alcohol consumed was

observed and related to
gender, gender composition
of drinking group, and the age
of people at local taverns.
Smoking Behavior in 11 Jacques Boudreau (1997) Junior high and high school
to 19 Year Old Students students were observed as
they left school. Smoking
(yes/no) was noted along
with the gender, age, and
number walking together.
Conformity to Stop Judith Finn (1997); Cars were observed as they

Sign Rule Patricia Mbowe (1997) approached stop signs and
the age, gender, and whether
the driver was alone in the
car were noted by the student
researchers.
*The above studies involved observations of public behaviors that were not staged; therefore, informed consent was
not required.
Step 1: Restrictions on Observations specic behavior (seat belt use) to its

general class (risk taking) and then mak-
Because this study concerns the use or ing predictions.
nonuse of seat belts, we would need to re- The following three hypotheses may
strict observations to those cases in be appropriate to the seat belt study:
which we can detect whether the driver
and passengers in a passing vehicle are 1. Female drivers are more likely to use
wearing seat belts. The rst restriction, seat belts than male drivers.
then, would be to limit the observations 2. Older drivers are more likely to use
to recent models of automobiles because seat belts than younger drivers.
trucks or older cars may use lap belts 3. Drivers with others in the car are more
that could not be seen as a vehicle passes likely to use seat belts than drivers
by. Second, we would do the observa- alone in a car.
tions in a place in which vehicles are By formulating hypotheses before col-
moving fairly slowly so that the observer lecting data, one ensures that one is not
has time both to see and record the in- inventing the hypotheses after the analy-
formation. Finally, a decision would have sis has been done.
to be made as to whether observations
will be limited to the driver (this will be
easiest) or include other front seat pas- Step 4: Defining Terms
sengers as well.
Before collecting the data, a careful de-
nition of each of the variables is required.
Step 2: Review of Literature Seat belt use (or nonuse) refers to the ap-
parent use (or nonuse) of a shoulder seat
At this point, a review should be under- belt by the driver (or others, if they are in-
taken of the literature to nd out what cluded in the study). The observer also
studies, if any, have been done on seat belt records the gender and age ( in cate-
use. Furthermore, if the failure to use a gories such as under age 25 years, age 25
seat belt is viewed as a form of risk-taking to 49 years, and 50 years and over) of the
behavior, the literature on risk taking driver. The category alone or with oth-
could be reviewed to nd out about the ers is added to classify drivers who are
variables that are related to risk taking. alone and those who are accompanied by
For example, a researcher could check one or more people.
the literature to see if adolescents show
higher levels of risk taking for this minor
infraction. If use of a seat belt is viewed as Step 5: Develop a Tally Sheet
a form of health-promoting behavior, a re-
A tally sheet should be developed to
view could be done on the health promo-
record the observations. The rst vari-
tion literature to see what other re-
able listed is usually the dependent vari-
searchers have found.
able (wearing or not wearing a seat belt),
followed by the independent variables,
Step 3: Developing Hypotheses gender, age, and alone or with others. The
sheet may look like the one displayed in
At this stage, the researcher attempts to Figure 7.1. The tally sheet should be de-
make predictions about expected out- signed so that the observer can quickly
comes. Such hypotheses may be derived tick off the categories of each driver. A
from common sense, what other recheck mark is easier than writing the esti-
searchers have found, or by relating the mated age or writing female for a female
Figure 7.1 Tally sheet for recording observations.
driver. If conditions or time of day are to master table (Fig. 7.2). In this example,
be recorded, one can place a bracket the dependent variable is arranged on
around a set of observations done during the left side of the master table using the
a rainy, cloudy day or a bracket around categories wearing seat belt and not
those done during rush hour trafc. wearing seat belt. Categories of the in-
dependent variables are arranged across
the top of the master table. In this case,
Step 6: Transferring Data to the we have three independent variables:
Master Table gender, age, and group size. Each obser-
vation ends up in oneand only one
After observations have been completed, cell of the master table. Thus, a female
the tally sheet data are transferred to the driver under age 25 years, who is alone in
Figure 7.2 Master table for seat belt study.
the car and who does not appear to have and 7.3 show the formats for the age and
her seat belt fastened, would end up in the group size analyses. Note that the de-
one cell only. Normally each observation pendent variable is arranged on the left
is entered into the cell with a short line side of the page and that the categories of
indicating one observation; these obser- the independent variable are arranged
vations are grouped together so that the across the top. Also note that the per-
fth line entered into a cell has a line centages are calculated on each of the
across the other four. Each group then re- columns. Details for the construction of
ects ve observations falling into the such tables are presented in Chapter 16.
cell. Later these can be counted and a to- More complex tables containing control
tal for each cell entered and circled at the variables may also be created from the
bottom of the cell. Table 7.1 shows how master table. However, before construct-
the table for seat belt use and gender ing such tables, the student should ex-
could be designed. amine the material in Chapter 17.
Step 7: Creating the Individual Tables

Step 8: Writing a Report
At minimum, three tables would be cre-
ated each showing the relationships be- The nal step is to write a report on the
tween the use or nonuse of seat belts and research efforts. Chapter 19 presents ad-
gender, age, and group size. Tables 7.2 ditional information on report writing.
Table 7.1 Seat Belt Use by Gender

Male Female Total
Seat Belt Use N % N % N %
Using seat belt 144 75.8 138 79.8 282 77.7
Not using seat belt 46 24.2 35 20.2 81 22.3
Totals 190 100.0 173 100.0 363 100.0
Table 7.2 Seat Belt Use by Alone or With Others

Driver Alone or With Others
Alone Group Total
Seat Belt Use N % N % N %
Using seat belt 121 76.1 161 78.9 282 77.7
Not using seat belt 38 23.9 43 21.1 81 22.3
Totals 159 100.0 204 100.0 363 100.0
F. ANALYZING QUALITATIVE author used NUD*IST to analyze inter-

DATA views, eld notes, and reective journals.
Not all qualitative researchers believe
Two very different techniques can help in in the merits of computer programs to an-
analyzing qualitative data: using comput- alyze their data efciently and effectively.
ers and performing content analysis. Some even caution that the use of com-
They are both discussed briey here. puters may alter the aesthetics of qualita-
tive research and increase labor time be-
1. Using Computers to Analyze cause storage and retrieval of more data
Qualitative Data can occur that will need to be accounted
for in the report of results. In addition,
If eld notes are entered into a computer, they argue that depth may be sacriced
it is possible to use special software pack- for breadth because the researcher may
ages to process the information. Cur- be tempted to study larger samples
rently a number of such packages are (Sandelowski, 1995). Computer programs
available, including ETHNO, Text Analy- do offer several advantages over tradi-
sis Package (TAP), QUALPRO, Ethno- tional ways of analyzing qualitative data.
graph, HyperQual, and NUD*IST. For an Let us explore some of the possibilities.
excellent review and description of some If a word processor is used, it is possi-
of these, the reader is referred to special ble to do a great deal of work simply by
issues of Qualitative Sociology on com- entering code words into the word proces-
puters and qualitative data (vol. 14, nos. sors search capability to locate where in-
3 and 4, 1991) as well as a review by stances of these code words appear. The
Weitzman and Miles (1995) of 24 pro- paragraphs associated with the words
grams currently available for use with PC may then be selected and moved to a sep-
and MAC platforms. Also useful is an arti- arate le. Of course, it is crucial to main-
cle by Hanna (1997) explaining how the tain a master copy of the original eld
Table 7.3 Seat Belt Use by Age

Age Category
1625 Years 2639 Years 40 Years Total
Seat Belt Use N % N % N % N %
Using seat belt 87 71.3 110 78.6 85 84.2 282 77.7
Not using seat belt 35 28.7 30 21.4 16 15.8 81 22.3
Totals 122 100.0 140 100.0 101 100.00 363 100.0
notes. And, as always with computer les, Memoing: Writing reective commen-
be certain to have multiple printed copies taries on some aspect of the data as a
and disk copies of the master le. Some basis for deeper analysis
word processors also have the ability to Content analysis: Counting frequen-
number paragraphs and lines; this is help- cies, sequences, or locations of words
ful in keeping track of ones records. A and phrases
good word processing package such as Mi- Data Display: Placing selected or re-
crosoft Word can perform the basic analy- duced data in a condensed, organized
sis required by the beginning researcher. format, such as a matrix or network,
It is important to identify ones analysis for inspection
goals and select a software package ac- Conclusion drawing and verication:
cordingly. Do not be lulled by exotic fea- Aiding the analyst in interpreting dis-
tures of esoteric software packages that played data and testing and conrming
you will not end up using for your project. ndings
Theory building: Developing system-
a. Functions of Software atic, conceptually coherent explana-
tions of ndings; testing hypotheses
No single software package can perform Graphic mapping: Creating diagrams
data analysis by itself. The researcher that depict ndings or theories
must still immerse herself or himself in Preparing interim and nal reports
the data, read and review the transcripts,
and ponder over the meanings and inter- It is beyond the scope of this chapter to
pretation of the data. However, the com- comprehensively discuss all of these
puter can facilitate this process for the functions. Serious qualitative researchers
researcher by performing some of the are directed to the Recommended Read-
manual cutting, pasting, and sorting func- ings at the end of this chapter for a com-
tions efciently, as well as assist in dis- plete discussion of the topic. Some of the
playing the data and preparing the nal functions, however, do warrant comment
report. Miles and Huberman (1994, p. 44) for beginning researchers as they attempt
outline a list of important functions that to analyze eld study data. To discuss
computer software can provide the re- these functions, let us take a practical ex-
searcher interested in analyzing eld ob- ample of a software program that may be
servations. These include: used in qualitative data analysis. The
Ethnograph and NUD*IST are probably
Making notes in the eld the most widely used in nursing research.
Writing up or transcribing functions This does not mean they are necessarily
Editing: Correcting, extending, or re- the best; rather, it reects the nursing re-
vising eld notes search communitys familiarity with the
Coding: Attaching key words or tags to programs. NUD*IST (non-numerical, un-
segments of text and making them structured data indexing searching and
available for inspection theory building) is used here to highlight
Storage: Keeping text in an organized the analysis features available for use
database with data generated through participant
Search and retrieval: Locating rele- observation and other eld studies.
vant segments of text and making them
available for inspection b. NUD*IST Software Program
Data linking: Connecting relevant data
segments and forming categories, clus- NUD*IST 4 is a software program used
ters, or networks of information for analysis of qualitative data. It is de-
signed to replicate the code-and-retrieve ofine documents such as books or videos

function performed manually by most that are imported into the system. The
qualitative researchers. NUD*IST, how- second component is an index system de-
ever, has the capacity to go beyond sim- signed to enable researchers to create and
ply retrieving text according to how it has manipulate concepts and emerging ideas.
been coded. It has two major compo- It is made up of nodes that are containers
nents for managing data as well as ideas for a researchers thinking about the proj-
(Box 7.5 provides an overview). ect. The nodes store the index categories
The rst is a document system which developed by the researcher. With the cat-
holds textual-level data about every docu- egory is stored information about the title,
ment, whether imported into NUD*IST or denition of the category, a memo of ideas
external to it; it also has the option of in- about it and references to the parts of the
cluding a memo about each document. document coded at the node. The nodes of
The documents may be online disk les or the index system are organized into a hi-
BOX 7.5 Overview Of NUD*IST
NUD*IST provides researchers with three ries. Its tools link documents and ideas in a way
basic tools for discovering and exploring the that allow the researcher to:
meanings of unstructured data. These include:
Search for patterns in coding and build new
Managing documents codes.
Creating ideas and managing categories Clarify ideas, discover themes, and store
Asking questions and building and testing memos about the data.
theories about the data Construct and test theories about the data.
NUD*IST manages data documents such as:
Generate reports, including the text, statisti-
cal summaries, and coding patterns.
Text (e.g., reports or minutes, transcripts of Display matrices and build models by linking
unstructured in-depth interviews, or historical with graphical display software.
documents, field notes, newspaper clippings) NUD*IST has two interlocking subsystems
Nontextual records (e.g., musical scores, pho- linked by search procedures:
tographs, drawings, tape recordings, films,
maps and plans) The document system containing every doc-
ument about a project
NUD*IST provides a working environment in
which the researcher can create, manage, The index system made up of nodes that
contain information about the researchers
and explore ideas and categories; minimize
thinking about the project
secretarial routine and maximize flexibility; and
discover new ideas and build on them: NUD*IST opens a project with Explorer Win-
dows on both these systems. By exploring
Explore documents, creating categories and
nodes and the coding associated with them, you
coding texts.
link them to documents.
Manage and explore ideas about the data.
NUD*IST uses search procedures to
Import and export data to link with statisti-
search document text and coding at nodes to
cal programs, spreadsheets, and so on.
look for patterns, themes, and regularities and
NUD*IST is primarily designed for asking inconsistencies in the data, and construct and
questions and building and testing theo- test theories.
SOURCE: Summarized from QSR NUD*IST User Guide (2nd ed). (1997). Thousand Oaks, CA: Sage.
erarchy of categories and subcategories the relationships among concepts. This is

referred to as a hierarchical tree (Fig. 7.3). supported by the ability to give nodes
The trees and the nodes in them can rep- denitions and textual memos that the
resent anything the researcher wishes, researcher can create and edit as neces-
such as people, variables, or ideas. Links sary. The search procedures allow users to
between ideas can be represented in fur- search either document text or coding at
ther nodes or in the node linkages in the nodes to discover and explore patterns
trees. In this way, the nodes can be treated and themes and construct and test theo-
as both textual level (coding documents) ries (Richards and Richards, 1994). Table
and conceptual level (recording things 7.4 summarizes the main features of
about the world). This duality is made NUD*IST that are useful in analyzing data
possible by the tree structure that shows from eld studies.
Figure 7.3 Hierarchical tree diagram adapted from NUD*IST.

c. Advantages and Disadvantages of an understanding of its capabilities and

Computer Analysis limitations. This in itself may intimi-
date many investigators who have be-
Computers and their software programs
come accustomed to doing data analy-
can be both assets and liabilities to qual-
sis manually, as well as beginner
itative researchers. They are not a sub-
researchers. Do not expect the com-
stitute for coherent thinking and compe-
puter to guide you through the process
tence in dealing with the data. The
of data analysis. It is useful in handling
computer is not an expert; it is merely the
data segments, but researchers must
servant of the researcher. With this in
still develop an initial organizing sys-
mind, computers can present potential
tem and present the analysis in an in-
disadvantages that include:
telligent narrative format.
In addition to learning the procedures The cost and maintenance of hardware
for doing qualitative data analysis, and software may be a concern even
novice researchers must also learn to though the cost of both have become
use the software program and develop more reasonable in recent years.
The analysis may be done a certain way lessly miss ideas present in the volu-
because the software dictates it rather minous pages of qualitative data that
than the researchers directing the are generated in most eld studies
analysis process. This may occur in sit-
uations in which a researcher becomes
comfortable with a certain process and 2. Content Analysis
prefers to use it for every project
whether it is appropriate or not. Content analysis is an analytical tech-
The convenience and efciency of the nique primarily associated with qualitative
computer to code data may result in studies. It makes inferences by objectively
inappropriate use and unnecessary and systematically analyzing written or
coding of data simply because the verbal communication. Various media
computer is available to perform the may convey the messages. One may wish
function. Although this is a disadvan- to understand the frequency with which
tage, the ease of coding and recoding sexist messages about nursing are com-
enables the researcher to be creative municated on television or the extent to
and interact with the data more freely, which traditional gender roles are por-
considering alternative coding and trayed in high school career counseling
concepts. materials, or one may wish to compare the
number of positive editorials on health-
Some of the many advantages to using care services now with the number 50
a computer for analyzing qualitative data years ago. The chances are that if these are
include: the kinds of research questions that are
being asked, some form of content analysis
The software has the ability to make will be used. Content analysis may involve
multiple copies of les and move around examining a sample of videos, books,
text within les, as well as duplicate text newspapers, or television programs and
and cut and copy from one le to an- attempting to categorize the messages
other. Before computers, these tasks that are being conveyed in them. In all
had to be performed manually. cases, one is attempting to assess the con-
It is possible to store and le material tent of the message that is being commu-
in one location that is readily retriev- nicated. Box 7.6 lists some projects com-
able. This is particularly important in pleted by students as class projects.
being able to locate material and sort Content analyses are appropriate for
cases promptly. many questions, can be done relatively
Quickly locating a specic piece of text inexpensively, and, if the material is sam-
or data related to a particular theme is pled appropriately, may be taken to rep-
possible without having to reread resent the material that is available. Fur-
through pages of text documents. thermore, the validity of the measures
The copy ability enables you to insert used may be fairly high because direct
several copies of the same text into dif- observations and classications are be-
ferent categories if you think it is ap- ing made. However, validity may also be
propriate. limited because the material presented
The computer forces the researcher to could be selected for its dramatic impact
read each line of text carefully to con- rather than its representativeness. For
sider possible categories or nodes for example, we cannot presume that the at-
coding. Without the computer and its titudes of a dissatised health-care client
software program, the researcher may reported in a story represent the views of
pay less attention to the text and care- the author or of the larger culture.
CONTENT ANALYSIS PROJECTS
Conformity to Thinness Kelly Yorke (1995) This project was a content

in Women analysis of television ads in
terms of the degree of thin-
ness by type of commercial.
Portrayal of Women in Robyn MacConnell (1995) This project examined the

Seventeen female models used in the
1960s, 1970s, 1980s and 1990s
in terms of body type.
Change in the Portrayal Kathleen Lumsden (1995) Using Macleans ads published
of Women, 19501995 over 4 decades, the study
examined changes in the
portrayal of women.
Personal Ads Shelley Conners (1996) Using the Halifax Chronicle

Heralds personal ads, the
study did a content analysis
(what people were looking
for) in heterosexual ads, 1994
to 1995.
Content Analysis of Logann McNamara (1997) This project examined changes

Advertisements in in the extent to which women
Vogue magazine, 1982 in advertisements were
to 1991 portrayed in sexually
permissive or dominant
positions.
Faceism and Gender Wendy Chisholm, Rita Gillis, This project examined changes
Roles Michelle Waye, and Flora in the extent to which women
Murphy (1997) in advertisements are
portrayed with full body
shots whereas men are more
likely to use face shots:
Time magazine in the 1950s,
1970s, and 1990s.
Content analysis is used in nursing and written and verbal communications. Cen-
a variety of related disciplines. It is the tral to all researchers using the technique
main approach used by Kalish and associ- is a concern to analyze the content of the
ates (1985) in their examination of the im- message. Sometimes emphasis is placed
age of nursing as portrayed in the media. on understanding the factors that inu-
Fundamentally, the technique involves ence the content of the message; in other
evaluating the content of communica- cases, the researcher is concerned with
tions, ranging from examining messages the effects of the message; and in still oth-
conveyed by various media to evaluating ers, the researcher is concerned with
variations in the message itself. In all choice of materials selected. Because the
cases, however, the content of the mes- analyst typically wishes to accurately de-
sage is part of what is analyzed. scribe a particular phenomenon, it is im-
portant that the procedures used repre-
sent all instances of the phenomenon
a. Basic Decisions in Content Analysis fairly.
Typically content analysis has both quali- ( iii ) What Is to Be Measured and
tative and quantitative dimensions. There How?
may be some counting of the frequency What is being measured? Is it a theme,
with which some phenomena occur. But the presence or absence of some phe-
classifying messages into various cate- nomenon, or counting how often some-
gories may involve qualitative choices as thing is mentioned? And, having decided
to which categories are relevant. Because what is to be measured, how are the clas-
something about a message is analyzed, sication decisions made? What other
the analyst must make a number of funda- pieces of information are to be recorded
mental decisions. Some of these are iden- about the unit being measured? For ex-
tied here. ample, it may be appropriate to record
the gender of an author, the type of a
( i ) Unit of Analysis
magazine, or the name of a program. Sug-
In starting to do content analysis, one
gested strategies for coding (placing into
must rst decide on the unit of analysis. Is
categories) information are discussed in
it an interview transcript, a speech, a
Chapter 13. One of the key issues is to en-
video; is it an issue of a magazine, a story,
sure that different people analyzing the
a book, a scene in a lm, or a whole lm?
same content will agree on the catego-
( ii ) How Are the Units to Be rization. This can be accomplished by
Selected? having the coders work through the
Chapter 15 outlines the basic procedures same items independently and then mea-
for selecting samples and for deciding on sure the extent of their agreement. A co-
sample size. Standard sampling proce- efcient of reliability may be calculated
dures attempt to remove bias from the as follows:
Number of units in identical category

Coefcient of reliability
Total number of units coded
47 agreements on categorization (gender)

50 individuals categorized (total)
.94
The coefcient of reliability provides a trained to achieve this level. A great deal
measure of agreement between the coders of common sense needs to be shown when
on the categorization of the items being adopting the 0.6 rule because the com-
analyzed. The proportion of times there is plexity and subtlety of the categories may
agreement is reected in the coefcient. A inuence the outcomes. To assess the gen-
rule of thumb is to have a minimum coef- der of a nurse in a recruitment advertise-
cient of 0.6. This means that coders should ment is one thing (in this example, you
agree on the precise category assignment would expect a coefcient close to 1.0!),
6 times out of 10. Coders need to be but to decide whether a senior citizen
housing complex is to be categorized as women were more likely to be shown in

appealing would produce less agreement sexy poses than men. The researchers
among coders. If you could not get a 0.6 re- also noted that the ratio between body
liability coefcient here, you should re- and face shots had not changed over the
view your denition of the different cate- decades nor had the types of products be-
gories to see if they could be rened ing promoted. Poses, however, had be-
further. come signicantly more sexy over time.
( iv) How Is Information to Be
Recorded?
Before classifying data, a tally sheet G. CONCLUSION
should be developed for recording the in-
formation. A possible model for such a Table 7.5 summarizes the strengths and
sheet would be the one used for record- weaknesses of eld study approaches.
ing eld observations (see Fig. 7.1). There Overall, these studies are weak in clarify-
should be space on the form to note all ing causal inferences and in providing
the variables, including the exact loca- samples representative enough for the
tion of the item being measured. ndings of a study to be generalizable;
they are strong in the area of validity and,
(v) How Is the Information to Be
in the case of participant observation and
Analyzed?
in-depth interviews, in the ability to probe
What techniques are to be used in pre-
deeply into the behavior being examined.
senting the information in a report? If the
researcher is interested in showing what
causes variation in the incidence of the 1. Advantages of Field Studies
phenomena, how is this analysis to be
performed? If computer analysis is antic- When we consider the strengths of eld
ipated, be certain that coding is done so studies that involve participant observa-
that transfer to the computer will be tion, in-depth interviews, or focus group
easy. If analysis without a computer is an- studies, there are numerous advantages.
ticipated, follow the steps outlined in the Such studies, for example, usually attempt
naturalistic observational section of this to provide a holistic understanding of the
chapter and in Chapter 16. total social system involved in the case. In
the case of participant observation stud-
ies, because the analysis is done over a pe-
b. Example of Content Analysis riod of time, social and other processes
can be observed (e.g., how friendships are
To illustrate content analysis, a study in formed and dissolved). The relationships
the general area of gender roles is featured between individuals and between parts of
in Box 7.7. This study was completed by their social world are of concern to the re-
undergraduate students who were inter- searcher. Emphasis is placed on obtaining
ested in changes in gender differences ex- careful, in-depth descriptions that can
hibited in magazine advertisements from help to develop hypotheses worth testing
the 1950s to the 1990s. The students ex- with other research strategies. Further-
amined advertisements in Time magazine more, because observations are made
testing six hypotheses: three for gender of actual, real-life activities, there is in-
portrayal and three for changes over time. creased validity in the measures: the re-
The students found that women were searcher does not rely on articial settings
more likely to be shown in body photos (as in experiments) or on respondents
rather than face shots than men and that abilities to report their behavior (as in sur-
BOX 7.7 Social Researchers at Work
FACEISM: GENDER DIFFERENCES IN MAGAZINE ADS

Method eyes with a straight mouth on males. Non-
The procedure entailed a content analysis of sexy poses would include fully dressed mod-
the ads in Time magazine from the 1950s to the els, smiling models, and standard poses.
1990s. Ads were analyzed and classified ac- Data Recording
cording to facial versus body prominence in ads,
sexy versus nonsexy ads. In all cases, differences The selected ads were content analyzed and the
in the portrayal were noted for each gender. information was recorded on a tally sheet,
The students were interested in whether there recording for each ad the year of the publica-
had been changes over time in how men and tion, gender of model, type of shot (face versus
women were portrayed. The first 3 years of 3 body), type of product, and type of pose.
decades were used in the analysis: 1950 to Data Analysis
1952, 1970 to 1972, and 1990 to 1992. The first
Information from the tally sheets were trans-
15 ads that met the rules for selection crite-
ferred to a master table and then cross-tabula-
ria were chosen for analysis:
tion tables were created to display the various
1. Ads had to have either a male or a female relationships and to show the tests of signifi-
adult subject; ads with more than one model cance.
were not used.
2. Ads had to have a facial or a full-body shot of Results
the person. A facial photo was one defined as Six tables were presented to test the hypothe-
one that showed the person only above the ses. Two of the tables are shown here. Table 1
chest area; body shots included those that indicates the relationship between type of
showed the person from head to toe. photo and gender. Note that the female mod-
3. Ads had to include a product that could be els were more likely to be shown with a full
classified as either domestic or nondomestic. body photograph. The difference is noted as
4. Ads had to have a person in either a sexy or statistically significant. Table 2 indicates that
nonsexy pose. Sexy poses were defined as there has been a shift to the use of more sexy
poses of any gender, seductive clothing on poses when the 1950s are contrasted with the
any gender, pouty lips in females and piercing 1970s and 1990s.
Table 1 Type of Photograph by Gender

Gender
Male Female Total
Type of Photograph N % N % N %
Facial photograph 16 57.0 4 23.5 20 44.4
Body photograph 12 43.0 13 76.5 25 55.5
Total 28 100.0 17 100.0 45 99.9
Chi-square 4.95; degrees of freedom 1; one-tailed test probability 0.05; decision: reject null hypothesis.
SOURCE: Summarized from Chisholm, W., Waye, M., Murphy, F., and Gillis, R. (1997). Faceism and gender roles.
Unpublished research design paper. St. Francis Xavier University, Antigonish, Nova Scotia, Canada.

BOX 7.7 Social Researchers at Work (Continued )
FACEISM: GENDER DIFFERENCES IN MAGAZINE ADS
Table 2 Type of Pose by Decade

Decade
1950s 1970s 1990s Total
Type of Photograph N % N % N % N %
Sexy 1 57.0 7 23.5 7 44.4 15 33.3
Nonsexy 14 93.0 8 53.3 8 53.3 30 67.7
Total 28 100.0 17 100.0 45 99.9 45 100.0
Chi-square 7.20; degrees of freedom 2; one-tailed test probability 0.05; decision: reject null hypothesis.
SOURCE: Summarized from Chisholm, W., Waye, M., Murphy, F., and Gillis, R. (1997). Faceism and gender roles.
Unpublished research design paper. St. Francis Xavier University, Antigonish, Nova Scotia, Canada.
veys) but, instead, records actual behav- coders, statisticians, editorial advisers,
ior. The researcher may also probe deeply and recording devices and computers be-
into a whole culture in order to come to a ing placed in the eld for extended peri-
full understanding of how its various parts ods and at considerable cost. Most often,
t together. For anthropologists, the ma- however, eld studies are of the less ex-
jor contribution of such studies is not just pensive variety. Although direct costs
their ability to enable us to understand the may not be that great, keep in mind, how-
world from the natives point of view but ever, that the time commitment is often
also their ability to describe a culture that extensiveoften several years.
may quickly be disappearing. Such de- Field experiments facilitate making
scriptions are of historic signicance be- causal inferences, are strong on validity,
cause if they are not made now, the phe- and are often fairly inexpensive to com-
nomena may be lost forever. plete. Naturalistic observational studies,
As noted in Table 7.5, in-depth inter- similar to eld experiments, are inexpen-
views have the advantage of exibility be- sive and have high validity because actual
cause the interviewer can probe deeply behavior is being observed in a natural
into areas that seem to be particularly setting. In such studies, subjects are not
relevant. The probes do require much even aware that they are being observed.
skill on the part of the interviewer and
cannot be done easily by inexperienced 2. Limitations of Field Studies
researchers.
Participant observation studies, focus Participant observation, in-depth inter-
groups, and in-depth interviews can usu- views, and focus group studies have im-
ally be done quite inexpensively (e.g., us- portant roles in answering many nursing
ing one observer with a pencil and lots of research questions. Many examples of
paper and time) or be on the expensive such research may be found in the nurs-
side, with several paid observers, a sec- ing literature. But they do have their
retarial transcription service, a crew of problems.

Pre-experimental
Quasi-experimental
Survey Designs
Group administered
Phone survey
Interview
Secondary data
Meta analysis
Qualitative Designs
Phenomenology
Grounded theory
Ethnographic
Field Designs
Participant observation
In-depth interviews
Field experiments
Naturalistic observational
One of the weaknesses of participant time, a number of conditions probably will

observation studies, in-depth interview have changed. Moreover, the observers
studies, and focus group studies is their themselves may have rather different im-
inability to tell whether the patterns that pacts on the organization. Thus, there are
emerge are representative or peculiar to many factors that confound an interpreta-
the institution or group being studied. An tion of any changes observed by different
additional problem is that such studies researchers. Making veriable causal in-
are impossible to replicate because these ferences is difcult because only one case
studies involve examining a unique com- is being examined.
bination of individuals interacting with In naturalistic observational studies,
one another at a particular time. Although the subjects are observed unobtrusively
it would be possible for another observer and are not aware that they are part of a
to go into the same institution at a later study. A notable disadvantage of such
studies is the limit on what can be ob- criteria different from those used to evalu-
served. Because subjects are not asked ate the studies of quantitative, positivist
questions, the variables observed are researchers. Many nurse investigators do-
limited to gender, age, and other charac- ing qualitative research try to understand
teristics that are manifested during the the subjective meanings an action has for
time of observation. For example, a study the person performing it; such research
that is examining the impact of group size emphasizes how an individual interprets
on smoking would not know whether it is situations. Thus, both the questions asked
group size, or the fact that an observed and the methodology used are different for
subject is out of cigarettes, that has in- qualitative and quantitative researchers.
hibited smoking. The validity of qualitative researchers
studies is determined by their ability to
3. Summary produce a description of, for example, an
individuals behavior that rings true to the
It is in the area of validity that eld stud- individual concerned. For the quantitative
ies shine. By observing behavior in nat- researcher, on the other hand, the issue
ural settings, the researcher can have may simply be whether or not a relation-
greater condence that what is being ob- ship is statistically signicant (meaning
served is not inuenced articially by the that the covariation of variables is at a
respondents knowledge that they are level unlikely to have been produced by
part of an experiment or by the image chance factors; see Chapter 12 for more on
management that may occur in studies this topic).
based on questionnaires. And although As this chapter illustrates, a range of
there are problematic issues related to qualitative and quantitative methodolo-
the researchers interpretation of the be- gies can be used in eld studies. What dis-
havior being observed (perhaps simplify- tinguishes these approaches methodolog-
ing assumptions that the same behavior ically is their reliance on observations of
means the same thing to various actors people going about their everyday activi-
or that the same person would react the ties, thus minimizing the inuence of the
same way the next time a similar situa- researcher on the behaviors that are being
tion arises are incorrect), there is little observed. Nonetheless, it must be recog-
question that observing spontaneous be- nized that it is the researcher who inter-
havior in everyday situations can provide prets the behaviors being observed. It is
convincing evidence that has not been probable that any two observers will
manipulated by a researcher. come to rather different conclusions. Fur-
Researchers doing participant observa- thermore, the researcher, particularly in
tion studies who emphasize a more quali- participant observation studies, cannot
tative approach to understanding behav- avoid having an inuence on how the
ior should have their work judged by group being observed works.
E X E R C I S E S
1. As an exercise in taking eld notes, at- would propose to record the informa-
tend a meeting of a political party, reli- tion gathered.
gious group, or special interest group.
Ask a classmate to attend the same
meeting. Working independently, re-
cord eld notes on the same event. Put 3. Try to come up with an idea for a eld
your eld notes into nal draft form in experiment or quasi-experiment. Out-
which you each carefully distinguish line what the variables would be and
between your descriptions and inter- how they would be measured and
pretations of events. Now compare the recorded. Develop a linkage to a the-
two versions. What are the main points ory of nursing or human behavior for
of difference? What principles are il- your project.
lustrated by the exercise? Do most of
the differences between the versions
concern points of interpretation? How
can the reliability of the observations 4. Develop a naturalistic observational
be improved? project in which you attempt to do one
of the following: nd out who plays
video gambling machines, who rents
horror movies, or who buys lottery tick-
2. Try to think of some activity in your ets in your home town. For your project
home community that impacts the indicate what restrictions would be
health of your community and that you necessary; what hypotheses could be
would like to investigate using partici- investigated; who would be observed
pant observation techniques. Develop and under what conditions; what your
a proposal indicating what group or tally sheet would look like; and how you
activity you wish to observe, what would design two of your nal results
kinds of questions you think might be tables.
interesting to pursue, and how you
Becker, H.S., Geer, B., Hughes, E.C., and vice on the diversity of ways to analyze
Strauss, A.L. (1961). Boys in White. Chicago: qualitative data.
The University of Chicago Press. One of the Emami, A., Torres, S., Lipson, J.G., and Ekman,
classics: a painstaking analysis of the life of S. (2000). An ethnographic study of a day
medical students and of their socialization care center for Iranian immigrant seniors.
into the role of doctor. Western Journal of Nursing Research, 22(2),
Berg, B.L. (2001). Qualitative Research Meth- 169188. This paper reports on the positive
ods for the Social Sciences. Boston: Allyn effect on seniors regular participation in a
and Bacon. An excellent introduction to day care enter.
qualitative research methods. Goffman, E. (1961). Asylums. Garden City, NY:
Coffey, A., and Atkinson, P. (1996). Making Doubleday. This is Goffmans classic study
Sense of Qualitative Data. London: Sage. of a total institution. A ne illustration of
This easy-to-read text examines a number qualitative research.
of key complimentary approaches to qual- Hanna, B. (1997). From raw data to chapters
itative data analysis and offers practical ad- via the support of NUD*IST. Health Infomat-
ics Journal, 3, 6165. This article reports on Schwarz, M., Landis, S.E., Rowe, J.E., et al.
using NUD*IST to expedite analysis of par- (2000). Using focus groups to assess pri-
enting practices of teenage mothers. mary care patients satisfaction. Evaluation
Kidd, P., and Parshall, M. (2000). Getting the and the Health Professions, 23(1), 5871.
focus and the group: Enhancing analytical This study is a good example of the use of
rigor in focus group research. Qualitative focus groups.
Health Research, 10(3), 293308. This arti- Silverman, D. (1993). Interpreting Qualitative
cle describes the analytical challenges in Data. Thousand Oaks, CA: Sage. This text fo-
the interpretation of focus group data and cuses on the issues of observation, analysis,
suggests approaches for enhancing the and validity in qualitative research. It exam-
rigor, reliability, and validity. ines a range of different approaches within
Maynard-Tucker, G. (2000). Conducting focus the qualitative tradition and considers the
groups in developing countries: Skill train- relationships between qualitative and quan-
ing for local bilingual facilitators. Qualitative titative research.
Health Research, 10(3), 396410. A practical Silverman, D. (1997). Qualitative Research: The-
guide to training focus group facilitators. ory, Method, and Practice. Thousand Oaks,
Mason, C., Orr, J., Harrisson, S., and Moore, R. CA: Sage. This provides a comprehensive
(1999). Health professionals perspective overview of studies using participant ob-
on service delivery in two Northern Ireland servation, in-depth interviews, and focus
communities. Journal of Advanced Nursing, groups.
30(4), 827834. This study uses participant Spradley, J.P. (1979). The Ethnographic Inter-
observation and content analyses relating view. New York: Holt, Rinehart and Win-
to service delivery systems. ston. A brief but practical guide to in-depth
Miles, M.B., and Huberman, A.M. (1994). Qual- interviewing.
itative Data Analysis. Thousand Oaks, CA: Stewart, M.J., Doble, S., Hart, G., et al. (1998).
Sage. The denitive source book on quali- Peer visitor support for family caregivers
tative data analysis that is a must for qual- of seniors with stroke. Canadian Journal of
itative researchers libraries. Nursing Research, 30(2), 87117. This paper
Morse, J.M., and Field, P.A. (1995). Qualitative reports on a home visiting program de-
Research Methods for Health Professionals. signed to provide support for seniors who
Thousand Oaks, CA: Sage. An excellent have experienced a stroke.
guide to issues involved in doing qualita- Walker, L. (1998). Weight related distress in the
tive research in the health sciences. early months after childbirth. Western Jour-
Morse, J.M. (Ed.) (1991). Qualitative Nursing Re- nal of Nursing Research, 20(1), 3044. This
search: A Contemporary Dialogue. Newbury paper uses content analysis to examine
Park, CA: Sage. This ne collection of arti- womens descriptions of their feeling about
cles presents a variety of perspectives on their weight after childbirth.
issues in qualitative nursing research. Weitzman, E.A., and Miles, M.B. (1995). Com-
Savage, J. (2000). Participant observation: puter Programs for Qualitative Data Analy-
Standing in the shoes of others? Qualitative sis. Thousand Oaks, CA: Sage. A compre-
Health Research, 10(3), 324339. This arti- hensive review of a large number of
cle parallels participant observation with programs, with summaries and appraisals
the clinical practice of nursing. of their functions and uses.
Chapter 8
New-Wave Research:
Contemporary Applied
Approaches
CHAPTER OUTLINE
A. Action Research C. Health Promotion Research

1. Characteristics 1. Characteristics
2. Steps in the Action Research Process 2. Scope of Health Promotion Research
3. Challenges in Conducting Action 3. Methods
Research 4. Future Directions
4. Example of Action Research Study D. Feminist Research
B. Evaluation Research 1. Philosophical Tenets
1. Purpose of Evaluation Research 2. Application of Method and Methodology
2. Steps in Evaluation Research 3. Example of Feminist Research
3. Pluralistic Approaches to Evaluation E. Advantages and Limitations of
Research Contemporary Applied Approaches
4. Examples of Evaluation Research F. Postscript to Part 2
Studies
KEY TERMS
Action research Feminist research Participatory action research

Comparative studies Formative evaluation Randomized clinical trial
Content analysis Health promotion research Reflexive critique
Dialectic critique Meta-analysis Summative evaluation
Evaluation case study Needs assessment Unilineal model
Evaluation research
263
T
here is a growing trend in nursing to- combination of inquiry, intervention, and
ward research that empowers par- evaluation that should lead to some im-
ticipants and leads to improved provement in practice or empowerment of
practice. The ideology of health-care re- participants. The prime intention of this
form, improved health-care delivery, the type of research is to have an impact on
need for evidence-based practice, and a policy making, be it at the level of the work
belief that nursing services should be in- unit, the community, or the government.
formed by research into the needs of re- New-wave approaches can use quantita-
cipients and shaped by collaboration with tive or qualitative approaches or a combi-
them all lead to a quest for new research nation of the two to achieve their research
methodologies and approaches that re- aims. All of the approaches within this new
ect this ideology (Hart and Bond, 1998). tradition are described in this chapter.
Various researchers have coined terms to
describe approaches that do not t tradi-
tional research designs. Labels such as col- A. ACTION RESEARCH
laborative research (Parahoo, 1997), ac-
tion activist-oriented research (Denzin and The term action research can be traced
Lincoln, 1994), new generation research back in the literature to the work of Kurt
(Streubert and Carpenter, 1995, 1999), par- Lewin (1944), who is generally considered
ticipatory evaluation (Feurstein, 1988), the founder of action research and change
new paradigm (Meyer, 1993), and commu- theory (Hart and Bond, 1998). Lewin, a
nity reective action research (Boutlier, Prussian psychologist who immigrated to
Mason and Rootman, 1997) are terms America in 1933, introduced the term ac-
found in the literature to refer to this new tion research as a way of studying a social
brand of research. In addition, there is a system at the same time as attempting to
growing interest in feminist research, not change it. Action research is simply de-
only in nursing but in the other social sci- ned today as the systematic collection
ences as well. And, similar to the other ap- and analysis of data for the purpose of tak-
proaches, feminist research has a concern ing action and making change. It provides
with issues of social change. We have practitioners, organizations, or communi-
christened such designs with new-wave ap- ties with tools to solve their problems.
plied research to reect their growing pop-
ularity and utility within nursing. For pur-
poses of this discussion, we have included 1. Characteristics
four designs in this category:
Seven criteria have been identied by
Action research Hart and Bond (1998, pp. 3738) to distin-
Health promotion research guish action research from other method-
Evaluation research ologies. Action research:
Feminist research
Is educative
What do these four approaches have in Deals with individuals as members of a
common? Why have we grouped them to- social group
gether? These are reasonable questions Is problem focused, context specic,
given the lack of consensus on groupings and future oriented
and classications of research designs. Involves a change intervention
Our response is that we see them as Aims at improvement and involvement
unique designs with unique purposes that Involves a cyclic process in which re-
do not t traditional methods. They share search, action, and evaluation are inter-
in an iterative process that involves some linked
NEW-WAVE RESEARCH: CONTEMPORARY APPLIED APPROACHES 265
Is founded on a research relationship purpose of bringing about change and par-

in which those involved are partici- ticipation by community members (either
pants in the change process. geographic community or people with
common interests and shared experi-
The aim of all action research is the gen- ences). PAR places great emphasis on col-
eration of practical knowledge that has the laboration with research participants by
potential to improve a specic situation or considering them as co-investigators
practice. It does not aim to generalize so- throughout the entire process. Through
lutions to problems that apply across a the process of participation, community
range of settings because the solutions are members become empowered to dene
designed for the particular setting in which their own problems and nd solutions.
the research is conducted. In nursing situ- PAR increases the participants under-
ations, an action researcher would at- standing of the issue they are working on.
tempt to nd a solution for a particular The research element provides an oppor-
problem by implementing changes in prac- tunity for them to test assumptions so
tice and then closely monitor and evaluate that they can be more condent of their
the changes during implementation. For ground as they go forward to implement
example, Gale and associates (1998) de- change. The process of participating en-
scribe an ongoing action research project ables members to build skills, condence,
designed to explore, develop, implement, and knowledge (Barnsley and Ellis, 1992).
and evaluate the role of the generic health-
care support worker in a high-dependency
unit for people with complex physical and 2. Steps in the Action Research
psychological needs. The introduction of Process
the generic health-care worker has poten-
tial to create a number of problems related Action research proceeds in cyclical
to professional role boundaries and bal- stages that involve planning, implement-
ance of skill mix in practice settings. The ing, reecting, and evaluating and involves
project is funded by the Department of collaboration between researchers and
Health, United Kingdom, and consists of participants throughout the entire pro-
three phases. Phase 1 includes an explo- cess. It is an iterative process that paral-
ration of the new role and the identica- lels the nursing process; this may explain
tion of problems and difculties created by its increasing popularity with nurses. It is
its implementation. Phase 2 is the imple- a particularly useful research approach in
mentation phase and involves the estab- todays rapidly changing health-care envi-
lishment of a Practice Development Unit to ronment because it allows nurses to im-
manage and develop the new role. Phase 3 plement changes in practice and simulta-
is the evaluation of the impact of the pro- neously evaluate the impact of such
ject with data collected from providers change. Below is a list of some of the steps
and users of the service. Practicing nurses that are necessary in most action research
will watch with interest as this project un- studies, including participatory action re-
folds. It has the potential to teach us a search projects. Action research is de-
great deal about use of this new role in scribed here as if it were a linear process,
health-care delivery. but in truth it does not follow a series of
Participatory action research (PAR) is stages. Rather it is a dynamic process with
a subset of action research. It consists of a spiral of cycles in which research, ac-
three elements: research, adult education, and evaluation interact with each
tion, and sociopolitical action (Parahoo, other at various phases (Waterman et al.,
1997). It emphasizes both research for the 1995).
Step 1: Entry into the Community and participants decide which data collection
Development of a Collaborative methods are most appropriate to the spe-
Relationship cic situation. Usually triangulated data
generation is recommended with at least
In action research, the community may three methods selected to transcend the
represent an actual geographic commu- limitations of each other and lead to more
nity; an institution or agency such as a effective problem solving (Streubert and
school, health-care facility, or government Carpenter, 1999). Participant observation,
department; or a grassroots group of peo- interviews, focus groups, diaries, ques-
ple with similar interests and issues. Ac- tionnaires, and surveys are valuable meth-
tion research aims to create an equal part- ods of data generation. At the heart of the
nership between the researcher and the planning stage is the development of an
researched in which representatives from action plan that will lead to problem reso-
both come together and form a research lution. The action plan emerges after data
team. The extent of the collaboration be- analysis.
tween the researchers and those re- Analysis varies depending on the data
searched may vary from review of the generation techniques used. Quantitative
problem and diagnosis to full collabora- data are analyzed in traditional ways using
tion in all stages of the process. Action re- statistical methods and qualitative data us-
searchers who facilitate collaboration at ing qualitative methods. Researchers and
all stages believe they are able to nd the participants work together collaboratively
most practical solutions to identied on the data analysis, but researchers often
problems (Whyte, 1991). take the lead in this step because of their
prerequisite knowledge and expertise re-
Step 2: Assessment of the Situation quired for analysis. It is critical, however,
and Identification of Issue to involve participants to make sure that
the interpretations of the data t the real-
This step involves gathering as much data ity of the situation. Depending on the
as possible from a variety of sources to analysis produced, an action plan is devel-
identify the problem or issue. At this step, oped by the co-researchers. The action
you are interested in getting detailed data plan details the planned change interven-
about the situation as it exists before a tion, the implementation process, the plan
change is implemented. Data generation for facilitating reection, and evaluation
continues throughout the entire process plan (Streubert and Carpenter, 1999). This
and involves a wide range of approaches, stage requires a number of meetings with
both qualitative and quantitative. Data col- the co-researchers to discuss and plan the
lection is usually a collaborative process details. Records and detailed notes are
involving both researchers and partici- made of all meetings and decisions by the
pants as members of the research team. At researchers and used in the writing of the
this point, steering groups or planning nal plan.
committees may be formed to inform the
process. You are concerned with identify- Step 4: Implementation of the Action
ing a problem (something that needs Plan and Reflection
changing).
This stage involves the actual implemen-
Step 3: Planning for Research and tation of the planned change. The imple-
Action mentation usually occurs over a specied
period of time. At this point, the re-
Planning begins with a detailed analysis of searchers may move into a more periph-
the present situation. Researchers and eral role and community members may
be more involved in the actual implements, unlike the old method that simply
mentation of change. Researchers remain required nurses to identify the assessment
present, however, to guide and facili- categories. Others complain that the new
tate the reection that occurs with the method takes longer because you have to
implementation (Streubert and Carpen- individualize the notes, unlike the old sys-
ter, 1999). tem that used a generic check-off system.
Reection is the process of thinking Hence, a number of explanations emerged
about the implementation of the new for nurses dissatisfaction with the new
change and its impact on the key players. system based on reexive critique. Using
It occurs simultaneously with implementa- dialectic critique, the researcher may ask
tion. Reection is promoted through group participants to talk about the time element
meetings and observation periods during involved in using the new charting system.
which the participants and co-researchers The researcher would look for conicting
may record their thoughts and experi- information about time. It may be that new
ences with the planned change. Two aids graduates who learned a variety of chart-
to reection are dialectic and reexive cri- ing methods in their educational programs
tique. Reexive critique is a process that say it does not take more time to imple-
enables participants and researchers to ment, but older nurses accustomed to the
make explicit, alternative explanations for traditional charting system say it is very
events or experiences. It facilitates discus- time consuming. Discussing these two op-
sion between the researchers and partici- posite positions would provide insight into
pants and leads to greater insights and ac- the variety of factors inuencing imple-
ceptance of multiple explanations for mentation of the new charting system.
events. Dialectic critique, in contrast,
makes explicit, internal contradictions in
the data rather than complementary ex- Step 5: Evaluation of the
planations. By exposing the contradictory Implementation
nature of phenomena in the change situa-
Coresearchers work together to evaluate
tion, the researchers and participants
the implementation according to the ac-
come to a clearer understanding of the
tion plan. Evaluation can occur at the end
change process (Streubert and Carpenter,
of the implementation stage as well as at
1999).
various points throughout the implemen-
For example, a new method of recording
tation stage. Evaluation includes data gen-
nursing notes is introduced in the maternal
erated during the reection process as
and child nursing unit. The new method is
well as additional information on which
evaluated by interviewing nurses working
the team decides. Evaluation data may be
on that unit about their satisfaction with
similar to data that was collected in the
the method. Most nurses interviewed are
initial assessment phase. Usually evalua-
dissatised with the new method, stating
tion meetings are held with the research
that it is time consuming and inefcient.
team and key stakeholders to discuss the
Participants are engaged in reexive cri-
interpretation of the evaluation data and
tique by the researcher, who queries fur-
verify conclusions.
ther about the source of their discontent.
Some nurses reply the charting time is
longer because they do not know what Step 6: Report and Reassessment
sheets to record specic information on
and subsequently must reread instruc- In action research, a report is written by
tions. Others complain that the new me- the research team that incorporates a plu-
thod requires nurses to provide objective rality of explanations for events observed
and subjective data to support assess- as well as questions for future considera-
tion (Streubert and Carpenter, 1999). The the research team) and who will be a
report is intended to promote dialogue research respondent (i.e., member of
among participants and stimulate addi- the broader population who provides
tional courses of action to be taken by par- data and validates results at various
ticipants, with or without the guidance of stages). The researcher must be sensi-
the researchers. This report on ndings tive to the participants agendas.
and process becomes the initial step in a Diversity of values, perspectives, and
new cycle of this iterative process. abilities among community members
and researchers must be valued. On a
Step 7: Planning Future Action practical level, this often presents dif-
culties based on the different perspec-
Based on the report and evaluation data, tives of researchers, practitioners, com-
planning future action continues. The cy- munity members, and academics. One
cle begins again with implementation, re- participant, for example, may be inter-
porting, and reassessment (Deagle and ested in broad social change, but an-
McWilliam, 1992). other may be interested only in the im-
proved practice situation that the
project should create. This feature of ac-
3. Challenges in Conducting tion research complicates the relation-
Action Research ships of collaborators and increases the
time required to focus the research
Action research offers many opportuni-
project and complete action research.
ties to nurse researchers. An obvious one
is the ability to implement change to im-
Issues of power imbalances and estab-
lishment of egalitarian relationships
prove practice and evaluate the change
require constant vigilance throughout
simultaneously. Another strength is the
the research project. Power balances
ability to empower participants through
may shift over the course of the project
involvement in the research process. Be-
because various community members
cause research produces knowledge, it
may change. Often the commitment to
empowers those who carry it out or com-
a research project by the community
mission it. Action research also provides
member may not be the same as the
practical knowledge that enables nurses
commitment of the university-based re-
to solve specic problems in their prac-
searcher who participates for the dura-
tice setting. Despite these important ad-
tion of a research grant.
vantages, action research presents a
number of challenges to the researcher,
All members of the research team
must be sensitive and responsive to
which include:
the need for different forms and types
The challenge of involving commu- of leadership at different stages of the
nity members as part of the research project (Lindsey and Stajduhar, 1998).
team. Engaging the appropriate stake- At times, community members will
holders and maintaining their commit- take the lead, particularly during the
ment to the research project over time implementation stage; at other times,
requires tact, effort, time, and knowl- the researcher may take the lead, such
edge of the community. The commu- as in data analysis.
nity must rst be appropriately dened Action research is extremely time
and then key stakeholders identied. consuming. Identifying and mobilizing
Decisions must be made as to who will participants and whole communities to
be a research participant (i.e., a co- be involved in the research process are
researcher and community member of time intensive and require a high level
of negotiation, tact, energy, and com- patory research project and highlights the
mitment on the part of the research ndings.
team. Time must be provided to train
members to be part of the research
team; enable full community participa-
tion; and allow the cyclical, iterative B. EVALUATION RESEARCH
process to work as it is intended.
The current emphasis on provision of
quality care and evidence-based practice
4. Example of Action Research and decision-making points to the need
Study for nurses and other health-care profes-
sionals to provide evidence that what
To illustrate action research, lets look at they are doing works and why it does.
the process of participatory research and Evaluation research provides a mecha-
community empowerment to reduce nu- nism to do this. Evaluation research tends
tritional inequities in a group of low- to focus on a particular program, product,
income urban women. method, procedure, event, or policy. An
Health education programs have tradi- important distinction exists between eval-
tionally focused on providing information uation and evaluation research. Evalua-
to individuals as a way of improving their tion is a process of assessing the value of
health through lifestyle modication. This something. Practitioners evaluate their
method has lacked demonstrated effec- performance, their delivery of services,
tiveness for enhancing or improving health the quality of care provided, and so on by
status. Travers (1997), following a critique reecting on what it is that they do and
of health education practices, suggests how well they do it. The difference be-
that an alternative orientation is needed in tween this and evaluation research is
health education. She uses the process of that the latter is a systematic appraisal us-
participatory research and community or- ing the methods of social research for the
ganization to design a project to address purpose of generating knowledge and un-
the inequities in nutritional practices for a derstanding that can be used for decision
group of low-income urban women. The ar- making. It is an applied form of research
ticle highlights the process of social action that provides utilitarian answers to prac-
by monitoring changes associated with the tical questions for decision makers, such
womens experiences. The article is of note as: Who is beneting from the program or
because it shows how the participatory re- service? Is the program cost effective?
search experience can be an empowering Should the intervention or program be
educational experience for participants. continued? Is the program achieving its
This article demonstrates how the princi- intended goals? In what areas does the
ples of participatory action research em- program need to be improved? Another
phasize valuing peoples knowledge, deriv- term used to refer to evaluation research
ing questions from the perspective of the is outcomes research.
people, and helping oppressed people to
reect on their situations. It aims to de-
velop critical consciousness and improve 1. Purpose of Evaluation Research
the lives of those involved in the research
process. The article also points out the re- Evaluation research is often classied ac-
alities and limitations of engaging in this cording to the purpose of the investigative
type of research project. Box 8.1 describes activities. Usually there are three main rea-
the methodology involved in this emanci- sons to engage in evaluation research.
REDUCING INEQUITIES THROUGH PARTICIPATORY RESEARCH

AND COMMUNITY EMPOWERMENT
Background tained from one source could be checked
This study was part of a larger research project against data obtained by another.
with the purpose of initiating nutrition education Qualitative interview data and field notes
for social change following an explication of the were analyzed, not so much for interpretation
social organization of nutritional inequities among of data but rather to show the process of eman-
socially disadvantaged women and their families. cipatory education for the womens group. To
do this, data were organized chronologically
Methodology
into chunks that showed the progression of
Participatory action research was the method- group activities.
ology used in this study with a group of low-
Results
income urban women to reduce nutritional in-
equities for themselves and their families. Through the process of participatory research,
the empowerment of research participants to
Method
initiate collective action for social change to re-
Research participants were low-income women duce nutritional inequities was possible. The re-
who self-selected to attend a coffee group at a sults section describes the details of the
local parent center. A total of 33 women actively process by which the participants came to un-
participated in the group throughout the dura- derstand their experiences and acted on that
tion of the study but never all at the same time. understanding to create change.
Although a sample of convenience, the group For the first few months of the group meet-
appeared representative of the low-income fe- ings, the discussion was unstructured allowing
male population in the city in which the research the women simply to talk about their experi-
was conducted. ences surrounding the difficulties of trying to
Participant observation was the primary feed their families. The women listened to each
method of data collection. The researcher was others experiences. Suddenly they were not
a participant observer at the community drop- alone. By listening to how others overcame dif-
in parent center in an urban neighborhood in ficulties and similar experiences, they began to
Nova Scotia, Canada. The researcher partici- learn coping skills from each other and build
pated in the life of the center, serving meals in hope by working together toward solutions. Al-
the soup kitchen, unpacking food from the food most imperceptibly, the group sessions moved
bank, and preparing and eating meals with the from complaining sessions to consciousness-
staff and volunteers. raising sessions. Consciousness raising is a prac-
One month after initiation into the parent tical process that begins with the experience of
center, the researcher initiated a series of oppression. Women come together to talk
group interviews with women attending the about their experiences and, in doing so, break
center to discuss their experiences of feeding the culture of silence. They begin to see prob-
themselves and their families. In total, 27 group lems not as individual failures but as sources
interviews lasting 2 hours each were conducted that are rooted in structures affecting the lives
over a 16-month period. Semistructured inter- of all women alike. The consciousness-raising
view guides were used to organize the discus- process mirrored the process of learning expe-
sions thematically. Participants also introduced rienced by the research participants in this
topics through discussion of their relevant ex- study. They began with their experiences of op-
periences. The use of triangulated data collec- pression, by gender, by class, and sometimes by
tion techniques (participant observation and race. They broke the culture of silence by shar-
group interviews) was useful as an internal va- ing their stories. They came to see the common
lidity and credibility check because data ob- and political roots of their oppression and thus
REDUCING INEQUITIES THROUGH PARTICIPATORY RESEARCH

AND COMMUNITY EMPOWERMENT
were able to shed their self-blame and take on address the inequities. First they wrote letters
a new way of relating to the social world. This to the stores addressing inequities in pricing,
was the beginning. quality, and service. Knowledge of the super-
Exploration of womens food purchasing ex- markets roles in construction of inequities was
periences enabled a process of education leading the catalyst for community action. They con-
to community development. At an early meeting, fronted both local supermarket chains with their
one woman discussed her experiences and per- findings and recommendations for action. Both
ception that food costs more while shopping in stores changed buying practices to decrease
low-income neighborhoods. With pricing in- price inequities between locations. One store
equities brought to the groups attention, it was introduced a bulk-food section in the inner-city
decided to do a comparison of chain stores in store, a move that eliminated price inequities
low-income and middle-income neighborhoods. between neighborhoods. These changes en-
The researcher taught the women to do unit abled women and socially disadvantaged families
pricing to enable them to make in-store price to purchase food more affordably. These
comparisons. The women split into groups to do changes live on long after the research process
the unit pricing comparison in four stores: two ended. Because the women were still dependent
in the inner city and two in the suburbs. Results on commercial outlets for the bulk of their food,
revealed that inner-city stores were consistently the researcher worked with a subgroup of
5% higher than prices in suburban stores of the women to secure funding for a grassroots co-
same chain. A 10.7% price differential was found operative grocery outlet. It is operated out of
between the store most frequently used by the the parent center and run by a committee com-
women and the store furthest from their neigh- posed of women from the community. This
borhood. The women learned that cost savings process was an empowering learning experi-
outweighed transportation costs. ence in a personal, political, and economic sense
In keeping with participatory action research for the women involved. This is emancipatory
principles, the women decided to take action to health education!
SOURCE: Summarized from Travers, K. (1997). Reducing inequities through participatory research and community
empowerment. Health Education and Behavior, 24(3), 344356.
These include needs assessment, forma- ples of questions appropriate for this
tive evaluation, and summative evaluation. type of evaluation research include:
Needs assessment is a form of inquiry
that assesses the needs, problems, con- What are the perceived needs of new
cerns, or conditions of a group, commu- mothers for information on infant care
nity, or organization that should be ad- during the postpartum period?
dressed in future planning activities. What are the continuing education
Needs assessments are usually prompted needs of RNs moving from acute care
when one is dissatised with the current to community-based service delivery?
situation and level of service delivery or What is the range of services required
quality. The assessment provides valu- to provide comprehensive health pro-
able information on what is needed to im- motion services to street youth in ur-
prove the situation for the future. Exam- ban centers?
What are the appropriate human re- are useful in determining whether to repli-
source requirements to staff commu- cate, discard, modify, or replace an inno-
nity-based primary health-care cen- vation. Unlike formative evaluation re-
ters serving rural communities? search, which produces knowledge that
can be used to improve the situation dur-
A needs assessment is useful in setting ing the actual implementation stage of a
priorities in health care and allocating project, summative evaluation research
scarce resources. Questionnaires, inter- produces knowledge that is useful in
views, and focus groups are popular ways terms of making decisions about whether
of collecting data for needs assessments. the innovation or activity should be con-
Nurse researchers have conducted nu- tinued or discontinued. Its purpose is not
merous needs assessments of various to make improvements to the situation
client groups. For example, Kulig and during the innovation implementation
Thorpe (1996) studied the teaching and like in action research or formative evalu-
learning needs of culturally diverse post- ation; rather, the purpose of summative
RN students using both in-depth, semi- evaluation is to make a decision after
structured individual interviews and fo- completion of the implementation of the
cus groups. McKay and Diem (1995) used innovation as to whether or not the inno-
a self-report questionnaire to study the vation made a difference to the group that
health concerns of 1416 adolescent girls received it or whether one innovation is
attending public schools to identify fac- better than another. Therefore, what dis-
tors and themes that would help in plan- tinguishes evaluation research is not the
ning health promotion programs for this method used but rather the purpose or in-
population. tent for which it is done (Cormack, 1991).
Formative evaluation is a form of in- Summative evaluation is usually con-
quiry that focuses on how well a new ser- ducted using a comparison procedure. A
vice or ongoing program or activity is popular design used for evaluation re-
meeting its objectives. The thrust of for- search is the randomized clinical trial
mative evaluation research is to identify (RCT) or experimental design, which was
what is and what is not working currently discussed in Chapter 4 in terms of its lim-
so that remedial action can be taken to itations and strengths. Randomized clin-
improve the situation at hand. The re- ical trials use one or more control groups
searcher is interested in determining if, and experimental groups, depending on
how, and to what extent the goals and ob- the number of interventions. Subjects are
jectives of the ongoing program are being randomly allocated to the groups. They
met. It yields practical information that are frequently used to test the effective-
can be applied to improve the activity, ness of drugs, therapies, and other inter-
service, or program. It is similar to action ventions. Summative evaluation may also
research in that it is focused on making include a comparison of two innovations
changes for improvement of the present or activities within the same program.
situation (Cormack, 1991).
Summative evaluation is the third cat-
egory of evaluation research. It is con- 2. Steps in Evaluation Research
ducted to determine the effectiveness,
value, and worth of an innovation. For Because evaluation research uses a plu-
many evaluation projects, this includes rality of approaches, the specic steps in
evaluation of the costs as well as the ef- any given evaluation research project
fectiveness of the program or interven- vary depending on the design chosen by
tion. The results of summative evaluation the investigator and the nature of the
evaluation. There is, however, consensus tionnaires? Semistructured interviews? Di-

that the traditional approach to evalua- aries? Will data be sought from sources
tion research proceeds through four other than the clients of the program to de-
phases: identication of program objec- termine attainment of objectives? All of
tives, measurement of program objec- these are possibilities.
tives, collection of data, and analysis an
interpretation of data.
c. Collection of Data
a. Identification of Program After the researcher determines how the

Objectives behaviors of interest will be measured,
the process of data collection begins. The
Objectives are usually developed from the amount of time and energy required for
broad statements of program goals. Ob- this step vary depending on the research
jectives should be stated in specic be- design selected in the previous phase.
havioral terms that are easily measured. For example, a time series study, a single
Behavioral terms refer to the specic be- group before-and-after study, or a ran-
havior change the researcher would ex- domized experimental study all require
pect to see in the people to whom the pro- two or more data collection periods to
gram was delivered. For example, if you get baseline ( i.e., before intervention)
are a community health nurse teaching a and comparative data. The researcher
disease prevention program to school- collecting evaluation data needs to be
aged children, your goal may be to have sensitive to the dynamics operating in
children use proper handwashing tech- the eld. People are often suspicious of
niques. The behavioral objective for this evaluators and might feel threatened by
goal might be, Children will wash their the presence of the researcher. This may
hands before eating meals. It is important lead to defensive responses and uncoop-
that there is consensus between the re- erative behavior. It is important that the
searcher and program stakeholders about researcher has good interpersonal skills
the objectives to be used for evaluation and experience dealing with such situa-
because the objectives set the stage for tions so that respondents can be put at
the other phases. ease. Participants may also respond the
way in which they think the researcher
would like them to respond and hence
b. Measurement of Program distort their answer. For example, if par-
Objectives ticipants were aware that the researcher
was interested in evaluating the positive
As in any other research project, the in-
impact of an early discharge program,
vestigator now selects a design that will en-
they might respond in a way to support
able him or her to determine whether the
the researchers view.
behavioral objectives have been attained.
A variety of designs are available, depend-
ing on the purpose and nature of the proj- d. Analysis and Interpretation of Data
ect (see the Pluralistic Approaches to Eval-
uation Research section of this chapter). This involves taking apart the data that
The researcher must determine how the were collected and reorganizing it in a
behavioral objectives will be measured manner that enables the researcher to
within that design. Will observations of determine if the objectives have been
clients behaviors be measured? Will met. Quantitative data are analyzed using
clients complete self-administered ques- statistical procedures and qualitative
data analyzed using a variety of qualita- them. Others believe the real issues in
tive approaches. evaluation research should emerge from
intensive onsite knowledge rather than
3. Pluralistic Approaches to formulating them before data collection.
Evaluation Research They favor a more qualitative approach
that involves inductively constructing ex-
The approach selected for an evaluation planations about the data rather than de-
research project depends on the purpose veloping hypotheses in advance and then
of the evaluation. It is important to be collecting data to test them.
clear about the objectives of the evalua- The evaluation case study is a method
tion project from the beginning. As previ- that is gaining popularity. These studies
ously stated, one approach to evaluation are in-depth explorations of phenomena,
is the classic experiment (or RCT) or usually over an extended period of time,
methods approximating the classic ex- using diverse data collection procedures
periment such as a quasi-experimental to collect detailed information about an
design. Recently, however, the value of individual, group, organization, program,
the experiment has come under attack as or social phenomenon. The major advan-
a form of evaluation research in nursing. tage to using a case study in evaluation
Experiments have been criticized be- research is the depth of data and the in-
cause they cannot provide full explana- tensive analysis of the situation that can
tions of why an innovation or program is result. Case studies can use both quanti-
effective. This is the type of information tative and qualitative approaches to fo-
that is important to the practitioner who cus on specic situations. Most case
wishes to implement the successful inno- studies yield descriptive data, identify re-
vation in a new setting. There are, how- lationships among the variables, and
ever, different opinions on the value of track trends in the data. Case studies are
the experiment to evaluation research in especially helpful in providing descrip-
nursing. Some believe it is the sine qua tive information about the process of pro-
non of scientic research and they see gram or procedure implementation. A
the RCT as the only way to evaluate case study approach to evaluation could
health-care interventions (McDonald and address such questions as: How is the
Daly, cited in Parahoo, 1997). Proponents program or intervention working? What
of experimental research argue that al- is the impact of the program in the prac-
though a variety of designs can be used to tice setting? Is the program operating the
do evaluation research, the experiment, way it was intended? What variables are
despite its limitations, is the only one resources or barriers to successfully im-
that can establish cause and effect. plementing the program? What are the
Those who oppose the use of the RCT strengths and limitations of the program?
for evaluation research in nursing believe What are the intended and unintended
it incapable of contributing to the under- outcomes of the program or innovation?
standing of human experience. Many fem- Case studies yield specic, in-depth,
inist researchers have argued that women holistic ndings. The emphasis is on un-
have been victims of experimental evalua- derstanding the specic case ( in this
tion research that is reductionistic in case, program or innovation), but it is
nature. They see physicians imposing likely that the ndings may be useful in
medical interventions and treatments on similar situations in which the program
women that have been evaluated using ex- may be replicated.
perimental methods that ignore the needs In brief, a mix of designs for purposes
and experiences of the women receiving of evaluation research is desirable. Each
design contributes a different perspec- on a modied level III perinatal unit. The
tive to the evaluation process. It is im- study is of note for its simplicity and be-
portant to be clear about the overall pur- cause few studies have evaluated this
pose of the evaluation project so that the care delivery system despite its recent
best design is selected. If the purpose of adoption by hospitals without strong em-
the evaluation project is to determine pirical evidence to support the numerous
clients perceptions of the new program, benets reported in the literature.
a qualitative design is often recom-
mended. When the goal is to determine
the effectiveness of an intervention by
conrming a cause-and-effect relation- C. HEALTH PROMOTION
ship between the treatment and the out- RESEARCH
come, a randomized clinical trial is best.
And if the goal is to determine the many The concept of health promotion has
factors that contributed to successful im- evolved in nursing over the past several
plementation of a new program, a case decades and continues to be dened from
study may be the best choice. a variety of perspectives and disciplines.
Similarly, health promotion research is an
evolving and dynamic process. It is dif-
4. Examples of Evaluation cult to succinctly dene what health pro-
Research Studies motion research is. For purposes of this
discussion, health promotion is dened
Box 8.2 describes a research study using according to the denition proposed by
formative evaluation to examine the ef- Pender (1996, p. 77) as activities directed
fects of changing the nursing care delivery toward sustaining or increasing the level
system in the neonatal intensive care unit of wellness, self-actualization, and fulll-
(NICU) of a large Childrens Hospital in cen- ment. This denition may be directed to
tral Canada. The evaluation measured the the individual, family, group, or commu-
impact of primary nursing versus total pa- nity level because nurses interact on all of
tient care nursing on nurses perceptions these levels in terms of their health pro-
of their work environment; on physicians motion role. Pender makes the important
and allied health professionals percep- distinction between health promotion
tions of communication and collaboration and illness prevention. Illness prevention
with nurses; on the parents perceptions of is synonymous with health protection
communication and care provided; and on and focuses on measures to prevent dis-
the nurses direct and indirect care activi- ease or illness. Illness prevention is moti-
ties. This study illustrates the point that vated by a desire to actively avoid illness.
one of the major contributions of forma- Health promotion, on the other hand, is
tive evaluation is the ability to monitor the motivated by a desire to maximize health
effects of the change process and simulta- and well-being.
neously provide feedback that can be used Health promotion research is dened
to address areas of concern so that ongo- as systematic investigation into the
ing improvements can be made during the processes and substance of health pro-
change process. motion action. It encompasses a broad
Box 8.3 features an example of an eval- range of studies including much of the re-
uation research project that used a com- search in health policy, lifestyle, and so-
parative study to determine if combined cioepidemiology. Some authors include
maternalbaby care was better or as ef- studies of community participation and
fective as the traditional method of care the anthropology of health in the classi-
FORMATIVE EVALUATION: IMPLEMENTATION OF

PRIMARY NURSING
Method Results
The formative evaluation used in this study used Findings indicated that significant differences oc-
a before, during, and after implementation de- curred over time (P 0.05) on the work envi-
sign to monitor the effects of changing from a ronment scale for involvement (F 3.67; df
total patient care system to a primary nursing 2.60; P 0.031), autonomy (F 5.02; df 2.60;
system and to identify problems resulting from P .010), innovation (F 6.57; df2.60; P
the change. This type of design permits the 0.003), and physical comfort (F 3.97; df2.60;
change in nursing system to be examined with- P 0.024), indicating nurses decreased satisfac-
out the difficulties inherent in comparing across tion in these areas. At 6 months after implemen-
different organizational and structural environ- tation, results suggested an adaptation phase to
ments, unit staff, and competency levels. It also the new primary nursing system because there
helps to identify issues that need to be ad- was a decrease in the percentage of physicians
dressed early in the change process. who perceived that the nurses were available to
Baseline measures were taken before the discuss and were knowledgeable about their pa-
implementation of the primary nursing delivery tients. A total of 92 percent of the physicians
system, then repeated at 6 and 12 months after stated that communication was excellent or very
implementation. good and 100 percent rated collaboration as ex-
Thirty-one nurses completed the Work En- cellent or very good at the 1-year postimple-
vironment Scale at all three testings. The 12 mentation interview. Allied health professionals
NICU physicians and seven allied health profes- consistently reported over the three periods
sionals were interviewed for the three repeated that nurses did not recognize the need for their
measures of the communication and collabora- services. And there was a consistent decline in
tion scale. Thirty-six parents were interviewed. the overall rating of collaboration during the
The effects of change were measured on study. There was no appreciable change in the
nurses perceptions of their work environment amount of time that nurses spent in direct or in-
as measured by the Work Environment Scale; direct patient care activities. The nursing audit,
communication and collaboration with other including parent satisfaction, indicated that the
health-care professionals as measured by the quality of care had improved. Nursing audit
nursephysician communication and collabora- scores rose steadily over time with a 12-month
tion interview and the nurseallied health com- postimplementation compliance of 99 percent.
munication and collaboration interview; parent The major contribution of this study was to
satisfaction and quality of care provided as mea- enable all participants to monitor the effects of
sured by the Medicus Quality Assurance Audit the change process and to immediately address
Tool; and direct and indirect nursing activities areas of concern. In this change process, evalu-
as measured by random observations of nurses ation has been the key to the continuing success
in the NICU. of primary nursing in the NICU.
SOURCE: Summarized from Alcock, D., Lawrence, J., Goodman, J., and Ellis, J. (1993). Formative evaluation: Imple-
mentation of primary nursing. Canadian Journal of Nursing Research, 25(3), 1527.
cation of health promotion research (Mc- to a growing interest by nurses in health

Queen, 1994). The majority of health pro- promotion studies that focus on health
motion research studies in nursing in the policy formation and community partici-
1980s and early 1990s focused on lifestyle pation such as occurs in social action re-
with the individual as the unit of analysis. search. There is also a growing emphasis
Recently there has been a paradigm shift on the family and the community as the
COMBINED MOTHER AND BABY CARE: DOES IT MEET THE NEEDS

OF FAMILIES?
Combined mother and baby care is thought to mothers with combined care but not traditional
be an effective way to prepare a family for the care; the reverse was true of the multiparous
changing roles and added responsibilities that the mothers scores.
arrival of a new baby entails, but few studies have Factors that may have confounded the re-
evaluated this care delivery system. Therefore, sults include insufficient time between institu-
the postpartum staff at Sudbury General Hospi- tion of the program and its evaluation and the
tal in Sudbury, Ontario, conducted a posttest quality of the prenatal education received. Mul-
control group study design with a self-selected tiparous mothers scored higher on self-care, in-
sample of postpartum mothers when the unit fant care, and maternal competence than did
was changing from traditional to combined primiparous mothers regardless of the care de-
mother and baby care. A total of 103 mothers livery system. Future research should take the
who received traditional care and 102 who had differences between the primiparous and multi-
combined mother and baby care completed a parous mothers into account and focus on the
questionnaire to assess perceptions of their own less immediate needs of mothers. Although the
competence and satisfaction with the type of results of this study do not statistically support
care administered. There were no significant dif- the greater efficacy of combined care over tra-
ferences between the two groups, although ditional care in meeting the needs of families,
there was a positive trend in mothers compe- evidence suggests that further study may pro-
tence and satisfaction scores for the primiparous vide more conclusive results in this respect.
SOURCE: Summarized from Bailey, P., Maciejewski, J., and Koren, I. (1993). Combined mother and baby care: Does
it meet the needs of families? Canadian Journal of Nursing Research, 25(3), 2939.
unit of analysis and the unit of interven- A concern with the conditions for
tion for health promotion initiatives by change and the obstacles and barriers
nurses. This is in keeping with the ethos to change
of health promotion that asserts that peo- A focus on patterns for policy making
ple (e.g., individuals, families, and com- in health, patterns of behaviors, and
munities) know what they want and need the antecedents of health in the social
to be healthy. contexts of behavior
Emphasis on diverse methods, multi-
1. Characteristics disciplinary approaches, and collabo-
ration of different sectors of society
Although it is difcult and perhaps impos-
with different traditions
sible to reach consensus on a universal def-
inition of health promotion research, some
Use of multiple indicators (to infer ef-
fect of health promotion strategy) and
authors (Bunton and Macdonald, 1992; Mc-
statistical analyses appropriate to the
Queen, 1994; Pederson et al., 1994; Pender,
indicators
1996) have offered helpful parameters and
characteristics that health promotion re-
Application of ndings in the context
of action is central to health promo-
search ought to reect. These include:
tion research
Emphasis on individual, family, com- A tool that helps the community to
munity, environmental, or societal communicate with policy makers and
wellness vice versa
Involvement of the community of in- vention. These studies, for the most part,
terest in the research aim to describe health promotion prac-
A set of end users that include policy tices of individuals, identify determinants
makers, decision makers, health-care of healthy lifestyle or predictors of health
professionals, communities, and indi- behavior, assess lifestyle practices, im-
viduals plement and evaluate health promotion
interventions delivered to individuals, or
Although some health promotion re- develop instruments to measure health
searchers may add other characteristics promotion. They are important in terms
to this list, most would agree that this rep- of developing the knowledge base for
resents a helpful, albeit incomplete, list nursing and use a variety of approaches
reective of key features of health promo- to address the research question. Sur-
tion research. veys are particularly popular in studies of
this nature. Box 8.4 provides an example
of a prospective survey that focuses on
2. Scope of Health Promotion the individual as the unit of analysis. The
Research study identies factors that predict phys-
ical activity behavior in rural fth grade
It is important that nurses appreciate the
children. The study is of importance to
complexity of health promotion and the
the eld of health promotion research be-
various levels at which it occurs. Health
cause understanding the factors that in-
promotion research within nursing has
uence physical activity behavior is im-
called for a move from the individualistic
portant in the design of intervention
approach that predominates in health
programs targeted at youth. The trend to-
behavior research and health care to a
ward increasing physical inactivity in
societal approach in which the responsi-
North American youth is one of the most
bility is shifted to society for making
pressing issues facing health profession-
available the resources that people re-
als today.
quire for health. Certainly the 1980s and
1990s witnessed the theme of a commu-
nity versus individual focus in discus- b. Family
sions of health promotion in nursing
Although the family plays an important
(Gott and OBrien, 1990). This theme also
role in the development of health behav-
needs to be reected in the eld of nurs-
iors, there is very little nursing research on
ings health promotion research agenda.
the role of the family in health promotion
Given the elds emphasis on holism and
(Pender, 1996). The family is often consid-
our acceptance of the environment as a
ered to be the context within which health
core variable integral to health, there re-
behavior is learned, but it is actually much
mains a need to broaden the focus of
more than that. The family is a unit of
health promotion research in nursing to
analysis in its own right. Progress in this
include not only the individual and family
area of health promotion research has
but also the community and societal
been slow because of methodological im-
level. Let us consider each of these levels.
precision and measurement problems
when the family is considered the unit of
a. Individual analysis or the unit of intervention. Longi-
tudinal surveys that track family lifestyles
A large number of health promotion stud- and their stability over time, use of small
ies in nursing focus on the individual as focus groups within the family environ-
the unit of analysis or the unit of inter- ment, and case studies are approaches
BOX 8.4 A Prospective Study poorly understood. It is an important area

of Physical Activity of investigation for nurses, but studies of
in Children this nature are scarce in the literature.
Gillis (1994) studied the relationship of
Methods mothers and fathers health promoting
A prospective study design was used to lifestyles to adolescent daughters life-
identify the predictors of vigorous physical styles using a survey design. The health
activity (VPA) and moderate and vigorous promoting lifestyle prole (HPLP) was
physical activity (MVPA) among 202 rural, used to measure lifestyle. Results indi-
predominately African-American children. cated that both mothers and fathers life-
Selected social-cognitive determinants of styles were positively correlated with
physical activity were assessed via ques- their daughters lifestyles. This suggests
tionnaire in the fifth grade. Participation in
there is some consistency among family
VPA and MVPA was assessed via the previ-
health attitudes, behaviors, and practices.
ous day physical activity recall 1 year later
in the sixth grade.
Results c. Community
For girls, participation in community sports,
self-efficacy in overcoming barriers, enjoy- Community-based health promotion is an
ment of school physical education, race exciting area of investigation for nurses.
(white black), and perception of mothers The idea of changing the behavior of an
activity level (active vs. inactive) were signif- entire community rather than a family,
icant predictors of VPA. For MVPA, partici- small group, or individual is a challenge
pation in community sports and self-efficacy that requires effort and coordination of all
in overcoming barriers were significant bar- involved. It includes a range of activities
riers. For boys, whereas self-efficacy in over-
aimed at creating a health-enhancing en-
coming barriers was the only significant pre-
vironment and healthy behaviors for pop-
dictor of VPA, beliefs regarding activity
outcomes and participation in community ulations or entire communities (Pender,
sports were significant predictors of MVPA. 1996). Research into community-based
health promotion offers nursing a wealth
Conclusion
of opportunities to be involved in aspects
Social-cognitive constructs such as physical of needs assessments, intervention stud-
activity, self-efficacy, access to community
ies, evaluation research, participatory ac-
physical activity outlets, and positive beliefs
tion research, and methodological re-
regarding physical activity outcomes are
important predictors of future physical ac- search that focuses on the development
tivity behavior among rural, predominantly of new methodologies and measurement
African-American children. techniques. Research into community-
based health promotion initiatives needs
SOURCE: Trost, S., Pate, R., Saunders, R., et al. to be carefully examined to determine
(1997). A prospective study of the determinants of
physical activity in rural fifth-grade children. Preven-
valuable information about what works
tive Medicine, 26, 257263. Cited with permission. and what does not work with various pop-
(From Am J Health Promotion, 1998, 12(4)) ulations and why. These are important
questions for nurses to answer.
The best examples of research into
community-based health promotion are
provided by the Healthy Cities initiative
that may be considered in the home con- in Canada, the United States, and Europe.
text. The World Health Organization (WHO)
The relationship between family health launched the Healthy Cities Initiative in
and individual health is multivariate and 1986 as a strategy for achieving health for
all. It focused on developing healthy pub- ment education program for seniors; or
lic policy, creating supportive environ- implement and evaluate a school-based
ments, strengthening community action, health promotion program on healthy
developing personal skills of community diet. The point is that different types of re-
citizens, and reorienting health-care ser- search objectives and questions lead to
vices. Interested readers are directed to very different study designs. Surveys,
the work of Flynn (1992) for a U.S. per- interviews, participant observation stud-
spective and Manson-Singer (1994) for a ies, focus groups, and experimental de-
Canadian perspective on these projects. signs all have something to contribute to
health promotion research. Qualitative
and quantitative methodologies are ap-
3. Methods propriate for different types of studies. Re-
gardless of your position in the methods
An interesting debate has continued in debate, you must be able to clearly justify
the literature for some time over methods your choice of design and method.
for health promotion research. Succinctly
stated, the debate is over the use of tradi-
tional public health measures versus 4. Future Directions
more nontraditional methods. There are
Health promotion research has devel-
many elements to the debate. One is the
oped tremendously over the past few
quantitative versus qualitative issue dis-
years, from a small area of interest pur-
cussed in Chapter 1. Another is the rigid
sued by a few investigators and practi-
adherence to research criteria estab-
tioners to a wider set of activities under-
lished by statisticians such as the sacred
taken by federal, provincial, and state
cows of randomness and signicance,
health departments as well as university
even though the likelihood that actual re-
centers that specialize in health promo-
search conducted in health promotion
tion research. Health promotion efforts
will be in violation of the statistical as-
are increasingly being established in mul-
sumptions underlying classical paramet-
tiple and diverse settings, and new direc-
ric procedures (see Chapter 12). A central
tions in research are being forged, such
challenge to health promotion research is
as those proposed by Pender (1996). Fu-
how to break free legitimately from a rigid
ture directions include:
orthodoxy that is increasingly inappro-
priate and inadequate for the special ar- 1. Identifying health promotion beliefs
eas of study needed (McQueen, 1994). and practices in diverse families and
We support the use of a multiplicity of communities as a basis for effective
methods for health promotion research. programming
The research questions and objectives 2. Developing appropriate indicators of
should guide the design selected. Because change (consequences) at the commu-
the range of questions is so broad in nity level that reect the effectiveness of
health promotion research, a rich reper- health promotion interventions (treat-
toire of approaches is required. For exam- ment)
ple, you may wish to describe the charac- 3. Developing consistent methods for as-
teristics of a target population such as the sessing health outcomes across a range
needs for peer support of homeless youth; of programs and communities
or explain an outcome from a health pro- 4. Determining the synergistic effect of
motion intervention such as changes in work site, school, family, and commu-
knowledge, attitudes, and health behav- nity health promotion efforts on popu-
iors after a community-based empower- lation health outcomes
5. Identifying the characteristics of health Feminist research has been dened in

strengthening environments in families, many ways. As the context and shape of
schools, work sites, and communities feminism is shifting, so is the manner in
6. Tracking the outcomes of families and which feminist research is dened. This
communities receiving care from multi- makes the task of dening feminist re-
service nursing centers with major em- search perspectives both challenging and
phasis on health promotion and dis- exciting. There are many different faces of
ease prevention feminism. In the early beginnings of the
second phase of the womens movement
In summary, the future looks bright for
(1960s onward), one could easily classify
health promotion research in nursing.
feminist researchers in terms of their po-
The eld of nursing is forging new fron-
litical views (e.g., liberal, radical, or Marx-
tiers in this area. The good news is that
ist), their preferred research styles, or
health promotion research is still new
their academic discipline; however, these
enough for all of us to be considered pio-
distinctions have become blurred. Politi-
neers. The multiple denitions and ap-
cal orientations are now blurred by inter-
proaches to health promotion indicate a
nal divisions in feminist thought; many
high level of interest in it. This is healthy
academic researchers are borrowing from
for nursing and for health care. As nurses
other disciplines, and many are mixing
broaden the scope of their research ac-
quantitative and qualitative approaches
tivities to include individual, family, and
or developing new methods to explore
community health promotion projects,
questions of interest (Olesen, 1994). One
their inuence in the social, political, and
cannot speak of feminism; one can only
economic arena will also increase, allow-
speak of feminisms and dene them by
ing them to impact factors inuencing hu-
their adherence to four basic principles:
man health.
1. A valuing of women and their experi-
ences, ideas, and needs
2. A seeing of phenomena from the per-
D. FEMINIST RESEARCH
spective of women
3. A recognition of the existence of con-
Feminist research has gained increasing
ditions which oppress women
momentum in nursing over the past de-
4. A desire to change these conditions
cade. Many argue that because nursing is
through criticisms and political action
a female-dominated profession that ad-
vocates for women and children, as well Feminist research perspectives use a
as men, the relationship between nursing variety of research methods in their
and feminism should be explored. The work. Many share the assumptions held
patriarchal society in which we live is a by interpretive (qualitative) researchers
part of the eld of nursings reality as a and those who adopt a critical perspec-
professional discipline, as a science, and tive. As Reinharz (1992) noted, most fem-
as an art. Many nurses are now focused inist scholars use a variety of research
on the conditions of oppression that ex- methods. They value inclusiveness more
ist for nurses and their clients and have a than orthodoxy and allow for creativity in
desire to address the social, class, gen- all aspects of the research process.
der, ethnic, racial, and other biases that Although feminist researchers use a
are inherent in their world and in their re- wide range of quantitative and qualitative
search. Feminist research perspectives methods, most emphasis seems to be on
offer one possibility for nursing to im- qualitative methods, particularly using in-
prove its science in this area. depth interviews. Oral histories, compar-
ative studies, and eld study methods are some authors in terms of its characteris-
also common. Similar to most interpretive tics. Duffy and Hedin, cited in Bunting and
perspectives, feminist research stresses Campbell (1994), provide a useful list of
personal experience embedded in the qualities that they attribute to studies
lives of the participants and subjectivity. conducted from a feminist perspective.
Feminist researchers do not stop here, These include:
however; they go on and ask the simple
but important question, And what about 1. A feminist consciousness rooted in an
the women? Where are the women in any attitude of equality, thereby replacing
situation being investigated? Why are hierarchies of traditional research me-
they not present in many situations? If thods with horizontal relationships. In
they are present, what exactly are they these relationships, the participants in
doing? How do they experience the situa- a study are seen as partners in the re-
tion? What do they contribute to it? What search process. The investigator works
does it mean for them? (Ritzer, 1988). in partnership with the participants to
Harding (1987), a feminist philosopher rene the research question, conduct
of science, adds clarity to the discussion the research, validate the ndings, and
surrounding denitions of feminist re- disperse the results.
search by providing some simple criteria. 2. The aim of feminist research is to in-
She takes the stance that in dening femi- clude women and study phenomena of
nist research, three aspects of the study practical concern to women. The focus
should be considered: (1) the purpose of is always based on womens experi-
the inquiry; (2) the explanatory hypothe- ences and the validity of womens per-
sis; and (3) the relationship between infor- ceptions as the truth for them. In sim-
mant or subject and the researcher. ple terms, the research should be for
These three criteria taken together should women.
distinguish feminist research from non- 3. Feminist perspectives involve con-
feminist research, rather than the methods scious partiality in which the biases
used to gather data. The purpose of femi- of the researcher are acknowledged
nist research is to create social change and openly discussed. The researcher,
that will benet women. This is core to any as in other qualitative approaches, ex-
denition of feminist research. Feminist re- amines her own values, beliefs, as-
search may also benet other sectors of sumptions, and motivations. The re-
society such as children, men, homosexu- searchers point of view and biases
als, oppressed groups, and others, but it become part of the data. This ensures
must have value for women. Feminist re- that the researcher is on a level play-
searchers use a variety of methods to col- ing eld with the participants.
lect data. These data are always analyzed 4. Feminist research uses a variety of per-
within the context of womens lives and are spectives and a multiplicity of meth-
done in such a way that women are em- ods, including both qualitative and
powered rather than portrayed in ways quantitative methods. This reects the
that stereotype them or reinforce tradi- complexity of womens lives that are
tional stereotypes. The relationship be- sufciently complex and diverse to re-
tween the researcher and the participants quire multiple perspectives.
is a horizontal one in which the participant 5. Feminist research aims to create social
is a partner in the research process. The change. In this way, it can be considered
participant is a legitimate knower of their a subset of critical social theory (see
experience and is an expert. Chapter 1). It is action and change ori-
Feminist research has been dened by ented and argues that science should be
used to improve the conditions of op- maintained through the transmission of an

pressed people. ideology justifying male domination (Saun-
6. Feminist research strives to represent ders, 1988, pp. 159160).
human diversity. It includes ethnicity, Feminist perspectives stress the idea
class, gender, sexual orientation, and that inequalities suffered by women stem
culture in the designing, conducting, from an ingrained, systemic patriarchy in
and interpreting of research. These societies; this patriarchy pervades many
are often blind spots in traditional re- institutions and practices, including, in
search approaches. particular, family socialization. Similar to
7. Feminist research critiques before researchers who subscribe to one or
scholarship, especially looking for another of the critical perspectives, femi-
drocentric (male bias) and ethnocen- nists see science as a tool that has largely
tric (race and ethnic) biases. been in the hands of male oppressors.
Feminists take a strong advocacy posi-
1. Philosophical Tenets tion in an effort to eliminate sexism from
society.
A philosophical foundation and a dened Feminists display a tendency toward
methodology have not yet been broadly synthesizing various critical concepts,
accepted by feminist scholars; however, symbolic interactionist ideas, and qualita-
many believe it has emerged from critical tive research styles. Similar to critical the-
social theory. Both feminist and critical orists, feminists reject the relativistic
approaches believe that knowledge is so- stances of positivism and the interpretive
cially constructed and acknowledge the approaches. This stance can be explained
oppressive nature of social structures. by their interest in achieving gender
Both focus on the emancipatory goals of equality. Instead, they argue that science
the research, but they differ in that gender should be used to improve the conditions
is not the central concern of critical theo- of the oppressed. Feminists wish to help
rists. Critical theorists emphasize ratio- eliminate sexism by understanding and
nality, write principally for other intellec- documenting its sources. An emphasis on
tuals, and maintain power inequalities achieving gender equality is central. Box
within their research teams. Feminist re- 8.5 summarizes some of the major propo-
searchers place gender centrally within sitions of feminist perspectives.
the research, make feminist writing acces- Feminist philosophers believe all wo-
sible to all, and promote equality within men are legitimate knowers and that
their research team (Webb, 1993). those experiencing particular complexi-
Feminist researchers claim that one ties are the most knowledgeable about
cannot adequately understand human so- that experience. The world of women
cieties without paying attention to the uni- deals with the particular, with the con-
versal role of patriarchy, which refers to crete experiences of child rearing, moth-
the domination of social groups by men ering, parenting, neighborhood interac-
who have greater power and privilege than tions, and so on. Knowledge of that world
women and children. Some feminist re- should start from the experience of wo-
searchers have argued that inequalities of men. Similar to phenomenological as-
the genders emerged because of males sumptions, feminist philosophers believe
greater physical strength. The control this that self-reection, self-awareness, and
enabled men to exercise resulted in wo- purity of language constitutes human na-
men playing socially subservient roles and ture. If this is true, men experience the
in stereotypes that portrayed women as in- world differently in the everyday experi-
ferior to men. Sexism is thus fostered and ence of friendship, love, sexuality, moral-
BOX 8.5 Key Feminist Ideas important distinction between method (a

technique for collecting evidence) and
In major cultural institutions (e.g., universi- methodology (how the methods should be
ties, media, advertising, the writing of his- used) in feminist research. She writes that
tory), mens standpoints are represented as what is important in feminist research is a
universal (Smith, 1987:19). methodology that is consistent with femi-
As one moves from elementary through nist epistemology (theory of knowledge).
to university educational institutions, the This includes a methodology that:
proportion of women on the staff declines;
the proportion of women in administrative 1. Places value on womens subjective ex-
positions declines even more as one moves periences. Feminist methodology val-
through the hierarchy of educational insti- ues women as knowers and seeks to
tutions. reect with precision womens experi-
Women have been systematically ex- ences as they view them.
cluded from the making of cultural ideol-
2. Values the importance of context in
ogy. An adequate social science must be
womens lives. Womens experiences
grounded in the everyday experiences of
both men and women. are embedded in the context of their
An adequate social science must recog- everyday lives. It is important that the
nize the universal role of patriarchythe method chosen provides a rich de-
domination of society by males who have scription of context. This should in-
greater power and privilege than women clude a description of the social and po-
and children. litical factors as well as other variables
that may inuence the phenomenon.
SOURCE: Dorothy E. Smith, The Everyday World as
Problematic: A Feminist Sociology. Toronto: Univer-
3. Values the relationship of the re-
sity of Toronto Press, 1987. searcher and the participants. This re-
lationship is characterized by respect
for the participants. They are viewed as
equal partners in the research process,
but there is considerable debate as
ity, intellect, conict, and challenge than to whether equal power can ever be
do females. Consistent with this point of achieved in the research relationship.
view, feminist researchers call for re- 4. Values the inclusion of women from di-
search methods that reect their episte- verse social and ethnic backgrounds in
mological beliefs, emphasize concrete ex- their samples so that perspectives of
periences, and stress the particular that all women can be exposed and under-
comprises the world of women (Rothe, stood.
1994).
a. Developing the Research Question
2. Application of Method and
Methodology Research questions appropriate for a
feminist inquiry are those that are of in-
Feminist research is not a method but terest and concern to women. The pur-
rather a multiplicity of methods. Feminist pose of feminist research is to create so-
researchers select a method that is ap- cial change that will enhance the lives of
propriate to answer the research ques- women. The questions will vary in focus
tion. Over time, feminist researchers have and form but they will be distinctive in
come to a reasoned examination of how that (1) the issues are of primary concern
all methods can be used to answer femi- to women, (2) the question requires
nist questions. Harding (1991) makes an women to report their experiences using
their own voices, and (3) the question al- with those researched, who then add their
lows for a structural analysis of the con- opinions to the research (Klein, 1983).
ditions of womens lives and should lead Oakley (1981) produced a landmark piece
to an improvement of it. Experiences of feminist writing that showed the reci-
such as accessing the health-care sys- procity that exists between a feminist re-
tem, womens health, constructing rela- searcher and her participants. In her
tionships, giving birth, mothering, rape, study, Oakley asked questions of the par-
incest, infertility, poverty, and family vio- ticipants, and they, in turn, asked ques-
lence are topics from which questions, tions of her. She found it impossible not to
appropriate to feminist methods of in- give her own views, answer participants
quiry, could be developed. Hartrick questions, and give advice based on her
(1997), a feminist researcher, used ele- own experience as a woman. Oakleys be-
ments of phenomenological research and havior reects the role of the researcher in
feminist inquiry to examine the experi- feminist inquiry.
ence of self-denition for women who are As with other qualitative approaches,
mothers. Specically, the research ques- reexivity is expected of the researcher.
tion she asked was, What is the experi- Reexivity is dened as the critical think-
ence of dening self for women who are ing required to examine the interaction be-
mothers? tween the researcher and the data occur-
ring during analyses. The researcher
b. Role of the Researcher explores personal feelings that may inu-
ence the study and integrates this under-
One of the dening characteristics of fem- standing into the study. The researcher
inist research is the nonhierarchical rela- needs to be reexive about her views so
tionships between the researcher and the that she can uncover deep-seated views on
participants. Feminist researchers strive issues related to the research and provide
for horizontal relationships among mem- a full account of her views, thinking, and
bers on the research team and partici- conduct. This is necessary so that readers
pants. The researcher views herself as a of the research report are aware of how the
partner with the participants and involves researchers values, assumptions, and mo-
them in the generation of the research tivations may have inuenced the frame-
question, the conduct of the research, and work, literature review, design, sampling,
the dissemination of results. Although data collection, and interpretation of nd-
there is considerable debate as to whether ings. Making explicit the participation of
the researcher and the participant ever the researcher in the generation of knowl-
experience an equal power relationship in edge adds to the relevance and accuracy of
the inquiry process (Webb, 1993; Ford- the results (Webb, 1993).
Gilboe and Campbell, 1996), there is
agreement that the role of the researcher c. Sample
in feminist inquiry is a vulnerable one be-
cause the investigator shares her experi- Sample selection involves choosing
ences and emotions with the participant. women who have experienced the phe-
Feminist researchers promote intersub- nomenon of interest and are able and
jectivity and interaction in their research willing to share their perspectives. Femi-
rather than the one-way communication of nist researchers embrace the diversity of
traditional methods. To do this, the re- womens lives and experiences and
searcher constantly compares her work therefore attempt to include women of di-
with her experience as a woman and a sci- verse social class, race, ethnic groups,
entist. The researcher then shares this cultures, and so on in their sample selec-
tion. Feminist scholars are critical of the 2. Understand the experience of women
ethnocentric bias in the sampling proce- from their point of view and correct a
dures of many nursing investigations that bias of traditional participant observa-
fail to report race or ethnicity and make tion that trivializes female activities
the assumption that a white, Anglo-Saxon and thoughts or interprets them from
perspective of the world is appropriate the male researchers perspective.
and universally experienced (Ford-Gilboe 3. Document the lives of women to en-
and Campbell, 1996). able the participant observer to see
women as full members of their social,
economic, and political worlds.
d. Data Collection
Feminist participant observation gener-
Feminist researchers use a variety of ally values intimacy and empathy. The re-
methods to collect data. Oral history in- lationship that emerges between the re-
terviewing is a preferred form of data col- searcher and those observed is often a
lection. It is a participant-guided investi- nurturing one. Unlike male observers who
gation of a lived experience in which few maintain a respectful distance between
prepared questions are asked (Sande- the researcher and subject, women inter-
lowski and Pollack, 1986). This method is act differently in the eld. They often act
based on the assumption that meaning on a nurturing impulse, which is recip-
comes from life stories expressed by rocated (Hochschild, 1983). The implica-
women in their own way, without the use tion of this point of view is that with topics
of structured questions. unique to women, only other women can
Multiple in-depth interviews are often understand their meaning (Rothe, 1994).
used to collect rich narrative stories. Be- Feminist researchers use techniques
cause of the close relationship that devel- such as focus group interviews; structured
ops between the researcher and partici- and semistructured interview guides; and
pants, feminist researchers believe that documents such as diaries, letters, jour-
multiple interviews develop trust among nals, photographs, and historical and med-
the partners. The researcher shares the ical records, to name a few. They also use
transcripts with the participants and in- quantitative instruments such as question-
vites them to contribute to the analysis naires and indexes. Hartrick (1997) used a
and interpretation process. This process variety of data collection methods in her
is believed to contribute to more accurate feminist investigation of the experience of
and sensitive descriptions of feelings, self-denition of self as mother. The pri-
emotions, thoughts, and processes as mary means was multiple interviews with
they unfold (Rothe, 1994). participants that were audiotaped and
Feminists who focus on gender issues lasted from 1 to 2.5 hours in length; in ad-
in female homogenous and heteroge- dition, focus group interviews were used
neous settings conduct participant ob- to compare and contrast the experiences
servation within feminist inquiry. It in- of women, and letters and writings by one
cludes the study of womens private of the participants that expressed her ex-
domains, workplaces, and organizations. periences were collected.
According to Reinharz (1992), it is con-
ducted to achieve the following goals:
e. Data Analysis
1. View womens behavior as an expres-
sion of social contexts that shows their The process of analysis is similar to the
behaviors as being shaped by social quantitative and qualitative procedures
context rather than context free. in general. Therefore, if a feminist re-
searcher is using phenomenology to ex- ports are distinctive in that they portray
plore womens lived experiences or womens voices, provide a structural
grounded theory to develop a theory of analysis of the conditions of their lives,
empowerment for women, the methods and include in the analysis the role and
of analysis previously discussed for inuence of the researchers themselves.
these perspectives would be followed. It is as important to feminists to have
The main distinction is that the content is their ndings published in popular wom-
likely to be analyzed in terms of artifacts ens magazines as in the most prestigious
produced by women (or men), about academic journals. Many feminist pro-
women, for women. The production and jects are published in book form as pop-
perpetuation of patriarchy and ethnocen- ular literature. This presents a bit of a
tric bias are major research themes ex- dilemma for nurse researchers who align
amined in the analyses. Feminists have themselves with this tradition. Political
been vigilant in analyzing texts for the and prestige issues related to research
presence of gender stereotyping that lim- in academic settings require nurse re-
its the opportunities of women, restricts searchers to modify their writing so that
their autonomy, and inhibits female de- they are acceptable to the academic com-
velopment (Im, 2000; Rothe, 1994). munity that judges the value of knowl-
Many feminist researchers invite par- edge produced through research. Some
ticipants to be active partners in the data feminist researchers deal with this dil-
analysis process. Others have partici- emma by publishing two different ver-
pants recycle the analyses and then re- sions of their report; versions reecting
ne them according to the reactions of the voices of participants are made avail-
participants (Webb, 1993). Not all feminist able to participants, but more traditional
researchers agree with this latter point. forms of academic writing may be used
Some believe the researcher has access to for submissions to peer-reviewed jour-
additional perspectives beyond the imme- nals. Such modications in style increase
diate research, and in order to carry out a the likelihood that results will be pub-
structural analysis of the research, as op- lished in a wide variety of forums.
posed to only reporting womens experi-
ences using their words, the researcher is
required to go beyond the immediate data 3. Example of Feminist Research
(Ribbens, 1989). Ribbens believes the fem-
inist researcher may interpret data differ- An account of a feminist inquiry is pro-
ently than participants, particularly if the vided in Box 8.6. The following comments
participants do not see themselves as fem- are offered about the study in Box 8.6:
inists. When this occurs, researchers This project drew on elements of inter-
must take responsibility for the decisions pretive phenomenology and feminist in-
they make and acknowledge this type of quiry in developing a methodology to ex-
power as a paradox in feminist research. plore the experience of self-denition for
women who are mothers. Principles of
feminist inquiry promoted sensitivity to
f. Reporting on the Findings the cultural aspects inuencing womens
development and provided guidance
Reports of feminist inquiry use descrip- about essential aspects in research with
tive, nonsexist language to portray the women. Interpretive phenomenology fa-
everyday lives of women. They are user- cilitated illumination of the lived experi-
friendly reports that make them accessi- ence of mothers with consideration of the
ble to all, not just intellectuals. The re- complexity of that experience.
WOMEN WHO ARE MOTHERS: THE EXPERIENCE OF

DEFINING SELF
Combining elements of feminist inquiry and phe- periential knowledge of self-definition, (3) par-
nomenology, this study examined the experience ticipants had the ability to be introspective
of self-definition for women who are mothers. about themselves within the context of their
Feminist poststructuralists maintain that the self social roles and relationships.
develops through the process of social interac- Data collection included in-depth multiple
tion and is created and reconstituted through interviews, focus group interviews, and letters
various discursive practices in which people par- and writings of one participant. The interview
ticipate. For mothers, Western culture provides process included three phases: establishing rap-
conflicting discourses. Discourses around healthy port, engaging in open-ended dialogue for the
self-definition stress autonomy, differentiation, purpose of gathering data, and successive data
and separation of the individual from others gathering in the second set of interviews to en-
throughout the life cycle. Discourses around able a more complete illumination of the par-
mothering emphasize womens abnegation of self, ticipants experiences.
selflessness, and self-sacrifice. Hence, women are Thematic analysis was used to analyze data
exposed to conflicting and confusing messages. from the first and second round of interviews.
The purpose of this study was to explore and Focus group data was used to compare and
describe the experience of self-definition for contrast themes and experiences of women.
women who are mothers. Specifically, it ex- The results revealed three elements within the
plored the question: What is the experience of mothers process of defining self, including re-
defining self for women who are mothers? Sub- flective doing, living in the shadows, and re-
questions the study examined included: How do claiming and discovering self. Within each of
women who are mothers define and express these elements, a number of other themes
self when they are influenced by and participate were described. Women described an integral
in conflicting discourses? Does the edict for self- relationship between the process of defining
sacrifice and abnegation of self hinder a womans self and their experience of health.
process of defining and expressing self? If the In describing their process of authoring and
process of self-definition is hindered through defining self, the women in this study, in essence,
the conflicting discourses, what influence does also described their experience of health promo-
this have on mothers overall health experience? tion and empowerment. As the women moved
Seven participants were recruited using pur- through the process of self-definition, they de-
posive sampling. A flyer was used to advertise scribed becoming more aware of the detrimental
the study to women who frequented various effect living out roles was having on their health
agencies and centers. A contact person was and well-being. Through the process of self-
identified in each agency to discuss the research definition, they became active participants, expe-
project with women and increase awareness. riencing a sense of empowerment in their life sit-
Selection criteria included: (1) participants uations and experiences. Health professionals
were mothers who had children between the may want to contemplate how they could sup-
ages of 3 and 16 years, (2) participants had export this health promoting, reauthoring process.
SOURCE: Summarized from Hartrick, G.A. (1997). Women who are mothers: The experience of defining self.
Health Care For Women International, 18, 263277.
The sample included women from a focused on the experiences of women in

range of social classes, including the their everyday lives as they sought to de-
working poor to the upper crust of so- ne their sense of self.
ciety. The contexts of the participants The data collection reected the
lives were described in detail. The study researcherparticipant partnership model
of feminist research. The interviews were with complex multivariate analyses. Field
open ended and conversational in nature studies strengths are in the area of valid-
with the researcher following the partici- ity, cost (in some cases), and in probing
pants lead. The researcher simultane- for depth. Applied research is important
ously listened for and explored central as a means of bringing about change and
themes and meaningful structures within improvements in practice and empower-
the womens experiences. The women ing subjects through participation. The
shared in the identication of themes and practical problem-solving nature of action
in the validation of data analysis. research makes it particularly appealing
Feminist inquiry has empowering po- to a nurse who has identied a problem in
tential for women. Through this investi- practice and sees the merit of investigat-
gation, the women continually alluded to ing it and improving practice. Evaluation
the complementary relationship between research leads to decisions about the ef-
their experience of self-denition and fectiveness of current practice. This type
their overall experience of health. As the of research is important because of the
women moved toward discovery, they utilitarian nature of its results and the sig-
described feeling more in control of their nicant role it plays as the cornerstone of
well being and were beginning to make policy research at the local and national
conscious choices to nurture and care for level. Health promotion studies provide
themselves. It appears that the process an opportunity for the eld of nursing to
of self-denition was a health promoting soar like an eagle in redesigning health
process whereby women became the au- care and creating new models and sys-
thors of their own life experiences. For tems for promoting the health of individu-
women in this study, the promotion of als, families, and communities.
self-denition and self-expression simul-
taneously promoted health. This feminist
inquiry revealed how through the pro-
cess of self-denition, they became active F. POSTSCRIPT TO PART 2
participants, experiencing a sense of em-
powerment in their lives. Although the most common designs used
in nursing and health-care research have
been reviewed in Chapters 4 through 8,
E. ADVANTAGES AND others have been omitted. A discussion
LIMITATIONS OF of methodological research (e.g., con-
CONTEMPORARY APPLIED trolled investigations of obtaining, orga-
APPROACHES nizing, and analyzing data) is not pre-
sented in this book. Case studies are
Table 8.1 summarizes the advantages and discussed in terms of evaluation re-
limitations of the various research designs search. Various kinds of intervention,
we have presented throughout Part 2 of polls, and special interest approaches to
this book. Although it is not possible to research have also not been included.
generalize about all the studies contained As you get ready to design a study,
within any one type of design, we can con- keep in mind that you should choose a
clude that experimental studies have an design that is appropriate to the research
advantage over other research designs question you are posing. On occasion,
when we need to make clear causal infer- more than one type of design can be used
ences. Surveys are particularly adept at to answer a research question. The ques-
representing populations with samples, tion is the key concern. Try to decide pre-
and such studies have become associated cisely what it is you want to accomplish

Pre-experimental
Experimental
Quasi-experimental
Survey Designs
Group administered
Phone survey
Interview
Secondary data
Meta-analysis
Qualitative Designs
Phenomenology
Grounded theory
Ethnographic
Feminist
Field Designs
Participant observation
In-depth interviews
Field experiments
Naturalistic observational
Goal-Directed Designs
Action research
Evaluation research
Health promotion research
Feminist research
and then gure out the design that is best ers. We would do well to recognize the
suited to your needs. gulf between these traditions and try to
As a postscript to Part 2, Box 8.7 pre- build bridges between them. Both have
sents Kathleen OConnells wry commen- enormous contributions to make to the
tary on the dynamic tension between development of nursing knowledge in the
qualitative and quantitative practition- 21st century.
E X E R C I S E S
1. Choose one of the following studies A nursing researcher determine if
and identify the data sources that you there is a relationship between
would use. Write a proposal for how physician status and the speed with
you will proceed with the analysis. which patients are admitted to hos-
How might: pital. Would it be possible to con-
A nurse scientist study the relation- duct such a study using secondary
ship between the percentage of peo- data analysis techniques?
ple with skin cancers and the per-
centage of immigrants in a regional
health district?
2. Rates of school dropout, motor vehicle
accidents, homelessness, marriage
breakdown, depression, self-esteem,
abortion, or adolescent pregnancy. Us-
A health economist compare trends
ing one of these concepts, identify
in American and Canadian purchase
problems that may occur in attempt-
of health insurance?
ing to compare their values in the
United States with those in Canada.
Can you recommend ways of minimiz-
ing the problems of using them in com-
A nursing student analyze the sex-
parative studies?
ual stereotyping of nurses found in
popular literature sources?
3. You are a school health nurse inter-

ested in evaluating the effectiveness of a
A health promotion nurse explore violence prevention curriculum among
how local television news struc- children in elementary school. How
tures the public and policy debate would you proceed to conduct an eval-
on youth violence? uation study if you were doing a sum-
mative evaluation? In what way would
the process be different if formative
evaluation was required?
A feminist researcher determine
whether American magazines reect
standards of attractiveness for wo-
4. Outline the steps that you would fol-
men similar to those found in compa-
low as a nurse researcher to help a
rable magazines in Canada, Britain,
community identify and address their
and France. How could the re-
health priorities, help citizens develop
searcher determine that the identi-
a plan of action, and evaluate both the
ed relationship is always the same?
process and impact of any interven-
tions. Of the research designs dis- 5. What unit of analysis might be appro-
cussed in this chapter, which one is priate for studying intergenerational
best suited to this assignment? Justify transmission of family health beliefs
your response. and practices?
Hart, E., and Bond, M. (1998). Action Research Klostermann, B., Perry, C. and Britto, M.
for Health and Social Care. Philadelphia: (2000). Quality improvement in a school
Open University Press. A succinct yet com- health program. Evaluation and the Health
prehensive handbook on action research Professions, 23(1), 91106. This paper re-
for professionals and researchers who are ports an outcome and process evaluation
interested in using research to improve of a school health program.
practice. Morris, J., Penrod, J., and Hupcey, J. (2000).
Hilton, A., Thompson, R., and Moore-Dempsey, Qualitative outcome analysis: Evaluating
L. (2000). Evaluation of the AIDS prevention nursing interventions for complex clinical
street nurse program: One step at a time. phenomena. Journal of Nursing Scholarship,
Canadian Journal of Nursing Research, 32(1), 32(2), 125130. This paper presents a par-
1738. A good example of an evaluation re- ticular method for evaluating nursing in-
search program conducted to show the im- terventions derived from a qualitative re-
pacts and outcome changes as a result of an search project.
intervention program. Pederson, A., ONeill, M., and Rootman, I.
Im, E. (2000). A feminist critique of research (1994). Health Promotion in Canada: Provin-
on womens work and health. Health Care cial, National and International Perspectives.
for Women International, 21, 105119. This Philadelphia: W.B. Saunders Canada. This
paper is a feminist critique of traditional text provides a historical and sociological
research into womens work and health. It examination of health promotion in Canada.
identies implications for future research Pender, N. (1996). Health Promotion in Nursing
on womens work and health. Practice (3rd ed). Stamford, CT: Appleton
Kaviani, N., and Stillwell, Y. (2000). An evalua- and Lange. A valuable resource in health
tive study of clinical preceptorship. Nurse promotion practice and research. Provides
Education Today, 20, 218226. This paper an in-depth discussion of the nurses role
uses methods commonly found in evalua- in promoting healthy lifestyle and prevent-
tion research. ing illness.

Part 3
EXPLORING BIAS
AND ETHICAL
ISSUES IN
RESEARCH
P
art 3 discusses issues that need to be thought about carefully
as you commence a research project: these issues have to do
with the problem of bias (Chapter 9) and with ethical ques-
tions (Chapter 10). Bias considers how the researchers expe-
riences, values, and worldview inuences the research process. By
fully understanding bias, its various sources and its impact on re-
search, one becomes free to challenge research knowledge construc-
tively. Ethics, the moral component of knowledge that helps guide re-
search decisions, is explored in relation to planning, implementing,
and reporting research ndings.
295

Chapter 9
Understanding Bias
CHAPTER OUTLINE
A. The Nature of Bias 3. Data Collection

1. Bias in Quantitative and Qualitative 4. Data Analysis
Research 5. Reporting of Findings
2. Triple Biases: Nursing, Science, and 6. Funding
Culture 7. The Use of Findings
3. Sexism: A Prevalent Form of Bias C. Advocacy versus Pure Research
B. How Bias Affects the Research Process D. Guidelines for Minimizing Bias
E. Conclusion: The Gap Between Myth and
1. Selection of the Problem
Reality
2. Sample Selection
KEY TERMS
Androcentricity Familism Random error

Bias Gender insensitivity Research bias
Data massaging Halo effect Researcher affect
Demand characteristic Normative bias Sexism
Expectancy Overgeneralization Systematic error
Experimenter effect Overspecificity
297
298 EXPLORING BIAS AND ETHICAL ISSUES IN RESEARCH
C
hapter 2 argued that humans like to ticular conclusion. In everyday life, we
generalize: we are always trying to have our likes and dislikes. Often inter-
come up with rules to understand pretations of our experiences serve to
our behavior. For example, we might hear confirm for us what we already knew or
someone say, She is weepy these days thought. In short, bias helps to organize
because she is pregnant. Its the hor- our interpretation of events and make
mones. Or we might read that unem- sense of a complex world (Jackson,
ployment leads to depression or that all 1999). So, when staff nurses on a pedi-
adolescents engage in risk-taking behav- atric unit fail to submit the self-schedul-
iors. These generalizations set up ex- ing work plan for the next month to the
pectations in us about the clients we unit manager, she interprets this as a
serve. If these generalizations are to be sign that the staff is under a lot of pres-
relied on in nursing practice, they need to sure. In addition, she concludes that re-
be based on unbiased evidencethat is, cent staff cutbacks have made it difficult
evidence uninuenced by what the re- for her staff to complete the scheduling
searcher would like to nd. But the ten- on time. In contrast, the vice president of
dency to bias conclusions in the direc- nursing sees this failure to submit the
tion of expectations or preferences is a work schedule on time as just another in-
danger in all research. This chapter ex- dication that self-scheduling will not
plores the nature of bias and illustrates work and staff nurses are unable to mu-
that, at all stages of the research process, tually plan work schedules. The point is
the danger of bias needs to be recog- that we often find corroborative evi-
nized. The chapter concludes with some dence for our predispositions. Whereas
rules for detecting and minimizing bias. the staff nurses and unit manager find a
Although bias is omnipresent, we must justification for the nurses behavior, the
not allow it to intimidate us as researchers. vice president of nursing finds yet an-
The need is to be on the alert for it and to other reason why self-scheduling should
minimize it where possible. As Somers be eliminated. Sometimes it takes quite
Roche has noted, bias, similar to anxiety is powerful contrary evidence to change
a thin stream of fear trickling through the our minds, and when we must change
mind. If encouraged it cuts a channel into our minds, we do so unwillingly, with
which all other thoughts are drained. We much moaning and groaning.
need to be careful not to let concerns over Research bias may be dened as the
bias drain our creative energies and pre- systematic distortion of research conclu-
vent us from posing imaginative questions sions. Typically these distortions are in-
in conducting our research. This chapter is advertent, but they can also be inten-
intended to sensitize you to the many tional. There is no doubt that they occur
sources of bias inherent in the research de- in all disciplines in which a set of per-
signs you have examined in the previous sonal preferences prevents an impartial
chapters. This chapter provides practical judgment; nursing is no exception. Bias
advice on controlling bias as you begin de- can inuence most phases of a project,
signing your research project. from problem selection, to the identica-
tion of variables, to developing measure-
ments, to collecting and analyzing data,
to interpreting the results of the re-
A. THE NATURE OF BIAS search. When the chance of bias in a re-
search investigation is not addressed,
A bias may be thought of as a prefer- the reliability of the ndings needs to be
enceor predispositionto favor a par- considered highly suspect.
UNDERSTANDING BIAS 299
1. Bias in Quantitative and necessary bias include selecting partici-

Qualitative Research pants that are unfamiliar to them; clearly
bracketing their own biases, personal be-
To suggest that there is bias, or distortion, liefs, and opinions about the phenomenon
implies that there is an unknown truth under study (see Chapter 6); and selecting
waiting to be described accurately by the a topic that is not too close to the re-
researcher. Bias represents the gap be- searchers on a personal level.
tween the unknown true value and the esti- Qualitative studies are also unfairly
mated value of a particular phenomenon. criticized for having an elite bias. This oc-
This is implicit in quantitative designs in curs because most qualitative designs
which the approach suggests that the re- rely on purposive sampling techniques or
searcher knows and simply wants to volunteers who have had experience
conrm his or her knowledge. For exam- with the research phenomenon under in-
ple, structured questionnaires offer the vestigation. Often the volunteers tend to
respondent a range of replies, made up by be the most articulate and highest status
the researcher, from which the respon- members of a group. Such an elite bias
dent must select a response. These re- can be prevented by ensuring that the in-
sponses often reect the previous knowl- formation you use is representative of
edge, values, and personal preferences of the experiences rather than participants
the researcher (Parahoo, 1997). per se that are important in qualitative
Bias can also be a signicant problem in studies (Norwood, 2000).
qualitative research unless researchers
recognize and incorporate it into the struc-
ture of the study rather than trying to 2. Triple Biases: Nursing, Science,
eliminate it as in quantitative methods and Culture
(Brockopp and Hastings-Tolsma, 1995).
The whole point of qualitative research is Nurse researchers face at least three key
to look at phenomena from the perspec- sources of bias: the biases inherent within
tive of the participant, yet this process is the domain of nursing, the biases of sci-
inuenced by the researchers perspec- ence itself, and the biases acquired as a
tive. To minimize bias, it is important to member of ones society. Although nurs-
clearly record the researchers perspec- ing research systematically challenges
tive. To do this, the researcher establishes our predispositions, we as researchers
an audit trail. This is a clear statement of also take our predispositions with us to
decisions made with documented ratio- our research projects. Furthermore, as
nales so that others reading the audit can nurses, we tend to incorporate certain re-
follow the decision-making process of the search ndings into our predispositions,
researcher. making them even more difcult to
An assumption of qualitative research is change. Secondly, nursing is partly a sci-
that researchers are an integral part of the ence, and science has a system of values
research design and of the experience of and preferences all its own (see Chapter
the participants world. This is sometimes 1). And nally, as a member of a cultural
referred to as going native. Accepting group, nurse researchers have certain
this assumption allows the researcher to predispositions. In a sense, then, nurse re-
build in checks such as an audit trail to al- searchers have a triple load of biasing
low the discovery of knowledge that is not predispositions: one acquired as a mem-
distorted by the researchers biases. Ex- ber of the nursing profession, one ac-
amples of safeguards that responsible quired as a participant in an academic dis-
qualitative researchers use to reduce un- cipline with a traditional and rigorous
method, and one acquired as a member of to note differences between women in

society. All sources of bias need to be rec- terms of factors such as health status,
ognized as potential blinders to a clear ethnic origin, and social class (Hart and
understanding of human behavior, unfet- Bond, 1998).
tered by expectations or preferences. Eichlers Nonsexist Research Methods
(1988) identies several types of sexism
in research. The major types the author
3. Sexism: A Prevalent Form of identies include androcentricity, over-
Bias generalization, and gender insensitivity.
A problem that has plagued nursing, a. Androcentricity

health, and social science research is that
of sexism. At all stages of research, but es- If a researcher presents the world from a
pecially at the design stage, care must be male perspective as if this perspective
taken to avoid sexism (discrimination on were universal, that person is guilty of
the basis of gender). For example, ran- androcentricity. Historically, much of
domized clinical trials (RCTs) have been the health research conducted in North
criticized by feminist groups as extremely America seems to have such a bias.
limited in their ability to produce data Women and ethnic groups have been
that can contribute to an understanding largely ignored in health intervention
of human behavior and experience. Some studies that have focused on white
believe that women, in particular, have middle-class male samples conducted in
been victims of this approach, used by academic centers. In 1990, the Office of
physicians, who are mainly men and tend Research on Womens Health was estab-
to see the world from a male perspective, lished at the National Institutes of Health
and who study women as objects, ignor- (NIH) to address historical inequities in
ing their needs and experiences (Para- research design and allocation of federal
hoo, 1997). Oakley (1989) points out that resources. In 1994, the NIH mandated
the frequency with which doctors im- that women and members of diverse eth-
pose on patients experiments of an un- nic groups be included in all NIH-funded
controlled nature has been one of the projects unless an obvious justification
strongest objections to professionalized for exclusion was present. However, an-
medicine made by the womens health drocentricity continues to be a problem
movement over the last twenty years in in much of the health research endeav-
Europe and North America. ors funded by other sources.
Similarly, it is misleading to suggest An androcentric view of the world is
that feminist perspectives in research congruent with the Western biomedical
are not without bias. For one thing, there ethos that is based on individualism, ma-
is no single unitary feminist perspec- terialism, and competitiveness. This ap-
tive that can be applied in a research proach is in contrast to the views of
situation. For example, African-American women, ethnic groups, and the poor, who
women, women of color, women of mi- usually focus on family concerns rather
norities, and women with disabilities than themselves. These groups do not of-
have been subsumed and made invisible ten risk putting themselves rst before
in writings of women in general. White, their families to obtain costly treatments
middle-class feminists in researching or time-consuming research appoint-
womens perspectives, as opposed to ments. Thus, they exclude themselves
mens, also have been biased in their re- from involvement in clinical trials and
porting of womens experiences, failing procedures that focus on themselves at
the expense of their families. An example Familism is a special case of gender in-
of androcentric research is the exclusion sensitivity and involves treating the family
of women, African-Americans, and Lati- as the unit of analysis when, in fact, it is in-
nos from AIDS studies. AIDS data show dividuals within the family unit that engage
that these groups lag behind white males in a particular activity or hold a certain at-
in studies of protease inhibitor therapy, titude. Familism is also a problem when we
clinical trials, medical services, and pre- assume that a particular phenomenon has
ventive programs (CDCP, 1999; Flaskerud an equal impact on all members of the fam-
and Nyamathi, 2000). ily when, in fact, it may affect different fam-
ily members in different ways.
The view of women in the develop-
b. Overgeneralization ment of psychology is an excellent illus-
tration of sexism. Shields (1988) argues
If a researcher claims to study all people
that much of 19th and early 20th century
but, in fact, samples only men or only
psychology was severely limited because
women, we have an example of an over-
it assumed female intellectual inferiority.
generalization. Eichler uses the example
Research was devoted not to questioning
of a sample of women to study parenting;
this assumption but rather to various at-
the author points out that when one uses
tempts to understand this inferiority.
the term parents to refer to mothers,
Thus efforts were made to:
one is guilty of overgeneralization be-
cause one is ignoring fathers. A parallel 1. Identify those parts of the brain that
problem is overspecicity, when single- were more poorly developed in women
gender terms are used to describe situa- than in men
tions applicable to both genders (e.g., 2. Understand how the greater variability
the doctor . . . he; the nurse . . . she) in male skills leads to a higher propor-
(Eichler, 1988, p. 6). tion of male geniuses
3. Understand the role of the maternal in-
stinct in maintaining women in pas-
c. Gender Insensitivity
sive, subservient roles
To ignore gender as an important variable
Shields concludes her article by noting:
is to display gender insensitivity. Re-
searchers should identify the gender com- Graves (1968, p. v) included among the
position of their samples and be sensitive functions of mythologizing that of the jus-
to the different impacts social policies tication of existing social systems. This
have on men and women (Eichler, 1988, function was clearly operative through-
pp. 67). For example, much of the be- out the evolutionistfunctionalist treat-
reavement research in nursing is biased in ment of the psychology of women: the
that it tends to overrepresent the views of discovery of sex differences in brain
one genderfemales (Cook, 1997). An ex- structure to correspond to appropriate
amination of samples used in bereave- sex differences in brain function; the bio-
ment research shows that the majority of logical justication (via the variability
participants are women. This means that hypothesis) for the enforcement of
much of our knowledge about the grief ex- womens subordinate social status; the
perience is based on the female experi- Victorian weakness and gentility associ-
ence. This had lead to womens grief ex- ated with maternity; and pervading each
perience being considered the norm and of these themes, the assumption of an in-
ndings related to male grief being com- nate emotional, sexless, unimaginative
pared with this norm. female character that played the perfect
foil to the Darwin male. That science Unfortunately, the study may not have
played handmaiden to social values can- been designed to test, and possibly to
not be denied. Whether a parallel situa- rule out, competing explanations. Read-
tion exists in todays study of sex differ- ers of the report may be convinced, in-
ences is open to question (Shields, 1988, appropriately, that the researcher has
p. 55). provided evidence for the particular ex-
planation offered: After all, look at all the
tables that are presented. Although a
great deal of data documenting a variety
B. HOW BIAS AFFECTS THE
of relationships may have been pre-
RESEARCH PROCESS
sented, the researcher may have pro-
vided little, if any, evidence for the expla-
Nurse researchers are frequently con-
nation offered, no matter how reasonable
fronted with interesting ndings for
it may seem or how many graphs and
which an explanation should be offered.
charts accompany it. We need to design
Let us recall our example on lifestyle pat-
studies that systematically test a variety
terns and socioeconomic status (SES)
of possible explanations for the relation-
from Chapter 2. In that project, high
ship under examination.
school students SES and their lifestyle
Researcher affect is a term we have
patterns were examined. Now suppose,
coined that refers to the danger of re-
during data analysis, that a robust rela-
searchers falling in love with a particular
tionship emerges indicating that the
explanation for some relationship or a
higher a students SES, the greater the like-
particular view of the world and inadver-
lihood that the student will adopt a health
tently using procedures that lead to con-
promoting lifestyle. At this stage, the re-
clusions supporting the preferred expla-
searcher may wonder what explains the
nation or worldview. All stages of the
pattern that has emerged. Possibilities
research process may be affected ad-
such as the following might come to mind:
versely by bias. In this section, we exam-
Peers of students with high SESs have ine how bias can affect the initial selection
healthful lifestyles themselves and in- of the problem to be researched, the sam-
uence their friends in such a way that ple design, data collection, data analysis,
they adopt similar behaviors and atti- reporting of ndings, funding decisions,
tudes toward health and health choices. and the use of research ndings.
The parents of students with high SESs
place a high value on health and
healthful lifestyles and expect their 1. Selection of the Problem
children to do the same.
Students with high SESs know that The issue of bias in problem selection is
they have the time and nancial re- that some phenomena are judged to be
sources to engage in healthful patterns more important than otherssome are
of lifestyle, such as joining a gym or t- considered worthy of exploration and oth-
ness class, eating nutritiously, and so ers are not. Because of this, the choice
on, and therefore plan to do so. of subject matter provides a clue as to
Students with high SESs have been ex- the values held by the researcher. Within
posed more frequently to role models the nursing culture, researchers are more
that engage in health promoting likely to study problems that have a con-
lifestyles and are therefore more likely sequence to client care. For example, a
to model themselves after such indi- nurse researcher is more likely to study
viduals. the variables that may reduce the inci-
dence of decubitus ulcers in institutional- Selection bias is also problematic in

ized frail elderly patients than the inci- studies in which individuals themselves
dence of family visitors to them. Whereas self-select to participate in an investiga-
the former is viewed as a critical clinical tion. For example, a nurse researcher
question that may lead to improved client wishes to assess if a new weight loss pro-
care, the latter is not. The researcher is gram contributes to improved weight loss
likely to see the prevention of decubitus and maintenance. If the program is of-
ulcers as an issue with signicant poten- fered to all women who attend a health
tial to improve quality of life for clients clinic, it is likely that the women who are
and increase professional practice knowl- more highly motivated to lose weight will
edge and one over which nurses have con- participate in the program. The problem
trol in practice. It can be argued that iden- is the investigator cannot be sure if the
tifying the most effective variables in new program increased the number of
reducing decubitus ulcers can be done women who lost weight and maintained
without bias. But there is probably no the weight loss or if only highly motivated
value-free, culture-independent way of women joined the program. One way to
choosing variables for a study. Bias re- avoid this selection bias is to design the
sults in the selection of the variables con- sampling plan so that random assignment
ventionally considered important and the of women to the new program and to the
exclusion of those conventionally consid- control group occurs (see Chapter 4).
ered unimportant.
3. Data Collection
2. Sample Selection
Extensive literature in psychology deals
Whether one opts for a survey, panel with the inuence the experimenter may
study, eld study, or experimental design, have in the data collection phase of a
there are potential sources of bias in the study. In their classic studies, Rosenthal
sample selected for study. For example, and Fode (1963) did a series of important
by choosing to survey attitudes toward experimenter effect studies in which stu-
abortion in an urban community contain- dent researchers were asked to collect
ing a free-standing abortion clinic, a re- data on the number of trials it took rats to
searcher will probably produce a study learn a maze. The student researchers
that shows fairly high levels of support for were informed that a new breed of labo-
abortion on demand. To pursue the ex- ratory rat was being developed and that it
ample further, a researcher might do a had been bred for intelligence; the obser-
case study of the attitudes of clinic case vations were to see if, in fact, there was
workers in an inner city community on any difference between the specially bred
public policy issues. Once again, attitudes smart rats and ordinary laboratory rats.
favoring abortion will probably emerge. And so the students set to work, running
Choosing both an inner-city community the two types of rats through the maze.
with a free-standing clinic and case work- It turned out that, indeed, the smart
ers increases the likelihood that the atti- rats took fewer trials to learn the maze
tudes expressed will favor abortion more than the ordinary ones. Apparently the
than would be the case in rural areas or breeding program was working. There was
with many other work groups. The expe- just one problem: Rosenthal did not have
rienced researcher, knowing how differ- any smart rats. The rats were simply as-
ent variables usually work out, could po- signed randomly to the smart and the
tentially select a sample that may bias ordinary categories. It was the students
results. who were the real subjects of this study.
Somehow their expectations about the out the answer; after three stomps, mem-
outcome of the trials had an impact on the bers of the audience would raise their
results of the study. If the experimenter ex- heads slightly, focusing their gaze on the
pects a rat to learn fast, somehow the data horses head. The horse simply watched
will come out that way. Thus, Rosenthal the audiencethey cued him when to
and Fodes ndings indicate that there is a stop. He was a pretty smart horse who
tendency to produce ndings that are con- was not very good at mathematics but
sistent with the experimenters expecta- was a good observer of body language.
tions. The explanation of the so-called ex- How is Clever Hans relevant to the
perimenter effect has been more difcult nurse scientist? Or, for that matter, what
to identify. can we learn from Robert Rosenthals ex-
Did the students fudge the data to periments? Suppose you are conducting
please the professor? Did they perceive an interview, and suppose you have just
errors differently for the two groups, asked a respondent how often she en-
perhaps being less likely to note an error gages in regular physical exercise. The re-
made by the smart rats? Did they han- spondent replies: I dont engage in phys-
dle the rats differently? The generally ac- ical exercise on a regular basis, I am just
cepted view is that experimenter effect too busy to do so. You say, raising your
occurs because of both the inuence of eyebrows slightly, Oh, so you dont ex-
behavior by expectations and slight, but ercise? In all likelihood, your respondent
systematic recording errors by the ob- now feels slightly uncomfortableyour
server. But no matter what the explana- eyebrow movement and your comment
tion, Rosenthals research is of critical have communicated a message. After
importance to the experimentalist as well this, future questions about exercise is-
as to all nursing researchers. sues may well encourage the respondent
to express more interest in exercise and
a. Expectancy physical activity than she actually has. It
is as if the interviewer is demanding a
A German mathematics teacher had a pro-exercise response. Respondents are
horse with unusual talents (Jackson, often interested in guring out what the
1999). The amazing horse, known as survey is really all about. They will,
Clever Hans, could solve simple mathe- therefore, be looking for cues and may be
matical problems by stamping his foot to inuenced by them. If a horse like Clever
indicate his answer. At rst, skeptics Hans can note a slight raising of heads in
thought that Clever Hans trainer was sig- the audience, it is probable that most hu-
naling to the horse, thus accounting for man subjects will be sensitive to a raised
the horses unusual ability. However, it eyebrow, a change in voice tone, or a shift
turned out that even when the trainer in the body position of the interviewer.
was removed from the room, the horse This sensitivity of research subjects to
could still do the trick for other people the researchers expectations of them is
posing the same questions. called expectancy.
The horse was, indeed, very smart. It Although it is appropriate for research
turned out that Clever Hans mathemati- assistants and study participants to be
cal skills declined dramatically when the made aware of the general purpose of the
audience did not know the answer or study, given the problem of expectancy,
when he was blindfolded. Apparently he the nurse researcher should avoid speci-
was watching the audience. If the answer fying the precise hypotheses of the study
was three, the audience would gaze in- either to the participants or the research
tently at the hoof of the horse clumping assistants. Instead, research assistants
should be given an honest explanation but business suit and using a laptop com-
one that does not include the precise out- puter. Similarly, if the researcher is col-
come expected from the treatment. In lecting data from a group of Native Amer-
some cases in which an ethics review icans, it is important to recognize that
board has approved a project involving they speak quite softly and view direct
limited disclosure of information, it may eye contact as rude behavior. Hence, un-
be required that study participants re- like Euro-Americans, who value direct
ceive a debrieng upon completion of the eye contact as a sign of interest and at-
study (see Chapter 10). It may also be fea- tention, the researcher should not be put
sible to use a single-, double- or triple- off by a lack of eye contact from Native
blind experimental design (as discussed in American respondents but rather recog-
Chapter 4), provided the study does not nize it as a sign of respect.
infringe upon participants rights. To re-
view, a single-blind design is one in which c. Demand Characteristics
either the participants or the researchers
are unaware of the assignment to groups. Questionnaires or interview guides may
A double-blind trial is one in which both also provide cues to participants and, un-
the participants and the researchers are less one is careful, these may distort re-
unaware of assignment to groups. A triple- sponses in the direction that the respon-
blind design is one in which persons other dent thinks the researcher prefers. Such
than the researchers evaluate the re- distortions are labeled demand character-
sponse without knowing the group assign- istics. In such situations, test results are
ments of the participants (Parahoo, 1997). distorted because people respond in the
Sometimes it is difcult, however, to way they think they are expected to re-
withhold information about expected spond. The term demand characteristic was
study outcomes from ones research as- coined by psychiatrist Martin T. Orne, who
sistants. As a principal investigator, you noted that subjects in experiments will
want to make them feel that they are a part sometimes play the role of helpful sub-
of the study and, therefore, feel that they ject and produce the results they think
should know what the study is really the researcher desires (Jackson, 1999). A
about. However, if you tell the research as- nurse researcher may phrase a question in
sistants what the hypotheses are, they such a way that the respondent is encour-
may inadvertently bias results either to- aged to answer it in one way rather than
ward the hypothesis (if they are friendly) another. For example, a biased question on
or away from it (if they are hostile). palliative care would be, Do you agree
with the president of the National Palliative
b. Environmental and Cultural Norms Care Association that provision of palliative
care services is directly related to effective
While collecting data, the researcher coping in terminal clients? Reference to a
needs to be sensitive to the norms of the respected nursing gure would no doubt
environment and culture in which the re- load the question in one direction inu-
search is being conducted. This may re- encing the respondents answer.
quire subtle shifts in dress, speech, man- During the data collection phase, a re-
nerisms, or actions. For example, if the search director must ensure that when a
researcher is conducting interviews with selected respondent refuses to participate
adolescents in a clinic in a low-income or is unavailable that the replacement is
neighborhood, imagine the reactions and selected on an equal probability basis (see
responses of the teens to a researcher Chapter 15). If this is not done, there is a
who shows up in a three-piece Givenchy danger that the data collection personnel
will simply choose the most convenient three ways: through coding errors, data
replacement, which will tend to bias the massaging, and hunting. Any of these
study by overrepresenting those people may be intentional or unintentional.
who spend more time at home. In the
words of Smith (1991), researchers often a. Coding Errors
have to work with the people who are
available; these may comprise volun- At the beginning of an analysis in surveys,
teers, hypochondriacs, scientic do-good- interview studies, and participant obser-
ers, those with nothing else to do, and so vation studies, researchers go through a
on. Even with probability sampling, bias is process of coding the information. What
present when certain groups such as this means is that if data are collected in
those too busy to participate in a mail or an open-ended format (questions with-
telephone survey or those too ill to be in- out xed-response categories), then the
cluded in a clinical study are excluded sys- information should be placed into cate-
tematically. Researchers need to acknowl- gories before analysis begins. The pro-
edge these possible sources of bias in cess of coding is subject to two kinds of
their study designs and provide readers of error, random and systematic.
their research reports with such informa- Random error refers to inconsisten-
tion so that they may judge the represen- cies that enter into the coding process
tativeness of the sample and the general- but that display no systematic pattern.
izability of the ndings. For example, suppose that you are cod-
ing people into the following educational
d. Halo Effect categories: (1) 8 or fewer years of formal
education, (2) 9 to 12 years of formal ed-
Bias may creep into data collection pro- ucation, (3) some postsecondary train-
cedures when rating scales are used. The ing, or (4) college or university gradua-
halo effect is a carryover effect in which tion. If you accidentally coded a person
the researchers rst rating may inu- with 5 years of education into category
ence the second and subsequent ratings. (2), you would have made an accidental
The general impressions the researcher or random error. In processing data, one
has of a particular situation may carry often enters the information twice, com-
over and inuence subsequent ratings of paring the rst version with the second in
similar situations. This is because the reorder to locate such random errors. Oc-
searcher has a tendency to reinforce her casionally, entry errors include values
or his impressions that are not within the range of possible
The halo effect may also result when values: in the above example, entering a 6
the researcher attempts to make ratings would be an example of such an error.
consistent. Such bias may be reduced by These errors are easiest to spot because
having different raters rate the partici- after data analysis begins, an out-of-range
pants or phenomenon or by having the value will become apparent as soon as
same raters do the rating at different one runs a frequency distribution of the
times without knowledge that they are variable. And because these errors sim-
rating the same participant or phenome- ply represent noise in ones data, they
non as done previously. are not as threatening to the conclusions
of a study as systematic errors.
4. Data Analysis Systematic errors are especially prob-
lematic. These errors are in danger of bi-
During the data analysis stage, bias is asing a study because they systemati-
most commonly introduced in one of cally distort the data in one particular
direction. For example, suppose that any two categories, low and high education,
person who does not answer the ques- where would you draw the line between
tion on education is assigned to the low- the two groups?) Data massaging is work-
est category, the category representing ing with the data until the analysis pro-
those people with 8 or fewer years of ed- duces the strongest association among
ucation. In such cases, we would be bias- the variables.
ing the data systematically for those who Such data massaging violates the prin-
refuse to answer the question by always ciples of objectivity. Unfortunately, it is
coding them as if they should fall into the practiced in all disciplines. Because
lowest educational category. (Normally every discipline has a range of approved
those who do not answer a question are methods of analysis, massaging the data
assigned a missing value code for the may seem perfectly legitimate. After all,
question.) Another kind of systematic er- the researcher wants to produce the
ror could occur if we discovered, after best analysis of the data, and in many
having developed code categories and disciplines, only statistically signicant
starting our work, that another and un- results are publishable. Even more unfor-
categorized kind of response is occurring tunately, such data massaging is rarely
halfway through the analysis. For exam- reported in formal presentations of the
ple, if, to save time, we do not go back to research.
the beginning and reexamine all of our
cases to see if some of the codings have c. Hunting
been inappropriately forced into certain
categories, we will have decreased the Given the modern computer, it is now
number of cases that fall into the newly possible to run tests for many different
discovered category systematically. This relationships. There may be four or ve
kind of difculty is likely to pop up in different operationalizations of a concept
studies that have many open-ended ques- (e.g., SES) available to the researcher in a
tions. (This is one reason, incidentally, data set; perhaps the various hypothe-
why many researchers avoid using a lot sized relationships are run using each of
of open-ended questions.) these possibilities. Finally, the one that is
most congruent with the researchers ex-
b. Data Massaging pectations gets reported. A rationale for
throwing out some of the results can
Particularly in surveys and secondary readily be found by arguing that they rep-
data analyses, a great amount of informa- resent a poor measurement of the rele-
tion is collected on a large number of vari- vant variable. Most researchers have,
ables. This data may be treated in a vari- from time to time, engaged in some se-
ety of waysmany of which may be lective use of data. If the researcher
viewed as alternative modes of analysis. hunts through a data set long enough,
The researcher may analyze the data in a an acceptableeven interestingnding
number of different ways, discarding the will surely emerge. At that point, the hunt
results that are less interesting and stops.
keeping those that make the most sense. In the physical and the social sciences,
If cross-tabulations are being done, a vari- hunting is common. What student in
ety of cutpoints may be tried; some will be chemistry or a physics laboratory has
retained, and others will be discarded. (If not checked the results obtained by
you had measured respondents educa- other students? If unexpected results oc-
tion by years of formal school completed cur, then the procedure is rerun on the
and you now wished to group them into grounds that something must have been
done wrong. In short, if anticipated re- may represent uke results (Sterling,
sults are not obtained, the results are dis- 1959). His argument is that there are a lot
carded. Even mature researchers have of researchers, and many may be working
the tendency to work with the data until on a similar problem at any time. If the
the right nding is obtained. The search 0.05 level of signicance has been used,
for reportable ndings is a continuous an average of one study in 20 will produce
process in all science, yet few papers ac- a statistically signicant relationship
knowledge this search. Reading research purely by chance, and such studies may
papers leaves the impression that most even get published ( journals ordinarily
projects are easy and straightforward, do not publish papers reporting no rela-
but few actually are. tionship ndings). The obvious question
The key issue in discussing bias as it is what proportion of journal articles are
relates to analysis is this: What scientic based on these uke studies. Although
principle determines when analysis is n- Sterlings point is an interesting one, it
ished? Do we stop analysis when we get a may overstate the problem (Jackson,
reportable, respectable nding? If we 1999). Nonetheless, the argument needs
stop analysis when we get the results we to be kept in mind as one possible source
like, the bias will be toward conrming of bias in published studies. In a recent
expectations. Although some of the data update of the 1959 research, Sterling and
may be consistent with expectations, associates (1995) note that little has
some may not be. Frequently, inconsis- changed in publication decisions with re-
tencies are not reported. spect to tests of signicance. They note
For survey researchers, there is a spe- that in a 1958 study of psychology jour-
cial problem. Given many variables and nals, some 97.3 percent of the articles re-
observations, it is possible to run any vari- porting tests of signicance rejected the
able against every other variable in a null hypothesis; in 1986 to 1987, the gure
search for signicant ndings. If we use stood at 95.6 percent rejecting the null
the 0.05 level of signicance (see Chapter hypothesis. Box 9.1 summarizes some of
12), then we would expect one in 20 of the the points raised in the article by Sterling
relationships examined to be statistically and associates.
signicant simply on a chance basis. Un- Sterlings argument can be extended
less the researcher reports the hunting into another area. There is a selection
that has occurred, the reader of a report process that nds certain ndings re-
will be in no position to regard the con- portable and others unreportable. In-
clusions with the skepticism they deserve. deed, it is unusual for ones rst analysis
Finally, it should be pointed out that of the data to make it through to the nal
there is nothing inherently wrong with report. Although there are many good
exploring data, looking at relationships reasons for analyzing data in different
that have not been hypothesized. But ways, the process of deciding which nd-
such analyses should not be reported un- ings to report may have more to do with
less it is made clear that no hypothesis aesthetics than science.
has guided the search. At least then the In nursing research, the issues of both
reader has been cautioned. statistical and clinical signicance must
be considered in determining the re-
5. Reporting of Findings portability of ndings. Clinical signi-
cance, which is discussed in Chapter 12,
In 1959, T.D. Sterling published a classic refers to the potential for research nd-
paper that suggested that much of what ings to make a real and important differ-
is being published in learned journals ence in clinical practice, in improving
BOX 9.1 Bias in Publication Decisions
The scientific method has traditionally been motes redundant work because few will be
touted as a model of objectivity, but a team of aware of the negative result until they find out
researchers at Simon Fraser University in about it firsthand.
British Columbia has revealed a significant bias Moreover, the increasing popularity of a sta-
in the way scientific results are reported. tistical method called meta-analysis makes it all
According to their work, scientists who of- the more important to know about negative re-
fer positive findings from their worksuch as sults. Meta-analysis pools the results of a num-
the verification of a hypothetical experimental ber of published studies without bothering to
outcomestand a better chance of getting carry out any original research. But this ap-
published. Meanwhile, those whose findings in- proach assumes all the relevant scientific find-
dicate the lack of a resultsuch as the denial of ings, rather than just the positive ones, will be
an expected experimental outcomeare unavailable in print.
likely ever to see their conclusions in print. Because access to study results is typically
You get this idiocy that you dont publish limited to published studies, the question of
negative results, says computer science pro- whether or not published studies constitute a
fessor Ted Sterling, who first noted the trend in representative sample of relevant studies is of
a research paper he published in 1959. At that concern, state the researchers in their Ameri-
time, he analyzed the contents of a selected can Statistician paper.
group of psychology journals and noted the The solution, according to Sterling, lies with
overwhelming prevalence of articles touting a new strategy for selecting scientific work to
positive outcomes. Dr Sterling says he was dis- publish. He advocates blind-to-outcome peer
appointed to find out that nothing had changed review, which would weigh submitted papers
more than 30 years later. on the basis of the importance and relevance of
For this article, Dr Sterling and his colleagues the work, instead of relying on the nature of the
sampled a wider variety of scientific journals, but outcome. In light of the fact that little about this
the prevalence of positive results had not aspect of scientific publication appears to have
changed. The article even cites a rejection letter changed in 30 years and because the possibly
written by the editor of a major environment misleading consequences are becoming more
and toxicology journal, which states bluntly, serious, his colleagues and he argue that this
The negative results translate into a minimal sort of drastic modification of editorial policy is
contribution to the field. necessary.
The Simon Fraser researchers disagree with In short, their study concludes, more rad-
that observation, suggesting that alerting scien- ical measures than public consciousness raising
tists to negative findings can prevent others are needed to curtail the influence of publication
from repeating the same experiment and ob- bias.
taining the same findings, a process that pro-
SOURCE: University Affairs, January, 1996, p. 16. This report was based on Sterling, T.D., Rosenbaum, W.L., and
Weinkam, J.J. (1995). Publication decisions revisited: The effect of the outcome of statistical tests on the decision to
publish and vice versa. The American Statistician, 49, 108112. Cited with permission.
health status, or to a problem identied failed to nd signicant results. This

as a priority for the discipline. Tradition- problem was investigated by Polit and
ally, researchers have relied on statisti- Sherman (1990), who analyzed 62 nursing
cal signicance to imply clinical signi- research studies published in Nursing Re-
cance (Jeans, 1992). This has lead to the search and Research in Nursing and Health
problem of nursing research being re- to determine if, similar to the researchers
jected by editors because the studies in other disciplines, nurse researchers
were at high risk of nding nonsignicant cipline and, therefore, have developed a
results even when their research hy- nursing perspective on issues. Concep-
potheses were correct. Results indicated tual and theoretical models prevalent in
that many nurse researchers are design- the discipline shape these perspectives.
ing inadequately powered studies in All conceptual and theoretical views
which their hypotheses will fail to be sup- limitand, at times, biasour perspec-
ported, even though the hypotheses are tive of the topic under study and inuence
correct. Two reasons explain this. One is our reporting of ndings. In other words,
that the majority of hypotheses tested by nurses who support the self-care model of
nurses involve effects that are small or nursing frame their conclusions in rela-
small to moderate. This is because nurs- tion to self-care, and those who support
ing phenomena are typically so complex the stress adaptation model of nursing
and difcult to measure that no single come to stress adaptation conclusions.
study can hope to explain a substantial Chapter 2 outlined some types of awed
portion of the variance. The second rea- arguments that may nd their way into the
son is the tendency of nurse researchers nal report of a study. Included among
to use small samples. Nursing studies these were inappropriate appeals to au-
typically use samples of fewer than 100 thority; provincialism; setting up a false
participants (Moody et al., 1988). Be- dilemma; missing, insufcient, or sup-
cause of the small sample sizes, often ne- pressed evidence; and unwarranted con-
cessitated by the clinical nature of nurs- clusions. Readers should be alert to these
ing research, a substantial number of aws in reading a nal report on a research
published nursing studies and presum- project.
ably even more of unpublished studies In addition to being evident in theo-
lack sufcient power to detect real ef- retical predispositions and improper ar-
fects. Therefore, Sterlings argument that gument, bias can also be reflected by in-
what is published may represent uke re- sensitivity to minorities, sexism, or in
sults can also apply in nursing research, tendencies to go beyond the limits of
in which many studies fail to detect sta- ones data when interpreting them. And
tistically signicant differences that are, although scholars are trained to be cau-
however, truly clinically important. tious in their interpretations, it is diffi-
The difculty is that some nonscientic cult to avoid suggesting extrapolations
considerations come into play when re- beyond those justified by the data. Some
ports are submitted for publication or for of these issues are explored more fully in
presentation at professional conferences. Chapter 19, which deals with report writ-
Are the ndings culturally acceptable? Are ing.
the ndings acceptable to ones peers? Bias is just one of many factors that
Because pressure exists to keep papers may inuence the outcome of a study.
and presentations short, only the major How important bias itself is varies across
nding is reported; this major nding, many conditions. When we fail to repli-
however, may not be representative of the cate a study ( i.e., fail to get the same nd-
ndings of the whole research project. ing as another researcher), we cannot as-
The point here is that not only does the sume that the difference in results is
choice of subject matter reect values but caused by bias alone. Many factors, in-
also the researchers theoretical predis- cluding bias, may have been responsible.
positions may inuence the conclusions Projects are rarely as straightforward
of the study as much as any data col- as the nal report on the project implies.
lected. Many nurse researchers have un- Consider sampling, for example. Only rare
dergone years of socialization in their dis- projects do not run into some sampling
difculties. First, there may be problems porting of the results. Otherwise, the cost
with refusals and lost questionnaires. of biased results may be very expensive
Some of the responses may not be clear; for the client and the health-care system,
when comparisons between the sample both nancially and emotionally.
and the known parameters of the popula- The value society places on different
tion are compared, there may often be un- kinds of research is also reected in the
comfortable disparities. Interviewers may relative amounts of research funds made
have cut corners in the interest of com- available. Traditionally, medical research
pleting the project (such tactics may that targets the diagnosis and cure of dis-
range from faking interviews to avoiding eases has received most of the federal
the normal random respondent selection health dollars targeted for research in
process, even going so far as to include both Canada and the United States. Nurs-
whomever happens to be available in a ing research continues to be under-
sample). In short, research is inevitably funded when compared with other pro-
messier than our reports of it. fessional disciplines. Jeans (1990, p. 2)
raises the question: Why is it that when
legislators clearly place priorities for
6. Funding health research on areas that are the very
substance of nursing research, these rec-
Granting agencies, although they may ommendations do not translate into the ac-
claim to fund projects on the basis of tual funding of health research? Who is
merit, are themselves subject to bias. In passing the buck or not passing the bucks,
Canada, for example, a variety of councils as the case seems to be? Some nurse re-
and foundations related to health care are searchers suggest it is the entrenchment
funded by the federal government, which of a system by a dominant group (med-
may favor projects that support legislative ical researchers) that prevents changes
agendas. Similarly, in the United States, in patterns of research funding and,
the National Institute of Nursing Research therefore, inuences the type of research
(NINR) in planning future research for the that is conducted.
next 5 years (2000 to 2004) has selected In addition, scholars in the traditional
seven key areas that will dominate the in- disciplines such as medicine and the nat-
stitutes research portfolio and be funding ural sciences are more readily funded
priorities. These include chronic illness, than those in newer areas such as nurs-
health promotion and disease prevention, ing, allied health professions, or interdis-
quality and cost effectiveness of care, ciplinary research in health promotion.
management of symptoms, adaptation to Indeed, in recent years, some funding
new technologies, health disparities, and agencies have set up special committees
palliative care. to deal with applications for people work-
If external funding has been provided, ing in newly developing areas. The estab-
it is important to determine who the lishment of the new review committees is
sponsors of a research project are be- a response to the perception that schol-
cause the ndings may be biased in the ars in these areas are at a disadvantage
direction of their interests. For example, when competing with scholars in tradi-
nursing studies are often sponsored by tional disciplines.
drug companies or medical supply com- Within the university system, peer-
panies who wish to sell a product. In review committees typically adjudicate
these situations, nurses and readers of project-funding decisions. These commit-
such research reports need to be cog- tees usually face a high demand for funds
nizant of the potential for bias in the re- and severely limited resources. In their
funding decisions, they usually select When researchers slip inadvertently

studies that reect current research into advocacy roles, they compromise
trends and that will advance their disci- their credibility as impartial scientists.
pline. In summary, nursing research oper- Readers should note an important dis-
ates in a social context. Researchers are tinction between the advocacy role,
constrained by peers and granting agen- which is viewed as pejorative in relation
cies to not engage in trivial research and to research, and the patient advocacy
instead focus on issues that are consid- role, which nurses assume in relation to
ered important. What is considered im- their practice. The latter is a covenantal
portant in one generation may be con- professional relationship that grows out
sidered as trivial by the next one. of a moral principle of delity to the client
and a commitment to protecting those at
vulnerable points in their lives (Delough-
7. The Use of Findings ery, 1995). In this role, the nurse and
client work together to reach decisions
Given the enormous condence that West- that t with the unique needs and values
ern culture has in science, it is no surprise of the client. It should not be confused
that the ndings of science are powerful with the citizenadvocate role in which
tools. Courts, politicians, the media, and one produces scientic evidence to sup-
the general public seem to respect sci- port a particular position or cause. Nor
ence. Nurse researchers are increasingly should the nurse researcher role be con-
interviewed on radio and television, ap- fused with either the nurse advocate role
pear as expert witnesses in courts, and, in- or the citizenadvocate role. Nurse re-
deed, provide evidence taken into account searchers must maintain objectivity and
when legal and public policy decisions re- a search for truth that is unbiased and im-
lated to health matters are made. Scien- partial if their work is to be taken seri-
tic evidence is taken seriously. ously and if nursing is to be considered a
Unfortunately, the research literature is respected, scholarly discipline that con-
easily misrepresented. The research liter- tributes to improved practice.
ature contains many ndings, some of An impartial approach would try to
which support a particular view and oth- disconrm a theory, try to rule out alter-
ers that do not. Researchers who are com- natives, and continually press any given
mitted to some social cause, health issue, theory hard in an effort to discover the
or a particular theoretical perspective limits under which it is applicable. Al-
may select evidence that helps establish a though this can never be achieved per-
particular positionsimilar to a debater fectly, those who attempt to follow this
seeking support for a particular conclu- approach are, nonetheless, trying to elim-
sion. Because of a desire to support a par- inate as much bias as possible in their re-
ticular position, the debater is not inter- search or in their reviews.
ested in contrary evidence; only preferred Peter W. Huber has explored the prob-
evidence is reported. The problem with lem of the scientist-for-hire as it relates to
using the debaters approach is that many the American court system in his book
people may think that the ndings are ob- Galileos Revenge: Junk Science in the
jective and impartial and that they reect Courtroom (1991). Numerous court cases
dispassionate scientic views. But they in which huge sums of money are at stake
may simply be a conclusion seeking cor- involve the use of expert scientic wit-
roborating evidence. Debaters enjoy the nesses. Huber argues that the system is
credibility of science while violating the at fault because it seems unable to distin-
principle of impartiality. guish good scientic testimony from that
of the science charlatans who make ca- C. ADVOCACY VERSUS PURE

reers out of court appearances. American RESEARCH
courts are faulted for setting insufcient
standards as to who may testify. Fre- In our exploration of sources of bias in
quently lawyers consider many scien- the various stages of research, we have
tists, settling nally on the ones who are noted how the outcomes of research may
willing to make appropriate testimony be inadvertently distorted. Findings tend
and to be coached. As personal injury to move in the direction of the culturally
lawyer, Dennis Roberts notes: acceptable, in the direction of our expec-
tations, and in the direction of our pref-
You get a professor who earns $60,000
erences. And if the pervasive attitudes of
a year and give him the opportunity to
both the relevant scholarly disciplines
make a couple of hundred thousand
and the larger society are liberal, inclu-
dollars in his spare time and he will
sive, and emphasize tolerance, then the
jump at the chance . . . (quoted in Hu-
research outcomes will tend to reect
ber, 1991, p. 19).
these views. On the other hand, a more
Huber argues that courts would do conservative, exclusive, and intolerant
well to pay more attention to establishing society will encourage research out-
the consensus on a topic in the research comes that are supportive of these views.
community and rely less on the personal Funding agencies and their selection
opinion of an expert who has been se- processes will particularly encourage
lected and groomed by a lawyer who is mainstream researchersthose whom
seeking a favorable outcome. As long as referees think meet the standards gov-
lawyers are doing the hiring, they are erning good research. A similar argu-
likely to stack the deck to improve their ment can be advanced about the publica-
chances of winning their cases. This prac- tion decisions of journal editors.
tice fails to represent the ndings of sci- In recent decades, a blurring of the line
ence objectively and, therefore, may between advocacy research on the one
compromise the ability of the courts to hand and pure or descriptive research on
come to reasonable conclusions. the other seems to have occurred. One
The role of the investigator in intro- possible explanation is that practitioners
ducing bias can affect every stage of the in nursing, health-related, and social sci-
research process. Investigator bias needs ence disciplines have become increas-
to be acknowledged and made explicit so ingly aware that all research carries with
that it does not undermine the quality of it cultural baggage. So, whether we talk of
research. Cook identies four areas of in- research design bias, funding decisions,
vestigator bias that may compromise re- data collection, analysis, interpretation,
search conclusions: emotional, norma- or publication decisions, research tends
tive, cultural, and professional bias. Box to reect its sociocultural milieu. This re-
9.2 delineates these four areas. alization perhaps helped to legitimize the
In summary, researchers should be use of research to advocate changes or
aware of the way in which bias may affect advance the personal or collective agen-
the research process as well as the con- das of its practitioners.
clusions, reporting, and use of a research The degree of bias reected in research
report. The danger of becoming overly reports ranges across a continuum. On the
concerned about bias, however, is that re- one hand, there are research reports
searchers may spend far too much time specically designed to study bias (Cook,
contemplating the difculties of research, 1997; Flaskerud and Nyanathi, 2000); the
and, as a result, not get the job done. psychology experiments exploring experi-
TYPES OF INVESTIGATOR BIAS IN BEREAVEMENT RESEARCH

1. Emotional bias: Objectivity becomes dif- nomenon. Research is often biased in that
ficult to achieve because of emotional in- it overrepresents one gender.
volvement with individuals or families or 3. Cultural bias: The cultural background of
personal (sometimes traumatic) experiences the researcher serves as an interpretive lens
associated with the phenomenon under in- that can limit the understanding of the phe-
vestigation. Many nursing studies deal with nomenon under investigation in a different
experiences and health behavior choices culture, ethnic group, or social class. When
that are taboo or laden with emotion. Inves- the investigator and research participants
tigators need to confront their own emo- come from different cultural backgrounds,
tional responses related to the research conflict and misunderstanding are likely to
topic because these can affect choice of re- emerge. Members of a dominant culture of-
search topic, interpretation of findings, and ten define normalcy in terms of their own
recruitment and responses of participants. experience. For example, nurses in clinical
Emotional bias can lead to anxiety in re- settings are likely to have more positive per-
searchers related to their own experiences ceptions of patients from their own cultural
and decrease their willingness to contact group than of other ethnic or cultural
participants. If brought to the researchers groups. Investigators who are sensitive to
attention, however, emotional bias can en- their own belief systems are better equipped
hance the researchers understanding of to detect cultural influences and acknowl-
their findings. This has been recorded in the edge cultural dimensions of the research
area of bereavement research in nursing, in phenomenon.
which personal loss by the researcher, when 4. Professional bias: The researchers discipli-
acknowledged, can facilitate understanding. nary training and perspective determine the
2. Normative bias: Value judgments are particular aspects of the research phenome-
made by the investigator regarding the non that are studied and the conclusions that
normalcy and appropriateness of phenom- are drawn. Researchers have been profes-
enon-related feelings and behaviors based sionally socialized into distinctive points of
on current conceptual models, theories, view; language; and sets of professional skills,
and available empirical data. Researchers perspectives, attitudes, and feelings. For ex-
often use conceptual models to guide re- ample, nursing perspectives are reflected in
search. The assumptions of these models certain research questions that tend to be ad-
go unchallenged and researchers tend to dressed repeatedly in nursing investigations
hold onto familiar models and theories regardless of the research topic. Questions
even in the face of contrary evidence. Gen- regarding ways of assisting families to deal
der issues need to be considered in defin- with stress is an example of a nursing per-
ing normative responses to the study phe- spective or professional bias.
SOURCE: Summarized from Cook, A.S. (1997). Investigator bias in bereavement research: Ethical and methodologi-
cal implications. Canadian Journal of Nursing Research, 29(4), 8793.
menter effects fall into this category leaves little doubt as to its underpinning
(Rosenthal and Fode, 1963, 1966; Jackson, assumptions and preferences. So, whether
1999). Although there may well be some in- we are dealing with a community group do-
advertent bias to nd bias in these studies, ing research to block a health center clo-
great pains are taken to measure the sure or a research project to oppose the lo-
sources and extent of distortions. At the cation of a landll site, there is no question
other end of the continuum are the social as to the stance that will be taken in the
activists and advocates whose research study. Bias almost certainly takes the form
of overstating both the size and impor- mandate that goes beyond an obligation to
tance of the problem. In between these ex- understand the world into an obligation to
tremes fall studies whose distortions are change it for the better. This mandate ne-
subtle but almost certainly present. cessitates a different order of knowledge
Increasing awareness of the problems that must be created and tested through
of bias in all research has, in some ways, research. As a practicing profession, the
helped to legitimate those who openly de- knowledge obtained through nursing re-
vote their research to advocacy. Critics of search must be meaningful to the nurse in
mainstream research believe it supports a particular situation and translated into
the interests of the established order in action. The notion of meaningfulness is in
society. Thus, those who consider health- itself a value, a source of bias. It comes
care delivery worthy of investigation are from the commitment to excellence in
considered to be responding to middle practice and from the realization that in-
class concerns about the preservation of formed biasmeaningfulnessenhances
the health-care system as a national pro- the quality of nursing research, rather than
gram; those who study declines in govern- detracting from it (Diers, 1979). Diers sug-
ment funding for higher education are per- gests that nursing research must press to
haps responding to middle class parents address meaningful clinical problems in
concerns about the cost of educating their meaningful ways in order to produce
children. Those supporting such advo- meaningful results to improve nursing
cacy research would argue that if main- practice. This in no way absolves nurse re-
stream research has supported the estab- searchers of the responsibility to do good
lished order in society, there should also science. Nurse researchers should present
be support for those who openly advance data clearly and in sufcient detail to allow
the interests of the underclasses. Why others to replicate the work. Regardless of
should researchers not try to advance the the topic, researchers need to be critical of
interests of minorities, women, people the ndings and attempt to falsify them as
with disabilities, the working poor, the well as offer alternative explanations for
homeless, or developing countries? Why them. This may be difcult if a researcher
should researchers in nursing, health- is attempting to evaluate a new program or
related, and social science disciplines not intervention or to engage in advocacy re-
apply knowledge to alleviate suffering? search, but it is required if the hard-won
Should pure researchers stop at describ- reputation of nursing as a serious scientic
ing and explaining the incidence of suffer- discipline is to be protected. Remember,
ing? Should the applied question, the only the researcher knows the intimate
question of alleviating suffering, be ad- details surrounding the context and pro-
dressed directly by the researcher? In- cess of data collection; therefore, the re-
creasingly, nurses, health-care profession- searcher is in the best position to identify
als, and social scientists have the view possible sources of bias and error. The
that pure research, although it has its question remainswill the researcher do
place, does not go far enough; someone so? Sometimes in the process of inquiry we
must advocate for the applied changes come to learn more about ourselves than
suggested by the research ndings. Thus, the ndings, and we are the richer for it.
the role of scientists in these disciplines The challenge to create an impartial
has broadened to include an applied di- nursing science has been enormous: some
mension. But this broadening has led to would deny its possibility; others would
confusion as to the line between pure re- claim that it is difcult, but not impossible;
search and advocacy. and still others would claim that all we can
Nursing, similar to medicine and law do is try to minimize bias in nursing re-
(among other disciplines), carries a social search. Realizing that there are many dif-
culties in doing research should not lead identied as such to avoid confusion.
students to despair. Nurse researchers In many cases, our role shifts several
deal with an extremely difcult subject times while doing a project. For exam-
matterthat of understanding complex ple, the choice of subject matter may
human responses to health and illness be inuenced by advocacy interests;
across all stages of the life cycle. However, the interpretation of the descriptive
all disciplines have their difculties and and pure aspects of the research may
challenges, and nursing is no exception. be largely impartial and value neutral;
and the policy recommendations may
be directed toward bringing about
changes meant to benet a particular
D. GUIDELINES FOR
social group. Try to be clear about
MINIMIZING BIAS
what hat you are wearing and alert the
reader when you switch hats. Make
The following sections provide some ten-
sure the standards of evidence do not
tative suggestions for dealing with re-
change as you move from one phase to
search bias. There are no easy answers
another.
here. These guidelines are intended to
Guideline 4: Eliminate bias in descrip-
provoke discussion and because some of
tive phase of research. In the descrip-
them are not acceptable to all nurse re-
tive phase of research, efforts should
searchers, they should be considered ten-
be made to ensure that the description
tative.
is not biased by the way questions are
Guideline 1: Alert key players to the asked and to ensure that proper sam-
problem of bias. We need to alert our pling procedures have been followed
society to the idea that much research, rigidly. Although subjective elements
in the end, tends to support some in- inuence choices about what should be
terest or interest group. Education described and how questions should
about bias needs to be directed to stu- be posed, after such decisions have
dents, research practitioners, public been made, the collection, analysis, and
policy personnel, courts, governments, interpretation of data should be as free
and the general public. To the extent of distortion as possible.
that research supports some interest Guideline 5: In explanatory research, let
or interest group, research activities disconrmation be your guide. In pure
may have an underlying structural bias research, be certain that competing ex-
that favors doing work on certain types planations are given a fair chance. Ef-
of research questions and arriving at forts should be made to disconrm re-
socially acceptable conclusions. lationships, not support them.
Guideline 2: Avoid sexism. In all phases Guideline 6: Identify policy recommen-
of research, avoid sexism. Do not over- dations as value based. In the policy
represent the experiences of one gen- recommendation phase, it should be
der. Examine your own gender biases made clear that the evaluations and
and increase your sensitivity to gender recommendations of the researcher are
issues in research. This will enable re- intended to achieve some end; given
searchers to be more open to wider in- this assumption, we should ask if there
terpretations of research phenomena. is any connection between the conclu-
Guideline 3: Identify roles played by the sions of the research and the re-
researcher. As researchers, we need searchers assessment of the likely suc-
to be honest about the role we are cess of the proposed intervention
playing. Advocacy research should be strategy. In some cases, the research
has specically evaluated competing logically or empirically from those that

intervention strategies; in other cases, are speculative. To avoid bias is also to
the intervention strategy proposed is recognize it. One must distinguish be-
purely commonsensical and has not tween conclusions that appear to have
been evaluated. In both cases, the ex- a sound basis and those that are specu-
tent to which the research has assessed lative and untested.
intervention alternatives should be Guideline 9: Distinguish advocacy from
made clear to the reader. In short, has pure research. Try to identify the re-
the research actually tested the likely searchers who are engaged in advo-
response to both intervention strategy cacy. By being able to identify some-
A and intervention strategy B? one who is mounting evidence to
Guideline 7: Be skeptical of research support a position, whether on theo-
ndings. Be skeptical of all reported retical or public policy issues, one can
research ndings, given the many recognize a debater at work. There is
sources of bias that may have inu- nothing inherently wrong with advo-
enced the conclusions. A healthy, cacy, but it must be recognized that its
questioning attitude toward ones own conclusions are not impartial. When it
and others ndings is appropriate, is clear that a case is being mounted,
particularly in disciplines such as the question to ask is: But what ex-
nursing in which there is so much periment could disprove your hypoth-
room for interpretation. esis? (Jackson, 1999).
When reporting research ndings of Guideline 10: Orient research to discon-
qualitative studies, it is important to rmation. If studies are designed to
distinguish between the knowledge of rule out alternatives or to disconrm
the participants, the knowledge the re- theories, then one is on the road to
searcher originally brought to the pro- minimizing bias.
ject, and the insights the researcher Guideline 11: Use theory to generate
has developed in the process of ana- testable hypotheses. Use theory as a
lyzing the data and reporting the nd- guide, as a tool for generating testable
ings. If these distinctions are explicit in hypotheses. Theories should be re-
the reporting of the ndings, then garded as efcient summaries of nd-
readers will be able to separate ac- ings and as tools for deriving predic-
counts of participants views from the tions about relationships between
views of the researcher (Dawes, et al., variables. Theories are attempts to
1999). Therefore, it should be obvious make general summary statements
in the reporting what the researchers and are to be revised continually. Re-
views are as distinct from the partici- member, theories are for testing, not
pants accounts. To minimize bias in supporting.
qualitative studies researchers must Guideline 12: Be sensitive to your own
make clear the process by which the outcome preferences. Contemplate
data were collected, analyzed, and the your own values. Recognizing (bracket-
conclusions formulated. By completing ing) your own preferences concerning
audit trails, researchers acknowledge the kind of nursing world you would like
how their perspectives inuenced the to live in will help to alert you to poten-
ndings. tial biases that you may bring to a re-
Guideline 8: Read the research literature search project. Being sensitive to your
cautiously. In reviewing literature, try own personal and professional biases
to distinguish between those conclu- enables you to design studies that are
sions that appear to be demonstrated more value neutral rather than studies
predisposed to generate results favor- viewed as a restrictive guideline, if fol-

ing your personal preferences. By set- lowed, it prevents unwarranted mas-
ting aside personal preferences and be- saging of the data.
liefs surrounding phenomenon under In qualitative studies, awareness of
investigation, the investigator prevents researcher bias about the topic helps
the information from biasing or inter- investigators identify a particular
fering with the recovery of pure de- frame of reference that may limit or di-
scription of phenomenon of interest. rect the data interpretation. Self-
Sensitivity to ones preferences and val- awareness through the process of
ues must be constant and ongoing in bracketing promotes honesty in nding
both quantitative and qualitative inves- the truth and decreases the inuence
tigations so that the purest form of re- of bias in data interpretation (Streu-
search ndings emerge. bert and Carpenter, 1999).
Guideline 13: Do not disclose hypothe- Guideline 17: Check for random and sys-
ses to subjects or assistants. Do not re- tematic errors. Researchers need to be
veal specic hypotheses to research aware of both random and systematic
subjects or assistants. It seems safest errors and put into place procedures to
to follow this guideline if expectancy minimize both types of errors.
bias is to be reduced. It may, however, Guideline 18: Report extent of data mas-
be necessary to provide some general saging. Report the number of relation-
idea of what the study is about, but it ships that have been explored in the
is best not to provide either research course of data analysis. Researchers
assistants or subjects with the details. should clarify the number of relation-
Guideline 14: Cover the attitudinal con- ships that have been examined and the
tinuum. When possible, avoid showing reasons why certain ndings have not
your hand by presenting a variety of been reported.
views. This will give the respondent a
The guidelines section is intended to
sense that all responses are accept-
suggest that we need to be more sensitive
able. Be certain to offer a full range of
to bias and to its role in both different
attitudinal response categories so that
types of research projects and different
no respondent is always forced to the
phases of a research project. Predisposi-
extreme of the continuum.
tions to favor certain outcomes will al-
Guideline 15: Be accepting of all re-
ways be present in human activities, in-
sponses. Interviewers must be trained
cluding nursing research. To appreciate
to appear to ask questions neutrally
that simple idea is a good start.
and to respond in the same way to all
respondents answers. Ideally, the in-
terviewer should convey an impression
of neutrality yet have a keen interest E. CONCLUSION: THE GAP
in respondents answers. Interviewers BETWEEN MYTH AND
should avoid coaching responses. REALITY
Guideline 16: Specify data analysis pro-
cedures in advance. In quantitative re- We take the predispositions, or biases, of
search, hypotheses must be specied our culture into our research. We also take
in advance as must data analysis pro- the biases acquired as members of the
cedures to avoid the bias of playing nursing research community to our re-
with the data. Analytical procedures search. So, regardless of whether we are
must be specied in advance of data positivist, interpretive, or critical, (see
collection. Although this may be Chapter 1) in our basic perspective, we
bring a set of assumptions, beliefs, theo- so that you will do good, impartial re-
retical orientations, and expectations to search. Are there any problems with this?
our research. Indeed, a text on methods The problem is that an important part
provides sets of techniques for conduct- of nursing research is missed. Although
ing research and attempts to socialize the most presentations of research sound
student into an understanding and accep- straightforward, there is much that hap-
tance of the latest approaches to re- pens in the course of the research process
search. In their education, research meth- that never gets reported. In short, the re-
ods students are provided with a set of search act itself can be the subject of re-
dispositions, a set of rules not only for do- search. A careful examination of most
ing research but also for judging the work nursing research projects would reveal
of other scholars. In short, the norms of re- how bias inadvertently plays a role in re-
search are being communicated. search outcomes. There is, then, a formal
But the ideas conveyed in a nursing system of sciencesomewhat mythical
methods text focus on the formal system and there is the real world of nursing sci-
of science. You learn how to do research ence. The gap between myth and reality
properly. You learn what techniques are exists in all academic disciplines.
appropriate in any given circumstance. Research is social behavior. There are
And most methods texts urge you not to expectations of others to be met, norms of
fall prey to bias, urge you to be fair and ob- behavior to be followed, and ndings that
jective, and urge you to exercise great care are anticipated. This social component of
Table 9.1 Myths and Realities of Nursing Research

Category Nursing Research Myths Nursing Research Realities
Value free Research is objective and value There are signicant subjective
neutral elements in all research
Stereotype of nursing Nurses primarily use experimental They also do eld studies, surveys,
data and follow the biomedical social action research, and
model of research evaluation research
Sampling Most studies involve representative Most studies are based on
samples nonrepresentative samples
Refusals Most people are willing to participate Refusals run from 0% to 95%; com-
in studies mercial market researchers have
the highest nonparticipation rates
Funding Open to all; based on peer review Researchers who are not part of the
and an evaluation of the quality of university system or a health
the proposal and the research research unit have little chance of
record of the applicant receiving funding; there are fads in
what kinds of projects are funded
Measurement There is agreement on appropriate Little standardization or agreement
way to measure most variables on measures exist
Report writing Final reports summarize the results Evidence is selectively reported;
of the observations some facts ignored or not reported
Tests of signicance They assess the extent to which They are often inappropriately used
results may be the result of chance when nonprobability sampling is
sampling uctuations used or when a whole populations
studied
science is frequently at odds with the fun- actual practices of science. An awareness
damental canons of science. Science as of this gap and of the sources of bias in re-
practiced is neither value free nor wholly search can only benet beginning re-
objective. If actual research practices are search methods students. Just as a good
observed, a whole host of nonscientic scientist is portrayed as a skeptic, so
factors enter the picture. Although the should we be skeptical of the methods of
achievements of science have, indeed, nursing science itself. Table 9.1 provides a
been impressive, it is nonetheless true few examples of gaps between actual and
that there is a gap between the ideal and ideal practices in nursing science.
E X E R C I S E S
1. What is meant by researcher affect, cepted responsibility to provide home
and how might it inuence the conclu- care for her brother who has AIDS. She
sions of a study? is exhausted and resentful of this situa-
tion. What type of bias might the nurse
researcher potentially introduce to
this research study? What safeguards
2. What are the fundamental differences could you build into the study design
between a debater approach to re- to minimize unnecessary bias?
search and an impartial approach?
What cues might alert the student to
the type of researcher being encoun-
tered? 5. Reect on your own nursing experi-
ence and identify a topic that would be
difcult for you to study in an impartial
manner. How would you deal with this
3. Identify a research problem that inter- in a responsible manner?
ests you. Describe the problem
briey. List a minimum of ve predis-
positions that you have concerning
this research problem, indicating your 6. You wish to evaluate a new health pro-
expectations in each case. Propose a motion program designed to increase
way of managing your expectations so the frequency of participation of young
as to minimize their effect on your re- women in regular physical exercise. The
search. program is offered to all women attend-
ing the local university. You hope to use
the ndings to advocate for additional
health promotion programs on campus.
4. A nurse researcher is designing a qual- In regards to this situation, answer the
itative study to assess the impact of following: (1) What role would selection
family caregiving for persons with AIDS bias play in this study? (2) Why is as-
on the children of primary caregivers. sessment of the effectiveness of this
The nurse researcher (a mother of two program problematic? (3) How could
preschooler children) has recently ac- selection bias be avoided? (4) What
would determine if this were advocacy Do you prefer to do aerobic or iso-

research or evaluation research? metric exercises?
7. Examine the following questionnaire Do you support the presidents posi-

items for demand characteristics. Cir- tion on advanced nursing practice?
cle wording that you think is problem-
atic ( if any). Reword items that might
cue the participants to respond in a
manner that would distort responses. Many experienced and expert nurses
How often do you drink and drive in believe they are overworked in to-
a typical week? days health-care environment. Do
you agree?
Cook, A.S. (1997). Investigator bias in be- vention research with practical sugges-
reavement research: Ethical and method- tions for inclusion of women and diverse
ological implications. Canadian Journal of ethnic groups in such projects.
Nursing Research, 29(4), 8793. An interest- Rosenthal, R. and Rosnow, R.L. (1991). Essen-
ing discussion of investigator bias as it re- tials of Behavioral Research (2nd ed). New
lates to the research process. York: McGraw-Hill. Chapter 6 of this book
Crookes, S., and Davies, S. (1998). Research is an excellent discussion of experimenter
Into Practice. Edinburgh, Scotland: Balliere effect bias as well as proposed ways of min-
Tindall. An excellent discussion of the ap- imizing the effect.
plication of research ndings to develop Sterling, T.D., Rosenbaum, W.L., and Weinkam,
good practice. It integrates a discussion of J.J. (1995). Publication decisions revisited:
bias and steps to minimize bias into several The effect of the outcome of statistical tests
chapters in the text. on the decision to publish and vice versa.
Flaskerud, J., and Nyamathi, A. (2000). Attain- The American Statistician, 49, 108112. This
ing gender and ethnic diversity in health in- article is a follow up to Sterlings paper pub-
tervention research: Cultural responsive- lished in 1959 and shows that little has
ness versus resource provision. Advances changed in publication decisions over the
in Nursing Science, 22(4), 115. An overview past 30 years.
of gender and ethnic bias in health inter-
Chapter 10
Ethical Issues in
Nursing Research
CHAPTER OUTLINE
A. What Is Ethics? C. Studies Illustrating Ethical Dilemmas

B. Ethical Decision Making 1. The Tuskegee Syphilis Study
1. Perspectives for Assessing Ethical 2. The Willowbrook Study
Acceptability 3. Unethical AIDS Research
2. Codes and Guidelines for Ethical D. Rules for Conducting Ethical Research
Decision Making
1. Rules for the Treatment of Participants
3. Principles of Ethical Research
2. Rules for Socially Responsible Nursing
4. Methods of Protecting Human Subjects
Research
KEY TERMS
Beneficence Ethics Nonmaleficence

Confidentiality Ethics review board Nuremberg Code
Debrief Informed consent Process consent
Declaration of Helsinki Institutional review board Utilitarian perspective
Deontological Justice
322
ETHICAL ISSUES IN NURSING RESEARCH 323
A
t every stage of the research goodness and badness of human actions
process, there are ethical consider- or with the principles of what is right and
ations for nurse researchers to con- wrong in conduct. The intent of ethics is
sider. For example, research questions to develop methods for clarifying confus-
such as the following are riddled with eth- ing questions even if nal solutions re-
ical challenges: How do I study mentally main elusive. Deloughery (1995) refers to
impaired subjects who may not be able to ethics as a study of the why? of the
give informed and voluntary consent? How moral principles by which a society lives.
can I ethically withhold a treatment from She elaborates that ethics deals with our
subjects that might relieve their suffering? concern for good behavior among people
How can I assess pain levels in children and our struggle to resolve ethical differ-
who are noncommunicative? Even the de- ences among members of society.
cision as to whether a problem should be The authors of this text believe ethics
studied or not has ethical implications for is equally pertinent to all health-care
investigators. Because research is central practitioners regardless of their profes-
to the evolution of nursing care, the avoid- sional designation as nurses, physicians,
ance of conducting nursing research may social workers, dietitians, and so on. How-
be considered unethical. Similarly, the fail- ever, in this chapter, we focus on efforts
ure to engage in evidence-based prac- to distinguish right from wrong in the
ticein other words, to base practice de- process of conducting nursing research.
cisions solely on tradition or intuition The terms ethics and morals are often
rather than on empirical ndingsdenies used interchangeably. Downie and Cal-
clients access to the best available care man (1987) identify three distinct mean-
and thus may also be considered unethi- ings of ethics in the health-care literature.
cal. These include ethics as moral philosophy,
As more and more nurses participate in ethics as ordinary morality, and ethics as
research activities, it is important that the codes of professional conduct. In the sense
profession operate from a sound ethical of moral philosophy, ethics refers to the
knowledge base. This chapter provides branch of philosophy that is concerned
nurse researchers with guidelines to ap- with the study of principles that govern
ply to ethical issues and problems in human behavior in our social world. It is a
health research. Let us begin by exploring theoretical exploration of practical moral-
what ethics is and what ethical decision ity. Its aim is to provide an intellectual
making in nursing research involves. The analysis and understanding of the funda-
chapter then describes several research mental principles of everyday morality. In
studies that illustrate ethical dilemmas the health-care environment, however,
faced by nurse researchers. Finally, the our goal is more than just an understand-
chapter provides a series of rules to guide ing of ethics. We hope to bring about an
nurse researchers in resolving ethical improved state of health through ethical
problems. action. This is true whether the action be
research or the provision of care.
This leads us to a discussion of the sec-
ond usage of ethics as ordinary morality.
A. WHAT IS ETHICS? From this perspective, it relates to the
problems of everyday life. As health pro-
The term ethics has a variety of meanings fessionals, the ethical challenges we en-
and associations. The Oxford American counter in practice are often an extension
Dictionary denes ethics as the study of of the moral problems of everyday life. To
moral principles. It is concerned with the be useful to problem solving, ethics in this
sense must be viewed broadly as includ- ceptability of a study? How do I make an

ing the spectrum of value judgments ethical decision? What are the charac-
about what is good and harmful in the teristics of ethical research? What are
world. Such judgments are all pervasive the rights of those who participate in
and consider a diverse range of factors, in- my study, and how do I protect those
cluding the health-care professionals own rights? This section begins to explore
values. Nurses bring with them their own some answers.
sets of values to professional life. Nurses
need to be aware of what these values are
and be ready to examine them in light of a 1. Perspectives for Assessing
changing health-care environment. Ethical Acceptability
In the third sense, ethics is dened nar-
Two ethical perspectives have been pro-
rowly as a list of dos and donts often
posed in western society to assess ethical
set out in a code of conduct. The nursing
acceptability of a research project: the
profession has several traditional codes
utilitarian and the deontological perspec-
that guide decision making around ethi-
tives. Each of these general perspectives
cal dilemmas, many of which are pro-
articulates ethical norms that transcend
vided later in this chapter. Codes are use-
disciplinary boundaries. They may prove
ful in outlining basic moral principles to
helpful to nurse researchers in reecting
direct professional behavior, but they are
on ethical issues in research situations.
limited in that they cannot address the
full range of moral complexities con-
fronting nurses and other health-care a. The Utilitarian View
professionals today.
The utilitarian perspective suggests that
ethical judgments about a research proj-
B. ETHICAL DECISION MAKING ect should be made by evaluating its con-
sequences for the participants, for soci-
Health-care research, in general, and nurs- ety, and for the academic discipline. This
ing research, in particular, is considered view entails the belief that the good of a
by most to yield positive benets to soci- research project is dened by the conse-
ety. However, the history of health-care re- quences of the results. Classical utilitari-
search teaches us that the pursuit of such anism was advocated by the well-known
a good has been abused by some re- philosopher John Stewart Mill (1806
searchers over the years. Beecher (1966) 1873). He proposed that actions should
and Pence (1990) illustrate the abuses that be taken to promote the greatest good
can result from unscrupulous and over- for the greatest number. An obvious
zealous researchers such as in the Tus- problem with the utilitarian perspective
kegee and Willowbrook studies discussed is that of how to dene good. The utili-
later in this chapter. Nurses are involved tarian perspective considers that the end
in a variety of research efforts, including justies the means. Based on this per-
research specic to nursing, multidiscipli- spective, a research project is judged to
nary research, and research that is con- be ethical if it produces the greatest good
ducted by other disciplines on clients for the greatest number.
receiving nursing care. In all of these situ- Small-scale studies with little ability
ations, nurses are expected to engage in for generalization of ndings to the larger
ethical research behavior. Questions for society may be judged from a utilitarian
ethical decision making abound: What perspective to be unethical. For example,
criteria can be used to judge the ethical ac- research into liver transplantation that
has potential to inuence quality of life pressure to permit the use of unproven
for a small number of participants who drugs or therapies on such patients. In
receive this costly medical intervention the area of AIDS and cancer treatments,
may not be considered ethical unless it we now permit the use of experimental
can be shown that the results can benet therapies because of the imminent death
a large number of people in the long term. of the patient. In one situation, use of
In contrast, research into the prevention such treatments may be viewed as a
of liver disease by lifestyle change is jus- blessing leading to a remission in the dis-
tiable from a utilitarian perspective be- ease process, yet in another situation it
cause lifestyle change is more likely to may have no benecial effect on the dis-
benet many more people than will ease prognosis.
costly surgical procedures available to It needs to be recognized that each re-
only a few. In this perspective, one is left searcher may have a different denition of
with the ethical dilemma of whether to what is good. For example, a cardiac crip-
provide the costly surgical procedure to ple awaiting a heart transplant has a per-
a patient if that is all that is possible. spective on the value and use of cardiac
Would it be ethical to deny that person transplant research that is quite different
the surgery in this case? from the perspective of a health educa-
The utilitarian view emphasizes antici- tion student doing lifestyle teaching with
pating the possibly unfortunate conse- disadvantaged women who smoke. Simi-
quences of the research. Thus, one wants larly, a cardiac surgeon who heads a
to assess the possible dangers to the par- transplant research team may have a dif-
ticipants. Will participation in the study ferent perspective on what constitutes
be in any way degrading, dangerous, or good resulting from expensive cardiac
expose the participants to undue levels surgery research than a hospital adminis-
of harm or stress? Can these adverse ex- trator who is balancing the bottom line.
periences be justied by the studys con-
tributions to our knowledge of health and b. The Deontological View
human behavior? Are we in any way trap-
ping individuals so that they do things The deontological approach to research
that they otherwise might not do? Are we ethics proposes absolute moral impera-
forcing them to take a position on an is- tives that must never be violated. Some-
sue that they have never thought about? one taking the deontological view might
Would participants involvement in the propose absolutes such as never using
study reveal unpleasant, unsavory things deception, always masking the identity of
about them that might otherwise remain participants, and never putting any pres-
hidden? Should research impose un- sure on respondents to participate in a
pleasant and perhaps unwanted treat- study. Rather than assessing the conse-
ments, experiences, or self-knowledge on quences of a given research procedure,
participants? And if this is sometimes the the deontological approach may pro-
consequence, can it be justied by other pose, in its most extreme form, that de-
pay-offs? The utilitarian view permits ception in experiments is never justied,
projects as long as reasonable precau- no matter what the positive contribu-
tions are taken and subjects are de- tions to our knowledge might be.
briefed so that no long-term negative con- Immanuel Kant (17241804) was a well-
sequences result. known deontologist who supported the
In the biomedical area, permission to categorical imperative; that is, basic laws
use unproven therapies on dying patients apply without exception. For example,
is now more readily available. There is telling an untruth is always wrong, re-
gardless of the consequences, or the wel- pected of nurse researchers and provide
fare of the patient must always be placed a standard for nurses working through
above the integrity of the research if a ethical conicts that may arise in re-
conict exists. Some deotologists take a search. They also provide guidance on
more exible view than Kant, accepting ethical decision making in clinical prac-
more than one basic law and considering tice, education, and administration. In the
context and circumstances in determin- 1950s, nurses in many countries adopted
ing priority. Alasdair MacIntyre (cited by the International Council of Nurses Code as
Holden, 1979, p. 538) argued that one can their rst code and then went on to de-
distinguish between types of harm to par- velop unique codes reective of practice
ticipants. He distinguishes between: in their own countries. Boxes 10.1, 10.2,
and 10.3 describe the Codes of Ethics of
Harm to a participants interests (e.g.,
the American Nursing Association (ANA),
reporting a case of venereal disease)
Canadian Nursing Association (CNA), and
Wrongdoing (e.g., lying to someone,
United Kingdom Nursing Association
which may not cause any damage)
(UKNA), respectively.
A moral harm (e.g., doing something to
In addition to these codes, nurses and
make the participant less good, such
others have developed professional guide-
as encouraging the person to tell a lie)
lines for the protection of human subjects
MacIntyre argues that whereas harms while conducting research. Because most
can be compensated for, wrongs cannot nursing studies deal with human subjects,
be; hence, the argument is that if doing a the protection of participants from harm
wrong is essential to conducting a reas a result of research participation is a
search project, then the project should major responsibility of nurse researchers.
be banned (cited in Holden, 1979, p. 538). For example, the CNA published Ethical
For example, if participant recruitment Guidelines for Nurses in Research Involving
into a study requires deception of partic- Human Participants (1994). This document
ipants, then the study should be banned provides nurses in all domains of practice
no matter what the benet to society in with guidelines that are relevant to the cur-
terms of knowledge development. The rent and complex reality of research. The
Tuskegee syphilis study discussed in Box guidelines are based on the relevant value
10.6 powerfully demonstrates the serious statements from the CNA Code of Ethics,
damage that can result when a degree of such as the respect for client choice and
deception is accepted in research. condentiality, protection of clients from
incompetence, and so on.
The ANA has produced two documents
2. Codes and Guidelines for that provide guidelines for nurses en-
Ethical Decision Making gaged in research: The Guidelines for the
Investigative Function of Nurses is summa-
Throughout its history, the eld of nurs-
rized in Chapter 1. The Human Rights
ing has been sensitive to the ethical di-
Guidelines for Nurses in Clinical and Other
mension of doing nursing research. Thus,
Research (1985) focuses on the rights of
many professional nursing organizations
human subjects involved in research and
have developed codes of ethics to guide
the ethical responsibilities of nurses in re-
nurses in their research endeavors. Al-
search settings. The key points of this
though these codes are not totally inclu-
document are listed in Box 10.4.
sive of every ethical concern a nurse re-
The Nuremberg Code (1949) was for-
searcher may encounter, they clearly
mulated as a result of unethical medical
articulate the level of ethical behavior ex-
research experiments conducted by Nazi

BOX 10.3 Code of Professional Conduct for the Nurse, Midwife, and
Health Visitor (UKCC)
Each registered nurse, midwife, and health visi- the privileged access allowed to their prop-
tor shall act, at all times, in such a manner as to erty, residence, or workplace.
justify public trust and confidence, to uphold 9. Respect confidential information obtained
and enhance the good standing and reputation in the course of professional practice and
of the profession, to serve the interests of so- refrain from disclosing such information
ciety, and above all to safeguard the interests of without the consent of the client, or a per-
individual clients and patients. son entitled to act on his/her behalf, except
Each registered nurse, midwife, and health where disclosure is required by law or or-
visitor is accountable for his/her practice, and, in der of a court or is necessary in the public
the exercise of professional accountability shall: interest.
10. Have regard to the environment of care
1. Act always in such a way as to promote and and its physical, psychological, and social ef-
safeguard the well-being and interests of fects on clients, and also to the adequacy of
clients. resources, and make known to appropriate
2. Ensure that no action or omission on persons any circumstances that could place
his/her part is detrimental to the condition clients in jeopardy or which mitigate against
or safety of the client. safe standards of practice.
3. Take every reasonable opportunity to 11. Have regard to the workload of and the
maintain and improve professional knowl- pressures of colleagues and subordinates
edge and competence. and take appropriate action if these are
4. Acknowledge any limitations of compe- seen to be such as to constitute abuse of
tence and refuse to accept any delegated the individual practitioner and to jeopar-
functions without first having received in- dize safe standards of practice.
struction in regard to those functions and 12. In the context of the individuals own
having been assessed as competent. knowledge, experience, and sphere of au-
5. Work in a collaborative and co-operative thority, assist peers and subordinates to
manner with other health care profession- develop professional competence in accor-
als and recognize and respect their contri- dance with their needs.
butions within the health care team. 13. Refuse to accept any gift, favor, or hospi-
6. Take account of the customs, values, and tality that might be interpreted as seeking
spiritual beliefs of clients. to exert influence to obtain preferential
7. Make known to an appropriate person or consideration.
authority any conscientious objection that 14. Avoid the use of professional qualifications in
may be relevant to professional practice. the promotion of commercial products in or-
8. Avoid any abuse of the privileged relation- der not to compromise the independence of
ship that exists with patients/clients and of professional judgement on which clients rely.
SOURCE: UKCC (1993). Code of Professional Conduct for the Nurse, Midwife And Health Visitor. London: United King-
dom Central Council. Cited with permission.
scientists predominantly on inmates of issues of informed consent, protection of

concentration camps before and during subjects from risk or harm, the right of
World War II. This code is appropriate to participants to withdraw from experi-
researchers in disciplines such as nurs- mentation, and adequate qualications of
ing, sociology, psychology, and educa- those conducting research. Box 10.5 lists
tion, as well as those in biomedical re- the key points of the Nuremberg Code.
search. The Nuremberg Code addresses The Declaration of Helsinki (1986)
BOX 10.5 Highlights of the Nuremberg Code
1. The voluntary consent of human subjects is exception would be where the experimen-
absolutely essential. tal physicians also serve as subjects.
2. The experiment shall be such as to yield 6. The degree of risk should never exceed the
fruitful results for the good of society, un- importance of the problem to be solved.
procurable by other methods, and not ran- 7. Proper preparations should be made to
dom or unnecessary in nature. protect subjects against even remote pos-
3. The experiment should be based on the re- sibilities of injury, disability or death.
sults of animal experimentation and a 8. Only scientifically qualified persons should
knowledge of the natural history of the dis- conduct the experiment.
ease or other problem under study that the 9. During the course of the experiment, the
anticipated results will justify the perfor- subject should be able to bring the experi-
mance of the experiment. ment to an end if he or she wishes.
4. The experiment should be conducted to 10. During the course of the experiment, the
avoid all unnecessary physical and mental researcher must be prepared to bring the
harm or suffering. experiment to an end if he or she believes
5. No experiment should be conducted continuation . . . is likely to result in injury,
where there is an a priori reason to believe disability, or death to an experimental sub-
that death or disabling injury will occur; an ject.
SOURCE: Nuremberg Code (1949). In J. Levine (Ed.). (1986). Ethics and Regulation of Clinical Research (2nd ed.,
pp.425426). Baltimore: Urban & Schwarzenberg.
was strongly guided in its development 3. Principles of Ethical Research

by the Nuremberg Code. This declara-
tion differentiates between therapeutic At times, ethical theories such as deontol-
research and nontherapeutic research. ogy or utilitarianism may not be adequate
Therapeutic research is research that to guide nurses in ethical decision making.
provides participants with an opportu- In these situations, ethical principles
nity to benet from an experimental rather than theories may prove more help-
treatment. Nontherapeutic research is re- ful. The CNA (1994) identies three main
search that generates new knowledge principles on which to base ethical deci-
that will bring future benets to society, sions in research situations involving hu-
but those acting as participants most man subjects. These include respect for
likely will not benet from their partici- persons, benecence, and justice. Each of
pation in the research. The Helsinki Dec- these principles warrants discussion.
laration provides three general rules to
guide investigators in conducting non- a. Respect for Persons
therapeutic research:
Inherent in the ethical principle of respect
1. Greater care must be taken to protect for persons are the concepts of autonomy,
subjects from harm in nontherapeutic dignity, uniqueness, freedom, and choice.
research studies. This principle forms the foundation of the
2. The investigator must protect the life participants rights to informed consent,
and health of the participants. privacy, and condentiality. This principle
3. Independent justication is required involves respecting peoples autonomy or
to subject a healthy volunteer to risk of right to choose freely for themselves. Par-
harm just to gain new empirical knowl- ticipants have a right to make choices
edge. based on their own values and beliefs and
A code of ethics applied to social, med- be free from undue pressure or coercion
ical, and health-care research in Canada is by others. Autonomous decisions imply
the Tri-Council Policy Statement on Ethical that participants have enough information
Conduct for Research Involving Humans to make informed choices and they com-
(1998). The Medical Research Council prehend the implications of the choices
(MRC), the Natural Sciences and Engineer- they make. Participants must be made
ing Research Council (NSERC), and the So- fully aware of any risks or benets antici-
cial Sciences and Humanities Research pated from participation in a study. Au-
Council (SSHRC) have jointly adopted this tonomy may also be inuenced adversely
new document, which provides standards by unequal power relationships between
and procedures for governing research in- the participant and the researcher. The
volving human subjects. principle of respect for persons requires
In the new millennium, we see that most that participants condentiality and
professional associations; all universities, anonymity must be safeguarded. The re-
colleges, and hospitals; and most govern- searchers responsibilities related to
ment agencies will have in place codes of these rights are discussed in the rules at
ethics to guide researchers. You should fa- the end of this chapter.
miliarize yourself with the applicable code
for your discipline or place of work. De- b. Beneficence
spite these codes and documents, unethi-
cal research still continues, as you will Two concepts are inherent in this princi-
read in Box 10.6. ple: nonmalecence, or the duty to not
BOX 10.6 What Happened in the Tuskegee Study?
In 1929, the USPHS conducted studies in the final examination they were now to receive was
rural U.S. South to determine the prevalence of a very special one and after it was finished they
syphilis among blacks and explore the possibili- would be given a special treatment if their con-
ties for treatment. Macon County, Alabama, in dition warranted it. Hence, participants contin-
which the town of Tuskegee is located, was ued in the study under the guise of treatment.
found to have the highest incidence among six In 1933, the USPHS decided to continue
counties tested. The study concluded that mass with the study despite the fact that their as-
treatment could be successfully implemented sumption regarding African-Americans atti-
among black Southerners. With the economic tudes toward treatment of the disease proved
collapse of 1929, the findings were ignored. wrong. There is no indication that treatment
Three years later, in 1932, the conditions in was ever considered an option for the study
Macon County were given renewed attention. participants. At this point, it was concluded that
The Chief of the USPHS Venereal Disease Divi- only autopsies could scientifically confirm the
sion decided the high prevalence of syphilis dis- findings of the study. Hence, a further series of
ease offered an unprecedented opportunity to deceptions and inducements were planned to
study the disease in nature. He reasoned that maintain the men in the study so that autopsies
because most blacks went untreated through- could be performed at a later date. To secure
out life, it seemed natural to observe the con- continued cooperation of the men, a black
sequences. This was in the face of contradictory nurse was hired to assess their health and se-
findings in every major textbook of the day that cure approval for autopsies. The nurse offered
clearly advocated treating patients with syphilis noneffective medications (i.e., aspirin and tonic)
even in the diseases latent stages. and provided hot meals and transportation on
The study design called for the selection of the day of their examinations. When it became
black men with syphilis who were between the difficult to convince the men to come to the
ages of 25 and 60 years. A physical examination hospital as their illness progressed, the USPHS
(including radiographs) and a spinal tap were promised to cover the burial expenses of the
done to determine the incidence of neu- men. This was a strong incentive because fu-
rosyphilis. The researchers had no intention of nerals were an important part of the cultural
providing any treatment to the infected study fabric of rural black communities in the South.
participants. The research subjects were di- By 1936, the first reports of the study ap-
vided into two groups: 400 men with untreated peared in the medical press. It was apparent at
syphilis and 200 men without syphilis who this time that the men with syphilis developed
served as a control group. many more complications than did the control
Difficulties arose in recruiting subjects to the group. By 1946, the death rate for the men with
study. Consequently, the men were told they untreated syphilis was twice as high as it was for
were ill and were offered free treatment. The the control group. In 1955, it was reported that
offer of treatment appeared to secure the co- more than 30% of the test group autopsied died
operation of participants. The USPHS did not indirectly from advanced syphilitic lesions of the
form participants that they were part of a re- cardiovascular or central nervous system. In
search study on syphilis. On the contrary, they 1950, it was concluded, We now know, where
were told that they were being treated for bad we could only surmise before, that we have
blooda colloquialism for syphilis. Subjects contributed to their ailments and shortened
were given mercurial ointment, a noneffective their lives. A black physician, Vernal Cave, later
drug, to maintain their interest in the study. The wrote, They proved a point, then proved a
final procedure of the study was a spinal tap to point, then proved a point.
test for neurosyphilis. To secure participants For 40 years, the USPHS denied subjects
participation for the potentially painful spinal treatment and took deliberate steps to see that
tap, a letter was sent to them suggesting that the participants did not receive treatment from
BOX 10.6 What Happened in the Tuskegee Study? (Continued )
other sources. For example, in 1934, local black care of government doctors already and, there-
physicians were sent a list of subjects and let- fore, saw no need to seek additional treatment.
ters requesting them to refer any black man on In 1969, the CDC reviewed the study and
the list back to the USPHS if he sought treat- recommended that it should continue. Only
ment. In 1941, the army drafted a number of one physician argued that the study should be
subjects and began treatment immediately. The stopped and the men treated. In July 1972,
USPHS supplied the draft board with a list of when the study was first published in the na-
256 names they wished to have excluded from tional press and provoked public outrage, data
treatment. The draft board cooperated. Most were still being collected and autopsies per-
men in the study thought they were under the formed.
SOURCE: Adapted from A.M. Brandt (1978). Racism and research: The case of the Tuskegee Syphilis Study. Hast-
ings Center Report, 8(6), 2129.
inict harm, either emotional or physical; complications) but in the process may in-
and benecence, or the duty to promote ict pain and extreme discomfort (bad out-
or do good. To apply the concepts of non- come). In these situations, the researcher
malecence and benecence, a standard must always try to maximize good (early
of care must be provided by researchers. recovery) and minimize harm and discom-
Standards of due care are usually estab- fort by providing good pain assessment
lished by society and by professional role and appropriate medication administra-
expectations. Nurses duties to act as a tion. Munhall suggests that when doubt is
client advocate requires that nurses in- evident, the therapeutic imperative must
tervene if a research intervention pro- take precedence over the research imper-
duces more harm than benet to a client. ative for the nurse as caregiver. This is be-
It is the researchers responsibility to cause nurses always have the responsibil-
minimize risks and maximize benets to ity to act as advocates for clients. The
participants; not doing so would contra- principles of respect for persons and
vene the ethical principle of benecence. benecence, as well as the deontological
When a nurse researcher is a care perspective, support this decision.
provider, the potential for conict be-
tween the duty to care and the duty to ad- c. Justice
vance knowledge may be imminent. Mun-
hall (1988) and Smith (1997) refer to this as The principle of justice requires that peo-
the research imperative (i.e., advancement ple be treated fairly. The protection of
of knowledge) versus the therapeutic im- participants from incompetence and the
perative (i.e., duty to care). Sometimes it is right to receive research treatments are
challenging to determine at what point the expectations of the justice principle. Jus-
therapeutic imperative is being compro- tice is an important principle to follow in
mised by the research imperative. Often the selection of subjects for research
good and harm are not easily separated. studies, particularly in expensive bio-
For example, a nurse researcher who is medical experiments. The risks and ben-
studying recovery patterns in frail elderly ets of a study should be fairly distrib-
patients after hip replacement surgery uted among participants based on the
may achieve a good outcome through subjects efforts, needs, and rights. If pos-
early mobilization of clients (prevention of sible, random selection of participants
avoids potential sources of bias and un- gest that the following information be pro-
fairness in sample selection. vided to potential participants in a re-
Nurses involved in any aspect of the research study:
search process have a responsibility to
1. Title of the study
see that research complies with the prin-
2. Personnel engaged in the study
ciples of respect for persons, benecence,
3. Invitation to participate in a study as
and justice. This is true for nurses own re-
opposed to a request to participate
search as well as for research conducted
4. Reason the individual is being asked
by others that affects their clients.
to participate
5. Clear statement about the studys
4. Methods of Protecting Human purpose
Subjects 6. Detailed description of studys proce-
dures (e.g., what exactly is required of
Two methods of protecting human sub- the participants in terms of time, pro-
jects have been developed over the past cedures, information requested)
three decades. These include the obtain- 7. Potential risks (e.g., physical, psycho-
ing of informed consent and the review of logical, emotional) to participants and
research proposals by an ethics commit- the steps that will be taken by the re-
tee or institutional review board. searcher to protect against the risks
8. Potential benets of the study to the
participants, society in general, and
a. Informed Consent whether or not there are direct sub-
ject benets
Informed consent is based on the princi-
9. Any economic expenses or consider-
ple of respect for persons and incorpo-
ations that may result to subjects
rates the ethical concept of autonomy or
from participation in research
the clients right to self-determine if he or
10. Steps that will be taken to ensure the
she wishes to participate in a research
condentiality of data provided by
study or not. The key elements of in-
participants
formed consent include disclosure of suf-
11. Freedom of participants to ask ques-
cient and appropriate information, un-
tions and to withdraw from the study
derstanding of information, and voluntary
at any time without penalty
participation.
12. The study will be discontinued if bet-
Sufcient and appropriate information
ter treatments become available dur-
implies that the researcher must produce
ing the study or if it is evident that the
a written consent form that ensures that
study treatment is harmful to the par-
subjects know they are participating in a
ticipants or the subjects.
research study as opposed to a normal
13. The institutional review board and
treatment procedure. Subjects must be
participants will be informed of any
18 years of age or older and fully under-
new information that may impact on
stand the information given to them. The
participant safety.
researcher should inform the partici-
14. Subjects refusal to participate in the
pants in writing as well as verbally what
study will in no way jeopardize their
the research protocol entails and what
treatment or otherwise harm them or
criteria were used to select the sample.
detract from their rights in any way.
The research consent form is signed, wit-
nessed, and dated. The sample consent form in Figure
Fain (1999), Jackson (1999), and Martof 10.1 addresses most of the above points.
(2000) are among researchers who sug- Children younger than age 18 years re-
Figure 10.1 Sample consent form. (Continued on next page )
quire a parent or legal guardian to give derstand the information provided to

written informed consent on their behalf. them. It is incumbent on the researcher
Children age 7 years or older may give as- to make sure the participants understand
sent to participate. Figure 10.2 is a sample both the risks involved as well as the ben-
assent form. ets. They must also be fully informed
Comprehension of information is essen- about the research protocols so that they
tial to informed consent. The researcher can make informed choices.
must determine that the participants un- Voluntary participation implies that par-
Figure 10.1 (Continued)
ticipants agree to be part of a study of their other situations, the participants may be-
own free will without coercion of any kind. lieve participation is their only option to
The relationship between the participant receive a particular new therapy or drug.
and the researcher is integral to voluntary In all these situations, the researcher
participation. At times, the participant needs to be clear that an invitation rather
may feel psychologically pressured to par- than a request or expectation has been ex-
ticipate because of the closeness of the re- tended to participants to be part of the
lationship that has developed between the study. The researcher should provide
two. Participants may not want to disap- clearly stated criteria and the rationale for
point the researcher or may be eager to subject selection. This step will protect the
support the work of the researcher. In researcher and the participants. It enables
Figure 10.2 Sample assent form.

the researcher to demonstrate that partic- posal or to terminate unethical studies in

ipants have been fairly selected and the progress. Membership in IRBs is usually
principle of justice enacted. multidisciplinary and designed to ensure
In qualitative investigations, the idea of expertise and independence essential to
process consent is suggested rather than competent reviews. Hospital IRBs may be
traditional consent, which is signed at the dominated at times by physicians who
beginning of a study and not reconsidered predominantly operate from a quantita-
unless participants raise a concern about tive research perspective. This presents
continued participation (Munhall, 1988). a unique challenge to nursing proposals
In process consent, the researcher rene- that use qualitative methodologies and
gotiates the consent with the participants may not be well understood by all mem-
as unforeseen circumstances arise. This bers of the IRB. Nurses have a profes-
allows the participants to be part of the sional responsibility to be members of
decision making as the study unfolds. IRBs and educate members about nurs-
ing approaches to research. Their main
b. Institutional Review Boards role is to make sure that the rights of hu-
man subjects are not violated.
Research involving humans should be sub- In developing a timeline for the project,
jected to approval by an ethics committee the researcher needs to consider the time
commonly referred to as an institutional it will take to obtain ethical approval for a
review board (IRB) or an ethics review study so that delays can be avoided. Most
board (ERB). The purpose of the review is review boards have regularly scheduled
to ensure that ethical principles are ap- meeting dates that are published in re-
propriately applied to research involving search newsletters. Researchers should
human subjects. Most IRBs reviewing bio- inquire as to what deadline date their sub-
medical research operate under the guide- mission requires in order to meet their an-
lines of the Declaration of Helsinki. ticipated start date. Some projects may
Most institutions, hospitals, govern- qualify for an expedited review that sig-
ment agencies, and universities in which nicantly reduces the time requirement.
research involving human subjects is car- In an expedited review, the chairperson
ried out require specic information to and at least one member of the IRB must
make informed and responsible deci- read and approve the proposal. Examples
sions regarding the ethical acceptability of projects that may receive an expedited
of a proposal. The CNA (1997) and the review include interviews, surveys, stud-
ANA (1985) suggest that the following ies of existing records in which the data
type of information should be included in sought are not of a sensitive nature, and
a submission to an IRB: situations in which the use of noninvasive
procedures are routine.
The purpose of the study and its sig-
nicance to health research
The study design and methodology
C. STUDIES ILLUSTRATING
Potential research participants and a
ETHICAL DILEMMAS
description of the selection process
Risks and benets to the participants
This section describes three studies that
Steps to ensure participant choice, con-
illustrate some of the ethical dilemmas
dentiality, and anonymity protection
faced by researchers despite the exis-
Areas of potential conict of interest
tence of methods of protecting human
Most IRBs are mandated to approve, subjects and of guidelines such as pro-
reject, or propose modications to a professional codes of ethics, the Helsinki
Declaration, and the Nuremberg Code. Af- ndings are summarized in the Belmont
ter this review, a series of rules is pre- Report.
sented to guide researchers in resolving In 1980, the Department of Health and
ethical problems that may emerge. Human Services (DHHS) developed a set
There are a number of studies that have of regulations based on the commissions
become landmarks in the discussion of re- recommendations. These regulations are
search ethics. Two of these studies came regularly revised to keep pace with the
to light in the 1970s and have become clas- changing health-care and research envi-
sic examples of unethical research. The ronment. They currently include:
rst is the Tuskegee Syphilis Study, which
1. Requirements for and documentation
began in 1932 and continued until 1972
of informed consent
when an account of the project in the
2. Membership of IRBs
Washington Star provoked public outrage.
3. Criteria for IRB approval of research
The project followed the natural course of
and exempt and expedited review pro-
syphilis in 400 black men who were un-
cedures for certain kinds of research
treated and remained so over the course
of the study, even after the efcacy of peni- Although many studies do not pose se-
cillin to treat the condition was well rious ethical dilemmas for nurse re-
established. The second study is the Wil- searchers, it is important for beginning
lowbrook study (1950s1970s) in which researchers to understand the impetus
mentally challenged children institutional- behind the development of the codes of
ized at the Willowbrook Institution were ethics and research guidelines adopted
deliberately infected with the hepatitis by various professional and governmen-
virus. More recently, AIDS research deny- tal agencies.
ing azidothymidine (AZT) to women in
African, Asian, and Caribbean nations
(1997) has raised concern about the ethi- 1. The Tuskegee Syphilis Study
cal acceptability of research funded by the
National Institutes of Health (NIH) and the We begin this section by examining a
Centers for Disease Control and Preven- study that raises important ethical is-
tion (CDC). sues. In 1932, the U.S. Public Health Ser-
The Tuskegee and Willowbrook stud- vice (USPHS) began an experiment in
ies had substantial impact on the devel- Tuskegee, Alabama, to determine the nat-
opment of regulations governing biomed- ural course of untreated, latent syphilis
ical and behavioral research. In 1973, the in African-American males. The study in-
U.S. Department of Health, Education, cluded 400 men with syphilis and 200
and Welfare (DHEW) published its rst uninfected men who served as control
set of regulations on the protection of hu- subjects. The study is now commonly re-
man research subjects as a result of the ferred to as the Tuskegee Syphilis Study.
Tuskegee atrocities. They also published The study continued for approximately
additional guidelines to protect vulnera- 40 years, through World War II, when a
ble participants such as mentally chal- number of men were called up for the
lenged, ill, and dying individuals. In 1978, draft and would have been treated for
the National Commission for the Protec- syphilis except that they were study par-
tion of Human Subjects of Biomedical and ticipants, and through the 1950s, even af-
Behavioral Research was established to ter the effectiveness of penicillin treat-
develop ethical principles to underlie re- ment was conrmed. The study survived
search on human subjects and formulate the 1960s and was untouched by the
guidelines based on these principles. Its Nuremberg Code, the Declaration of Hel-
sinki, the Civil Rights Movement, and the and informed consent. The report failed to
USPHS Code of Research Ethics. This re- recognize that the entire study had been
search begs the question of how could predicated on nontreatment. Administra-
such a seemingly unethical investigation tion of medication would have violated the
be approved, funded, and allowed to con- rationale of the experimentto study the
tinue for so long at such cost to partici- natural course of the disease until death.
pants and society? A summarized excerpt The report also failed to expose the criti-
from the Hastings Center Report of the cal fact that not only was voluntary and in-
Tuskegee Syphilis Study is included in formed consent not obtained but rather,
Box 10.6. the men were told and they believed that
Published accounts of the Tuskegee they were getting free treatment from ex-
study in the national press in 1972 pro- pert government doctors. According to
voked both critical comment and support Brandt, the failure of the report to expose
from members of the scientic commu- this critical fact calls into question the
nity. Many have opposed the research on credibility of the investigation.
ethical grounds; the statements of Brandt Kampmeier, cited by Brandt (1978),
(1978), Rothman (1982), and Caplan and wrote a spirited defense of the Tuskegee
colleagues, (1992) are particularly inci- study in the Southern Medical Journal in re-
sive. The following quotation from Brandt sponse to the HEW Final Report. His argu-
provides a sense of his position: ment centered on the limited knowledge of
effective therapy for latent syphilis when
. . . The Tuskegee Syphilis Study Ad Hoc
the experiment began. He argues that by
Advisory Panel Report ignores many of
1950, penicillin would have been of no
the essential ethical issues which the
value to these men and suggests that the
study poses. The Tuskegee study re-
men were fortunate to have been spared
veals the persistence of beliefs within
the highly toxic treatments of earlier peri-
the medical profession about the nature
ods. The USPHS maintained that the study
of blacks, sex, and disease-beliefs that
was ethical on the grounds that it consti-
had tragic repercussions long after their
tuted a study of nature. The USPHS in-
alleged scientic bases was known to
sisted that Macon County was a ready-
be incorrect. Most strikingly, the entire
made laboratory and the researchers
health of a community was jeopardized
would only be watching the inevitable.
by leaving a communicable disease un-
They assumed subjects were not going to
treated. There can be little doubt that
obtain treatment anyway and there was,
the Tuskegee researchers regarded
therefore, no reason to miss the opportu-
their subjects as less than human. As a
nity to trace the effects of their disease
result the ethical canons of experi-
process.
menting on human subjects were com-
The Department of Health, Education,
pletely disregarded.
and Welfare (HEW) Final Report (1973) fo-
. . . The injustice committed by the
cused on two ethical objections to the
experiment went well beyond the facts
Tuskegee study. These include:
outlined in the national press or the
HEW Final Report. The degree of de-
1. The study violated the ethical rule that
ception and damages have been seri-
a person should not be subjected to
ously underestimated.
avoidable risk of death or harm unless
Brandt goes on to argue that the nal re- he or she freely consents in an in-
port betrayed a basic misunderstanding of formed manner.
the experiments purposes and design by 2. The study was scientically unsound
focusing on the issues of penicillin therapy and its results meager in comparison
with the known risks it imposed on the may benet them. Nonetheless, the study
subjects involved. raised many agonizing ethical questions
that still remain unanswered.
There appear to be ve key points in the
Tuskegee study that raise ethical issues:
2. The Willowbrook Study
1. The researchers did not obtain in-
For a period of approximately 20 years
formed consent from the participants.
(1950s1970s), mentally challenged chil-
2. The researchers deliberately withheld
dren at the Willowbrook Institution in
information from participants about
Staten Island, New York, were participants
their involvement in a research study
in an unethical research investigation de-
and deceived them into thinking they
signed to improve the understanding of
were receiving effective treatment for
hepatitis. The study was directed toward
bad blood.
determining the period of infectivity of in-
3. Unknown to the participants, the re-
fectious hepatitis. All of the subjects were
searchers denied them penicillin when
mentally handicapped children who were
it became available in the 1940s as an
given either intramuscular or oral doses of
effective treatment and took deliber-
the hepatitis virus. Saul Krugman of New
ate steps to prevent them from getting
York University was the head of the re-
treatment from other sources.
search team. His team systematically in-
4. Their names were circulated on a list to
fected new child residents with the hepati-
various physicians, treatment sources,
tis viruses over the 20-year period of the
and the U.S. draft board; therefore, par-
study. Parents were forced to give permis-
ticipants rights to condentiality and
sion for their children to be study partici-
anonymity were violated.
pants if they wished to gain their childs ad-
5. Participants did not voluntarily partic-
mission to the institution. Parents gave
ipate but rather were coerced under
consent for the administration of the virus,
the guise of free medical treatment, as-
but they were not told the appreciable haz-
sessment by expert government doc-
ards involved. Parents were told that their
tors, free hospitalization, and provi-
children would be placed on the Krug-
sion of burial services.
man ward that was cleaner, better super-
The exposure of the Tuskegee study by vised, and had a higher nurse-to-patient ra-
the national press had a positive inuence tio than the general wards. They were not
on the development of guidelines to pro- told of the debilitating ulike symptoms,
tect the rights of human subjects in bio- skin rashes, liver damage, and resulting
medical and behavioral research through morbidity and mortality patterns that
the establishment of the National Com- many of the children eventually experi-
mission for the Protection of Human Sub- enced. During the 20-year investigation,
jects of Biomedical and Behavioral Re- the Willowbrook Institution closed its door
search. It also raised signicant questions to new admissions because of overcrowd-
about professional self-regulation and sci- ing, yet the research ward continued to ad-
entic bureaucracy. It pointed out the mit new residents. The institute denies this
need for greater vigilance in assessing the claim, referring to documentation of new
specic ways in which social values and admissions on the research unit as admin-
attitudes affect professional behavior and istrative errors.
raised awareness of how groups already This experiment and the study protocol
disadvantaged in society need to be pro- were well publicized in medical literature.
tected from such unethical investigations The study was documented among the 22
as well as included in clinical trials that examples of unethical investigations pub-
lished by Beecher (1966) in the New Eng- lation of the parents to secure consent
land Journal of Medicine. Nevertheless, this of the subjects if they wished to have
investigation was allowed to continue for their child admitted to the Willowbrook
20 years, even after the efcacy of gamma Institution and the withholding of vital
globulin to weaken (if not prevent) an at- information about the appreciable haz-
tack of hepatitis had been established. The ards of the experiment. Parents were
Krugman team defended the investigation not fully informed about the potential
as a natural experiment. They argued consequences of the inoculation or the
that residents would most likely contract types of harm that could come to their
the disease naturally within a few weeks of children as a result of the disease
admission because of the substandard hy- process. It was apparent that the rights
gienic conditions and staff shortages that of the subjects guardians to informed
made contagion probable. Furthermore, consent were agrantly disregarded.
they believed the study would bring im- 2. In the Willowbrook investigation, the re-
mediate advantages to the participants search team continued with the study
such as provision of care on a cleaner, bet- even after the discovery of gamma glob-
ter supervised treatment ward, a higher ulin to weaken or prevent an attack of
nurse-to-patient ratio, and closer supervi- hepatitis was publicized. This action
sion of their condition with fewer opportu- clearly violates the resolution adopted
nities for complications. by the World Medical Association,
Although there is a valid category of re- which states: Under no circumstances
search called natural experiments in the is a doctor permitted to do anything
health sciences, the question remains: which would weaken the physical or
Was the Willowbrook experiment an ethi- mental resistance of a human being ex-
cal study in nature? Rothman (1982) an- cept from strictly therapeutic or pro-
swers this question succinctly. He explains phylactic indications imposed in the in-
that when no cure or treatment exists for a terest of the patient. The risk to the
disease, then observing its course with the participants was clearly too high and
informed consent of participants is a study not outweighed by benets for the par-
in nature. In his opinion, the Willowbrook ticipants or society. Clearly the dictum
experiment does not belong in this cate- that one must not inict harm on one
gory of research because there is an es- group for the advantage of another was
sential difference between taking advan- violated by this study. Social depriva-
tage of social, as opposed to biological, tion and continued predictions of it (i.e.,
conditions. Poverty, ignorance, lth, and that Willowbrook residents would con-
institutional miseries are not in any way tract hepatitis naturally anyway) be-
comparable to the inevitable course of a came a self-fullling prophecy in this sit-
mysterious disease or the unknown risks uation.
posed by carriers. Indeed, the Tuskegee
and Willowbrook experiments offer both
practical and principled support for main- 3. Unethical AIDS Research
taining as rigid a distinction between so-
cial deprivation and biological conditions More recently, there has been contro-
as possible (Rothman, 1982, p. 6). versy raised by Public Citizen (1997), a
The ethical issues raised by this inves- consumer group afliated with Ralph
tigation warrant comment. They include: Nader, that has criticized U.S. AIDS re-
search as being dangerously awed. It
1. The informed consent process was sus- has claimed that the United States is con-
pect on two levels, namely, the manipu- ducting unethical research that has
caused the deaths of approximately 1000 expenditure on women is about $10 per
children overseas. At issue are nine stud- year, and so it is important that any results
ies involving approximately 9000 women of this investigation be relevant and af-
from African, Asian, and Caribbean na- fordable.
tions that are involved in AIDS research This investigation raises ethical con-
trials in which some of the women are be- cerns because of parallels with the Tus-
ing denied access to AZT; instead, they kegee and the Willowbrook studies. It
are being given placebos. reinforces the important need for institu-
The studies are underwritten by either tional review boards to review studies
the NIH or the CDC. Based on federal gov- conscientiously and for researchers to act
ernment documents, the Public Citizen ethically. Researchers need to be cautious
claims that the U.S. government is delib- that they do not place themselves in ethi-
erately denying AIDS-infected women ac- cally untenable positions by engaging in
cess to AZT, a drug that U.S. studies have research that builds on social deprivation
shown clearly reduces the transmission of others. Rothman (1982) cautions that as
of the AIDS virus from pregnant women to soon as researchers attempt to take ad-
their unborn babies by two-thirds. Simi- vantage of the social predicament in
lar studies involving U.S. women offer ef- which participants are found, they be-
fective anti-HIV therapy to all partici- come accomplices to the problem, not ob-
pants. Sidney Wolfe, Director of Public servers of it. This is because researchers
Citizen, claims, It is Tuskegee part two. usually have the ability to alter the social
Dr. Peter Lurie of Public Citizen states, deprivation of the study participants, al-
Its certainly as bad as anything that has though not the larger class that they rep-
occurred since World War II in terms of resent. Could the research team have ad-
the violations of the basics of medical ministered AZT or a comparable drug to
ethics. He goes on to comment, It is a the participants in the study? If re-
clear-cut double standard, and for the ba- searchers approach providing fair and eq-
bies involved it will be lethal. We are de- uitable treatment to participants with the
manding today that the federal govern- same zeal that they pursue their experi-
ment provide AZT or similar drugs to all ments, it may be possible to secure funds
of the women in these studies. or grants to do so. Rothman concludes,
The claims of the Public Citizen sparked Where the essential cause of a health
debate by U.S. ofcials. In defense of their problem is social deprivation, it is gener-
work, U.S. researchers claim it is not feasi- ally within the power of the research team
ble or scientically proper to give AZT to to remedy the situation for their subjects.
all patients involved with the research. Hence, they cannot be observers to a
Philip Nieburg, an AIDS researcher with plight they can improve (p. 7).
the CDC, claims that AZT is unavailable
and unaffordable in developing countries.
He stresses the importance of comparing
new treatments with the existing stan- D. RULES FOR CONDUCTING
dard of care in a particular country and ETHICAL RESEARCH
concludes that AZT is not the standard of
care in developing countries. The govern- The following ethical rules are presented
ment insists a placebo comparison is the as guidelines and are organized around
only way to prove potential new therapies two themes: (1) nurse researchers ethi-
are better than no treatment. Helen Gayle cal responsibility to participants and (2)
of the CDC defends the study, stating that nurse researchers responsibility to the
in the countries involved, the health-care discipline of nursing and society.
1. Rules for the Treatment of In qualitative investigations in which

Participants sample sizes are limited and the depth
of interviewing signicant, it becomes
Rule 1. Protect the condentiality of par- challenging to protect the condential-
ticipants. The researchers promise to ity of the participants. This is particu-
respect the condentiality of partici- larly problematic when one is con-
pant responses to a questionnaire or an ducting research in areas that require
interview or the identity of a participant the researcher to probe deeply into
in an experiment or eld study is to be the life circumstances of the partici-
treated as a sacred trust. It is based on pants. Direct quotations, rich in detail,
the ethical principal of respect for per- are often used to answer the research
sons. Most surveys, interviews, experi- question. The researcher must be vigi-
ments, and eld studies are completed lant about not revealing information in
on the understanding that individual re- the research report or publications
sponses and information that would that result from it that may reveal the
permit the identication of the individ- identities of participants. Pseudonyms
ual will never be released. Researchers should be used in both the data collec-
have not only an ethical responsibility tion process and in the transcription of
to preserve the anonymity of partici- interview tape recordings so that the
pants but also have a practical interest participants true identities cannot be
in doing so: Their ability to collect ac- associated with the data. The re-
curate information would be impaired if searcher should also periodically re-
the public believed that responses mind participants that they are free
were not kept in condence. not to respond to questions or probes
When it is necessary to identify indi- that they believe are inappropriate. If
vidual names with particular question- someone other than the researcher is
naires (e.g., as in panel studies) number transcribing the taped interviews, par-
codesrather than names should be ticipants have a right to know this and
used. The questionnaires and a master be assured that their identity will be
list listing names and identication protected.
numbers can then be stored separately. Rule 2. Do not place pressure on partici-
Such master lists should be destroyed pants. No pressure should be placed on
after the study has been completed. participants to cooperate in a study. Re-
If data are released to other respondents must feel free to refuse par-
searchers, steps should be taken to ticipation, to withdraw at any time, or
mask the individual identities of par- to refuse to answer any particular ques-
ticipants. This can be achieved by re- tion. Researchers should not put pres-
moving highly specic identiers such sure on participants or cajole or harass
as area of residence, specic job, or them in an effort to get cooperation
employer identications. with the study. This maxim is based on
The Tuskegee study clearly violated the principle of respect for person.
the participants condentiality. Not Although it is appropriate in mail
only were the identities of participants surveys to follow up with those who do
in the experimental group known, but not respond with a phone call, these
lists containing their names were de- contacts should provide prospective
liberately circulated to potential treat- participants an opportunity to seek
ment centers and the U.S. draft board additional information about the proj-
to prevent participants from seeking ect to help them decide if they wish to
treatment elsewhere. participate.
Participants should not be presented be a central concern of researchers. It

with either excessive rewards or threats is based on the principle of bene-
of punishment to secure their participa- cence, which requires that one should
tion. To do so violates the participants do good and, above all, do no harm.
right to self-determination. The Willow- This requires the researcher to create
brook study appeared to place a lot of a situation in which the benets to the
pressure on parents to consent to have respondents outweigh the risks or
their children inoculated with the he- harm that may result from the study.
patitis virus. Indeed, some say it was the Harm can take many forms such as
only way a resident could be admitted physical, emotional, social, spiritual,
to the overcrowded institution after the or nancial forms.
general wards were closed. The partici- The Willowbrook study has been
pants in the Tuskegee study were ca- faulted for its adverse effects on the
joled over a period of 40 years to coop- participants health state. Should the
erate with the syphilis investigation. investigators have injected a virus into
This included repeated promises of free the bodies of children that would no
treatment, supervision by medical ex- doubt produce hepatitis and other ad-
perts, free hospitalization, transporta- verse reactions in the participants?
tion to clinic assessments, and paid Should the experiment have been
burial services. In both these investiga- aborted after the efcacy of gamma
tions, it appears pressure was inten- globulin to weaken, if not prevent, an
tionally placed on participants to coop- attack of hepatitis was conrmed? The
erate. same question must be asked of the
Rule 3. Make the subjects participation Tuskegee study after the efcacy of
painless and free from discomfort penicillin as a treatment for syphilis
and harm. Completing a questionnaire was established. Other researchers
or interview or participating as a sub- have decided to abort projects when it
ject in an experiment should be a safe was clear that the study had potential
and painless experience for the partic- dangers to participants.
ipant. Researchers must not expose Piliavin and Piliavin (1972), for ex-
participants to needlessly long experi- ample, aborted trials on the bystander
mental trials or questionnaires or ask intervention studies after it was clear
questions that pry unnecessarily into that there were dangers involved in
personal matters. This constraint is staging phony emergencies. The re-
not meant to suggest that researchers searchers stopped one study when it
should not examine certain areas of was observed that bystanders, seeing
human behavior but suggests that a person fall with blood trickling from
only relevant information be collected. his mouth, activated an emergency
(Often when people are working on stop alarm. It was simply too danger-
their rst project, they suggest many ous and disruptive to continue.
variables; however, when asked what Professor C. Boeck, Chief of the Oslo
they would do with the information, Venereal Clinic between 1890 to 1910,
they are unable to articulate a mean- when the famous Oslo study of un-
ingful analysis.) Moreover, considera- treated patients with syphilis was ongo-
tion should be given to using alternate, ing, is another scientist who abandoned
or indirect, indicators for the ques- a study in progress. As lead researcher
tions that may offend participants. in the world famous Oslo Study of un-
Maintaining the safety, comfort, and treated patients with syphilis, Boeck
self-esteem of the participants should had withheld treatment to nearly 2000
patients infected with syphilis. He was to know if the selection was made by
convinced that current treatments with means of a probability sampling proce-
mercurial ointment were of no value. Af- dure or if it was based on special char-
ter arsenic therapy was discovered and acteristics (e.g., age group, health sta-
made widely available in 1910, he aban- tus, medical diagnosis, membership in
doned the study (Clark and Danbolt, a particular organization, the job one
1955). has) related to the research problem or
Rule 4. Identify sponsors. There should phenomenon under investigation. Ran-
be no deception concerning the spon- dom selection can eliminate some of
sorship of a project. Participants must the researcher bias that may inuence
be informed about who is doing the subject selection. Fair selection of par-
study. On this basis alone, they may ticipants is based on the ethical princi-
choose not to participate in the study. ple of justice, which states that every
Many people have opened their person should be treated fairly and
doors to a person claiming to be doing given what they are due.
a survey on satisfaction with delivery Rule 6. Place no hidden identication
of health-care services, only to dis- codes on questionnaires. Researchers
cover that they are not dealing with a should not use hidden codes on ques-
reputable scientist but rather with a tionnaires to assist in the identication
door-to-door health insurance sales- of those who have or have not returned
person. Feel free to encourage the re- questionnaires. Such codes are some-
searcher to leave rapidly! Similarly, times used to enable researchers to
an individual may choose to not par- nd out who has not returned a ques-
ticipate in a study conducted by an or- tionnaire; such subjects may be sent
ganization whose aims and objectives another request. Although such codes
are unacceptable to him or her. may save the researcher much time
Rule 5. Disclose the basis on which par- and money, they are unethical. If indi-
ticipants have been selected. The con- viduals are to be identied, this infor-
sensus in the nursing research com- mation should be placed directly on the
munity is that participants have the questionnaire itself (not hidden) and
right to know how they were selected discussed in a cover letter. For exam-
for participation in a study. In the past, ple, a questionnaire may contain an
unfair sample selection resulted be- identication code number such as 001.
cause of social biases or biases on the The participant must be informed that
part of the researcher or funding this number is used only by the re-
source. Biases take many forms in sam- searcher to associate that particular
ple selection. They may include treat- questionnaire with a specic individ-
ing subjects poorly with little regard ual. The identication codes and corre-
for their safety or providing benets sponding participant names should be
unfairly to subjects that could offer kept in a secure area to protect the
power or money to researchers. The identity of participants. Only the re-
Tuskegee Syphilis Study is an example searcher should have access to this in-
of how racial and social biases played a formation.
role in unfair sample selection. Rule 7. Honor promises to provide par-
The prospective participant should ticipants with research reports. When
be given a reasonable amount of infor- an offer to provide a research report to
mation about the selection process on respondents has been given, the
which a decision to participate can be promise must be fullled. The relation-
made. For example, respondents need ship between the researcher and par-
ticipants or respondents should be participants ability to comprehend

reciprocal. In practice, it is to the ad- the implications of participating in a
vantage of the researcher to fulll such research study. It is not enough to pro-
obligations because doing so will en- vide appropriate and sufcient infor-
courage the continued cooperation of mation; the researcher must also en-
participants in long-term projects. sure that participants understand the
When individuals are offered the op- information and know they have the
portunity of receiving a report on the freedom to participate or refuse to par-
study, a separate stamped envelope ticipate in the study. The concepts of
and request form should be provided self-determination and autonomy are
for the respondent. The separate return central to the principle of respect for
envelope not only keeps the respon- person. This principle forms the foun-
dents name disassociated from the dation for informed consent.
questionnaire but also conveys to the With studies involving children, ill pa-
respondent that the researcher takes a tients, or incompetent adults, the orga-
promise of condentiality seriously. nization or individual responsible for
Although there may be no ethical re- the prospective participants should
sponsibility to pay participants who provide consent in writing. In addition,
volunteer their cooperation, such pay- children age 7 years or older should
ments are encouraged in order to rein- give assent to participate. Assent means
force the idea of a reciprocal contract a child has voluntarily agreed to be a
between researchers and participants. participant in a research study. By age 7
Participant observation studies fre- years, children are capable of concrete
quently involve rather long periods of operational thought and, therefore, can
observation, and it is particularly im- provide meaningful assent to partici-
portant to reciprocate by providing a pate. As children age, they should be
payment or a report on the research given more responsibility in the consent
project. process.
Rule 8. Informed consent is a key con- The issue of informed consent may
cern. In dealing with competent adults, be a problem for researchers doing par-
participation should be based on in- ticipant observation studies. It would
formed consentthat is, potential par- not be reasonable to insist that all mem-
ticipants must be informed about the bers being observed would have to con-
nature of the study, what kinds of is- sent to being observed. The test in such
sues will be explored, how participants cases perhaps should be (1) whether
were selected, what the risks and ben- sufcient steps are taken to protect the
ets to the participants are, and who is anonymity of those who are observed
sponsoring the research. Moreover, and (2) that no negative consequences
prospective participants should be could reasonably be seen to result from
informed about condentiality and the activity of the research project.
anonymity and they should feel free to To put a blanket prohibition on such
withdraw from the study at any time studies would mean that works such
without prejudice if they wish to do so. as Erving Goffmans Asylums, Margaret
In surveys, respondents should be told Meads Sex and Temperament in Three
to feel free to skip any questions that Primitive Societies, or Madeleine Lei-
they consider inappropriate or ignore ningers Transcultural Theory of Caring
interview questions that make them could not have been done and the ben-
uncomfortable. et derived from such work would
Implicit in informed consent is the have been lost to society.
Naturalistic observational studies, no justication for failing to inform par-

such as those observing whether the ticipants of how they will be assigned
drivers of automobiles have their seat to groups. Such information is essen-
belts fastened, can be performed as tial to informed consent to participate
long as there is no apparent danger to in the study (Neuberger, 1992).
the subject being observed or to the Anytime selective nondisclosure of
observer. In such cases, it is not possi- information or deception is used in a
ble to gain the permission of the per- research study, the researcher must
son being observed. When deciding if question how acceptable it is to do so.
the study should be undertaken, the With deception, the risks to the partic-
best test the researcher can make is an ipants should be minimal and out-
attempt to balance any good that may weighed by the benets. Beauchamp
come out of the project against its po- and Childress (1989) identify four con-
tentially negative consequences. ditions under which deception may be
Some researchers believe covert considered justiable in research:
methods are sometimes appropriate in 1. Deception is essential to obtain im-
social science research (Warwick and portant information.
Douglas; cited in Punch, 1986, p. 29). In 2. No other substantial risk is involved
some situations, the job of the re- to participants.
searcher is to expose the powerful who 3. Other ethical principles are not vio-
prey on weaker members of society. lated such as benecence or non-
This occurs, at times, in research with malecence.
AIDS patients and victims of violence, 4. Subjects are informed that decep-
child abuse, drug addictions, and vari- tion is part of the study before they
ous forms of deviant behavior. How- consent to participate.
ever, researchers traditionally have Rule 9. Debrief subjects. When experi-
been harshly criticized for acting as un- ments or eld studies involve decep-
declared participant observers. Laud tion, subjects should have the study
Humphreys (1970) study of male ho- explained to them after the session.
mosexual behavior in which he inl- The researcher should note what de-
trated the gay community as a covert ception was used and why it was nec-
observer is an example of such decep- essary, and the subjects should be re-
tion. assured that their participation was
In experimental studies, particularly appreciated and helpful.
those involving some deception, there
is a problem concerning informed con- 2. Rules for Socially Responsible
sent. If the experimental manipulation Nursing Research
requires deception, it is not possible
to inform the subject fully, in advance, Nurse researchers have ethical responsi-
of the deception: to do so would spoil bilities not only to the participants in a
the study. In such cases, Rule 9 needs project but also to society at large and to
to be applied. the nursing discipline. Research-minded
In randomized clinical trials (RCTs), practitioners, educators, administrators,
it is not necessary to inform partici- nursing students, and other health-care
pants if they have been assigned to a professionals expect nurse researchers to
treatment or control group; however, it do ethical research. They depend on the
is necessary to inform participants that results of nursing studies to shape their
the research involves use of a placebo practice and promote excellence in nurs-
or nontreatment group. There is also ing care. This situation places a consider-
able ethical responsibility on nurse re- being scanned for interesting ndings,
searchers to conduct unbiased studies these cannot be reported unless the
and report results in trustworthy manners. process by which they have emerged is
made absolutely clear.
Rule 10. Researchers should distinguish Rule 12. Be aware of potential sources of
between science and advocacy. In bias. Becoming aware of the sources of
your role as a nurse scientist, do not bias may help you avoid bias in your
work on projects in which you are own work and spot it in the research
asked to develop a scientic case for reports of others. Researchers have an
a conclusion. Given the legitimacy of ethical responsibility to report their
science in Western culture, it should work fairly, attempting to avoid bias as
be no surprise that both scientists and much as possible. Review the discus-
nonscientists are tempted to use this sion of sources of bias in Chapter 9.
legitimacy to achieve personal or Rule 13. Represent research literature
group goals. Evidence that is viewed as fairly. In the interest of objectivity and
scientic carries a lot of weight in ar- ethics, researchers must attempt to ac-
gument; therefore, when a presenta- curately portray the body of literature
tion is being prepared, collecting or rein their area of research. Reporting
ferring to scientic evidence is ndings selectively is not acceptable.
tempting and sensible. Rule 14. Do the best research you can.
If you are hired to develop a scien- Research must strive to be competent
tic case for . . . then you have an eth- and impartial, and its results must be
ical problem. You are being asked to reported objectively. Use qualied per-
provide scientic evidence to convince sonnel and consultants. Keep up with
others of a particular position. You are, developments in your eld and use the
in this case, being asked to make oth- best techniques of data collection and
ers believe that you have scientic evi- analysis. Always seek to do the best re-
dence for the particular position. Such search you can; do research with care.
research should not be presented as Rule 15. Acknowledge all your sources.
science; to do so would be unethical. Acknowledge people who have played
As a citizenadvocate, it is perfectly a role in your research and acknowl-
legitimate to comment on the evidence edge all literature sources that have di-
of others and to produce evidence that rectly inuenced your study. Excluding
is appropriate to the issue under dis- the respondents (who have been as-
pute. It should be noted, however, that sured of anonymity), all other people
the information is being presented who have assisted in the project should
from the perspective of a citizen- be acknowledged by way of footnotes in
advocate, not from that of an expert your report. Similarly, when literature
nurse scientist. We would be well ad- has been used in developing the proj-
vised to maintain a distinction be- ect, each source should be cited.
tween citizenadvocate roles and re- Rule 16. Seek advice on ethical issues. If
searcher roles. Too often the line ethical issues arise, seek the advice of
between them is blurred. appropriate professional bodies or in-
Rule 11. Do not hunt through data look- stitutions involved in the project. Most
ing for pleasing ndings. The surest studies do not pose difcult ethical is-
way to be guilty of misrepresentation is sues; however, when the research team
to search for support for your own identies an ethical dilemma, outside
views. To do so would be both bad sci- consultations are appropriate. Such
ence and unethical behavior. If data are consultations weigh the benets of the
research to society against the costs appropriate adjustments to the project

that bending ethical guidelines may en- (for good discussions on these matters
tail. Before any study begins, all such see Neuberger, 1992; and Downie and
ethical dilemmas should be resolved by Calman, 1987).
E X E R C I S E S
1. If you were designing a survey dealing the issue of protecting human rights.
with the sexual behavior of university What did the researcher do to protect
students, what would you see as the the rights of participants in the study?
major ethical problems in such a study? Did the author give sufcient informa-
What steps would you recommend be tion to assess this? If not, what informa-
taken before data collection begins? tion was missing to adequately assess
the safeguarding of participants rights?
If you were to replicate this study, what
suggestions would you make to en-
2. Examine either the CNA, ANA, or UKCC hance the protection of subject rights?
code of ethics. In view of the selected
code, what changes would you recom-
mend if you were proposing to per-
form a study of adults with syphilis? 4. Suppose you were planning a study of
substance abuse in a small native com-
munity. What ethical dilemmas might
you face in doing such a study? What
3. Select a nursing research study from a procedures would you recommend be
current research periodical. Consider followed?
Crookes, P., and Davies, S. (1997). Research ticle identies subtle ethical quandaries in
Into Practice. London: Balliere Tindall. This conducting research. The tension between
book contains an informative chapter on science and ethics emerges in the authors
critiquing ethical issues in published re- discussion of issues such as inclusion and
search studies. It is highly recommended exclusion criteria, efforts to enhance re-
for beginning researchers. sponse rates, risk disclosure, and the ben-
Fain, J.A. (1999). Reading, Understanding, and etrisk ratio.
Applying Nursing Research. Philadelphia: Parahoo, K. (1997). Nursing Research: Princi-
F.A. Davis. This interactive text succinctly ples, Process, and Issues. London: Macmillan
discusses ethical guidelines for research Press. This general research text integrates
involving human subjects. an interesting discussion of research ethics
Neuberger, J. (1992). Ethics and Health Care: as it relates to various research designs
The Role of the Research Ethics Committee throughout the text.
in the U.K. London: Kings Fund Institute. A Pranulis, M. (1997). Nurses roles in protecting
helpful discussion of the role of a research human subjects. Western Journal of Nursing
ethics committee. Research, 19(1), 130136. This article high-
Pallikkathayil, L., Crighton, F., and Aaronson, lights practical strategies for nurses in pro-
L. (1998). Balancing ethical quandaries tecting patients, patients families, and
with scientic rigor: Part I. Western Journal others who may be subjects in research
of Nursing Research, 20(3), 388393. This ar- projects.
Part 4
A STATISTICS
PRIMER
C
hapters 11 and 12 review the statistical concepts necessary
for beginning researchers. Although the elementary view pre-
sented here is helpful for beginning students, a more pro-
found understanding will be necessary for students attempt-
ing more advanced work (see Munro, 2001; Polit, 1996).
Those with a background in statistics can skip these chapters or

review them quickly; those without such a background should work
through the material carefully to acquire a foundation on which they
can build an understanding of advanced techniques. A reference to
Statistical Package for the Social Sciences (SPSS) commands is in-
cluded after each statistic is introduced so that readers are able to
nd the procedure in Appendix A quickly.
Chapter 11 reviews elementary descriptive statistics (procedures

for describing individual variables and for describing the relationship
between variables). A second group of statistics is known as inferen-
tial statistics and deals with how we come to conclusions in testing
hypotheses. An introduction to inferential statistics is presented in
Chapter 12. Both types of statistics are of relevance to nurse re-
searchers.
349

Chapter 11
A Primer for
Descriptive Statistics
CHAPTER OUTLINE
A. Levels of Measurement 5. The Normal Distribution

1. Nominal Measurements C. Describing Relationships Between
2. Ordinal Measurements Variables
3. Ratio Measurements 1. Cross-tabulation Tables
B. Describing an Individual Variable 2. Comparing Means
1. Data Distributions 3. Correlational Analysis
2. Measures of Central Tendency 4. Plotting the Data
3. Measures of Dispersion 5. Computing Spearman Correlations
4. Standardizing Data 6. Computing Partial Correlations
KEY TERMS
Analysis of variance Measures of central Proportionate reduction in

Bimodal distribution tendency error
Cross-tabulation table Measures of dispersion Range
Correlational analysis Median Rate
Correlation coefficient (r) Mode Ratio measurement
Curvilinear Nominal measurement Ratios
Data distribution Normal distribution Regression analysis
Descriptive statistics Ordinal measurement Regression line
Frequency distribution table Partial correlation Spearman correlation
Gamma Percentage Standard deviation
Inferential statistics Platykurtic t Test
Lambda Prediction Variance
Leptokurtic Proportion Z scores
Mean
351
352 A STATISTICS PRIMER
N
urse researchers ask questions measurementnominal, ordinal, and ra-
about many kinds of phenomena of tio. (The integration of the traditional
interest to the discipline; some of equal interval with the ratio level is dis-
them are about individuals (e.g., age, type cussed in Chapter 13.) By level of mea-
of lifestyle, income, job satisfaction, and surement, we simply refer to the different
attitudes toward health), some questions categories of measurement. Essentially,
are about communities (e.g., how much these categories represent a continuum
chronic illness is present, what type of varying from the arbitrary assignment of
comprehensive school health is available, values to categories (nominal measure-
if there has been a change in the propor- ment) to those where there is a one-to-one
tion of low-income people living in the relationship between the mathematical
community), and some questions may be meaning of the number and the phenome-
about nations (e.g., how many births and non being measured (ratio measurement).
deaths were recorded or have the in- An understanding of levels of measure-
comes of the people increased or gone ment is necessary because how one
down over the past 10 years?). To answer should go about analyzing a variable is
such questions, nurse researchers use a constrained by the measurement level
variety of descriptive statistics. These sta- achieved in data collection. One way to be-
tistics, including various tools, conven- gin to understand levels of measurement
tions, and procedures for describing vari- is to ask if the variable being measured has
ables, are discussed in this chapter. an underlying continuum (e.g., does it
Besides describing individual variables, vary from low to high?). If there is no un-
nurse researchers are also interested in derlying continuum and the variable is
examining relationships between variables made up of a number of discrete cate-
(e.g., is there a relationship between gories (e.g., men versus women), then
gender and healthy food choices? Does the measurement will be at the nominal
this relationship persist when men and level; if the variable has an underlying con-
women from different socioeconomic tinuum (e.g., number of years of formal ed-
backgrounds are compared?). Again, the ucation), then the level of measurement
researcher wishes to decide what rela- will be either ordinal or ratio.
tionship exists, and if it does, determine In order to facilitate computer analy-
whether it holds in virtually all circum- sis, researchers use numeric values to
stances. There are a variety of conven- identify different values in variables. For
tions used to describe relationships, and example, we might assign the number 1
these, too, are examined in this chapter. to represent men and the number 2 to
Keep in mind that statistics are tools to represent women. Or we could reverse
help in describing and understanding so- those numbers, having the 1s represent
cial relationships. Be a master of these women and the 2s represent the men. If
toolsbut treat them as tools because you think about it for a moment, it would
they are no substitute for theory and can- make no sense to calculate the average
not make up for poorly designed studies value on gender, yet it might well make a
or sloppy measurement. lot of sense to calculate the average
salary of tall people versus short people.
Why is that? The reason it makes sense to
A. LEVELS OF MEASUREMENT calculate an average annual salary is that
the numbers used to represent salary
Measurement is the process of assigning represent a one-to-one relationship to the
numbers to phenomena according to a set concept being measured. Here the num-
of rules. This text presents three levels of ber 20,000 represents $20,000. In the case
A PRIMER FOR DESCRIPTIVE STATISTICS 353
of gender, however, the numbers as- Childhood immunizations are an important

signed are arbitrary and have no mathe- means of primary prevention.
matical meaning, so it would not make Strongly Disagree 1 2 3 4 5 6 7 8 9 Strongly Agree
sense to calculate an average of these val-
ues. Researchers learn very quickly to un- This kind of item provides ordinal mea-
derstand levels of measurement because surement. Although we know that high
the measurement level determines how numbers indicate greater agreement with
the variable may be analyzed appropri- the statement, we do not know that the
ately. We distinguish between three lev- distances between the values are equal:
els of measurement in this text. the distance between 4 and 5 on the scale
may not be the same as the distance be-
tween 8 and 9. Ordinal measurement or-
1. Nominal Measurement ders values but does not assure equal
gaps between the measurement points.
Nominal measurement involves no under-
lying continuum, and the numerical values
assigned are arbitrary and have no arith- 3. Ratio Measurement
metical meaning. Religious afliation is a
nominal variable. Although there may be Finally, there is ratio measurement. With
an underlying continua related to religious ratio measurement, the zero point is
practices (e.g., degree of religious commit- aligned with true zero. For example, it is
ment or frequency of church attendance), possible to represent an individuals in-
by itself, religious afliation is a nominal come level exactly with a single number.
category: it may be Baptist, Buddhist, In this case, zero reects no income and
Hindu, Lutheran, Roman Catholic, or Jew- all other numerical values reect other
ish. When a respondent checks off which income levels. In addition, with ratio level
(if any) religious afliations he or she was measurement, it is possible to add and to
associated with while growing up, the mea- subtract constants as well as to multiply
surement level attained is nominal. One or to divide by them, without changing
category is neither higher nor lower than the proportionality among the values.
any otherthey are simply different cate- For example, it is correct to say that an in-
gories. Other examples of nominal vari- come of $50,000 is twice as much as an in-
ables include gender, program of study, come of $25,000.
political party afliation, and ethnic origin. As another example, if a researcher
were to have respondents indicate with a
check mark which of the magazines on a
2. Ordinal Measurement list they had scanned or read in the pre-
vious month and a total were taken of the
Ordinal measurement involves an under- number ticked off, the result would be a
lying continuum in which the numerical variable that varies from 0 to the number
values are ordered so that small numbers required if a respondent checked off all
refer to lower levels on the continuum the items listed. This measure would be a
and larger numbers to higher points; how- ratio level measure. Similarly, when com-
ever, the distances between the assigned munities are being studied, measures
numbers and the underlying continuum such as the proportion of visible minori-
are not in a one-to-one relationship with ties, the proportion of retired people in
each other. For example, suppose a state- the population, and various rates and ra-
ment like the following is presented to re- tios (e.g., suicide rate, dependency ratio,
spondents: gender ratio) are all ratio level measures.
B. DESCRIBING AN INDIVIDUAL in two tests. Examine the grades. If you

VARIABLE were asked to indicate the test on which
the nursing students did best, what would
In this section, we consider how individ- you say? The task is a frustrating one be-
ual variables may be described. We look cause the mass of detail is overwhelming.
at data distributions, measures of central The need for some method of summariz-
tendency, how variation within a variable ing the grades quickly becomes apparent.
is measured, approaches to standardiz- It is inefcient to read an unordered list-
ing variables, and the characteristics of a ing of numbers.
normal distribution. The challenge for the researcher is to
nd ways to characterize a distribution in
1. Data Distributions a simple way. One of the most basic tech-
niques for simplifying a large data set is to
Quantitative researchers need to simplify create a frequency distribution table.
the complexity of their data to communi- This is a systematic listing of all the values
cate their results effectively. Many tools on a variable from the lowest to the high-
are available to researchers to help in est with the number of times (frequency)
this endeavor. Data may be displayed in each value was observed. Suppose the re-
graphs, summarized by showing how searcher has recorded the weight, size of
many cases fall in each category, or de- home community, and gender of subjects.
scribed using any number of descriptive The researcher may show a frequency dis-
statistics. tribution in each category of gender (ap-
A data distribution simply refers to a propriate for nominal and ordinal data),
listing of all the values for any one vari- use a bar graph to show how many re-
able. However, to simply provide a list of spondents came from communities of dif-
the data would not be terribly helpful to ferent sizes (also appropriate for nominal
the reader of a report. Usually there are and ordinal data), or use plots of the data
simply too many numbers and it is dif- to describe the distribution of egalitarian-
cult to make much sense out of a listing of ism scores among the respondents (ap-
raw data. For example, Table 11.1 shows propriate for ratio variables). Table 11.2
the scores that nursing students received shows a frequency distribution and a bar
Table 11.1 Two Sets of Nursing Grades for a First-Year Class

Test A. Results
60 60 82 71 60 58 64 81 58 58 70 57 56
56 69 58 55 82 46 54 62 61 77 70 59 74
87 47 57 63 37 67 55 59 63 59 55 52 58
63 72 54 54 62 69 66 58 53 73 57 68 52
75 47 52 73 72 65 64 63 59 57
Test B. Results
53 64 83 60 61 61 61 83 54 58 68 55 49
60 59 55 53 69 44 56 48 54 74 71 49 54
86 51 67 63 59 63 55 40 65 63 62 55 49
53 72 59 59 54 69 73 57 59 72 26 65 70
60 45 60 69 66 63 51 59 63 63
Table 11.2 Frequency Distribution and Bar Chart for Size of Home Community
Size of Home Community Frequency Percent Valid Percent Cumulative Percent
Under 1,000 18 9.7 10.3 10.3
1,000 to 4,999 20 10.8 11.5 21.8
5,000 to 9,999 17 9.1 9.8 31.6
10,000 to 19,000 15 8.1 8.6 40.2
20,000 to 49,999 29 15.6 16.7 56.9
50,000 to 99,999 21 11.3 12.1 69.0
100,000 to 499,999 10 5.4 5.7 74.7
500,000 to 99,999 7 3.8 4.0 78.7
1,000,000 and over 37 19.9 21.3 100.00
Total 174 93.5 100.00
Missing 12 6.5
Total 186 100.0
graph for the size of home communities this plot. Note how the plot line indicates
for respondents in a study. few low scores, many more in the middle,
The egalitarianism scores (Fig. 11.1) with a tapering off at the high score end.
could also be plotted, but this time we will This distribution is described as a normal
also include a distribution curve line on distribution. Normal distributions are de-
EGALITARIANISM
50
40
30
20
10
Std. Dev = 8.74
Mean = 30.8
0 N = 182.00
10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0
Figure 11.1 Frequency distribution with a normal curve, egalitarianism scores.
scribed in detail later in this chapter; for ability, the distribution curve tends to be
now, simply note how the distribution at and wide and is called a platykurtic
roughly forms a bell-shaped curve. distribution. An example of these two
But there are other distribution curves. plots is shown in Figure 11.2. The recog-
Not all variables are normally distributed. nition that distributions can be quite dif-
If, for example, we were to plot the weights ferent from one another will be important
of freshman students, we would almost when the time comes to learn about sam-
certainly nd that the result would be a pling (Chapter 15) and tests of signi-
bimodal distribution. The reason for this cance (Chapter 12).
is that female students have lower average Besides plotting the data distribution
weights than the males. Essentially we or showing the frequency distribution,
would end up combining two normally dis- there are other standard ways in which
tributed plots, one for women and one for data distributions are described. A vari-
men, and one that would have two peaks ety of conventions are used to describe
and considerable overlap between the the central tendency, the variation in or
male and female weights (Fig. 11.2.) dispersion of a variable, as well as com-
If a set of values has little variability, monly used methods used to standardize
then the distribution is peaked and it is variables to facilitate comparisons. We
said to be leptokurtic; on the other hand, rst explore the main tools used to de-
if the distribution has a great deal of vari- scribe a variables central tendency.
BI-MODAL
LEPTOKURTIC PLATYKURTIC
Figure 11.2 Other distributions.
2. Measures of Central Tendency Canada was the most frequently occurring

response to the question asking respon-
A measure of central tendency is a single dents to indicate their country. Table 11.3
numeric value that summarizes the data presents a frequency distribution (simply
set in terms of its average value. Just as reports the number of respondents who
baseball fans cite the batting average to fall into each category) of respondents
summarize a players hitting ability, nurse who report themselves to be from each of
researchers may use the value of 98.6F the ve countries listed. From Table 11.3,
(or 37C) to describe the average adult one can identify the modal category by
body temperature. Three measures of looking at the Number column, and pick-
central tendency are commonly used: the
mode, the median, and the mean.
Table 11.3 Distribution of Respondents by
Country
a. The Mode Country Number Percent

Australia 63 15.5
The mode is the category or value of a
Canada 165* 40.5
nominal variable with the most cases in a
distribution of values. If one wishes, for ex- New Zealand 58 14.3
ample, to describe average respondents United Kingdom 43 10.6
in terms of their country of origin, then the United States 78 19.2
mode would be the appropriate statistic.
Total 407 100.1
In the case of country of origin, it may be
Canada. This would simply mean that *Mode.
ing out the category that has the highest Table 11.4 Median for Nurse Graduate
Students: An Extreme Value Problem
frequency; in this case, the category is
Canada. Case Number Income, dollars
For nominal or ordinal variables, it is 1 5,400
appropriate to determine the mode and 2 6,600
the median values as well as the fre-
3 7,700
quency distribution of the variable.
4 10,200
5 13,400
b. The Median 6 16,400
7 16,700
The median represents the midpoint of a
distribution. One half the cases fall above 8 18,300*
the median value and one half below it. 9 19,000
The median is normally used for ordinal 10 20,000
level variables or in cases where the use
11 20,500
of a mean would be problematic because
a few extreme values would give an inap- 12 22,900
propriate impression of the typical case. 13 24,600
The steps in determining the median are: 14 31,500**
Arrange the cases in order from highest 15 580,000
to lowest or lowest to highest values.
Number the values ( ignoring no re- *Median value is $18,300
**Mean value is $54,213
sponse or missing data).
If there is an odd number of cases, then
the middle value is identied and that
value is the median for the distribution. c. The Mean
If, however, there is an even number of
The mean is a measure of central ten-
cases, then the mean of the middle two
dency that typies a set of observations
is calculated and that value represents
with a single value. Recall the grades in
the median.
the two rst-year tests shown in Table
Table 11.4 illustrates a situation in 11.1. Earlier you were asked to indicate
which a median would be a better de- the test on which the nursing students
scription of a sample than would a mean. did best. The task is a frustrating one be-
Here the data list the annual incomes of 15 cause the mass of detail is overwhelming.
graduate students enrolled in nurse mid- The need for some method of summariz-
wifery programs in California. When we ing the grades quickly becomes appar-
compare the extremely high salary of ent. It is inefcient to read an unordered
$580,000 with the much more modest in- listing of numbers.
comes of the other nursing students sur- Computing means may be a rst step
veyed, we realize that citing the mean in comparing the grades in the two sec-
would give the false impression that the in- tions. The mean (more formally known as
comes of graduate students are quite high. the arithmetic mean and less formally as
The median value of $18,300 is a much bet- the average) is computed by summing
ter description of the income distribution the values of a variable and dividing the
than the mean of $54,213. When atypical result by the total number of cases. A
values are present, the median may pro- mean is used on ratio level data. To nd
vide a better description of the data. the mean grade in each test, we would
simply sum the grades and then divide by stable when different samples are drawn
the number of students. This would re- from the same population. Table 11.6 in-
sult in a mean grade for each section of dicates that the range of Beths grades is
the course. Table 11.5 reports the means, 28, but Marys is 10, indicating much less
showing that the average performance variability in Marys grades compared
was somewhat higher in test A. Students with Beths.
scored 2.0 points higher on the rst test
than on the second one. That difference b. Standard Deviation
is not easily apparent in looking at the
raw scores. Researchers rely heavily on the standard
deviation to give them a sense of how
much dispersion there is in a distribution
3. Measures of Dispersion
of scores. The standard deviation is used
Besides describing a variable in terms of with ratio level data, considers all the val-
central tendency, it is helpful to know ues in the data set, and reects the average
something about the extent of the vari- amount that each value in the distribution
ability, or dispersion, of the values. Are varies from the mean. The higher the stan-
most of the values close to one another, or dard deviation, the more variability in the
are they spread out? To illustrate, sup- distribution; low standard deviations re-
pose we have two students, Mary and ect greater homogeneity in the data set.
Beth. Their grades are indicated in Table The standard deviation formula is:
11.6. Although both students have an iden- _

(X X ) 2
sd
tical 82 percent average, the distributions N1
are quite different. Marys grades vary lit-
tle, but Beths vary considerably. We will
Table 11.7 shows how the standard de-
explore three measures of dispersion in
viation for Beths grades could be com-
the two students grades: range, standard
puted. An examination of the table reveals
deviation, and variance.
that the standard deviation of Marys
grades is 3.7, considerably less than
a. Range Beths 12.2. An examination of the actual
grades indicates that, indeed, there is
The range is a measure of the gap be- much more variability in Beths than in
tween the lowest and highest value in a Marys grades. The standard deviation re-
distribution. It is computed by subtract- ects the variability in a set of values.
ing the lowest value from the highest one. Beginning researchers should be famil-
Although the range is a simple one to cal- iar with how standard deviations are com-
culate by using only the extreme scores, puted. This statistic is an important one. It
it fails to reect variations between the plays a role in computations such as de-
other scores. It tends also to be rather un- termining sample size and in doing various
Table 11.5 Summary Statistics for Nursing Grades

Summary Statistics Test A Grades Test B Grades
Mean 62.02 60.02
Median 58.00 59.00
Mode 60.00 60.00
Cases, n 62 62
Table 11.6 Two Grade Distributions In the illustration using Marys and
Subject Mary Beth Beths grades, the variances are 14.0 and
Sociology 78 66 150.0, respectively (see Table 11.6).
Table 11.8 summarizes the single vari-
Psychology 80 72
able statistics that have been presented
Nursing 82 88 in this text. Note that it is possible to use
Anthropology 82 90 all of the summarizing statistics when a
Philosophy 88 94 ratio level of measurement has been
achieved. Indeed, when there are unusual
Summary Statistics
features in the data, such as a few ex-
Mean* 82 82 treme values or a lot of identical scores,
Range 10 28 it sometimes makes sense to use a sum-
Standard Deviation 3.74 12.25 mary statistic other than the one high-
lighted in the table.
Variance 14.0 150.0
*Mean Sum of values divided by number of cases. 4. Standardizing Data

Range Highest value minus lowest value.
Standard deviation (see computation in Table 11.7).
Variance sd2. To standardize data is to report data in a
way that comparisons between units of
different size may be made; data may also
be standardized to create variables that
tests of statistical signicance. In addition,
have similar variability in them (these are
the standard deviation is an element in
called Z scores; they are introduced later
most advanced statistical techniques. Af-
in this chapter).
ter data have been entered into the com-
puter, it will not be necessary to hand com-
pute standard deviations; nonetheless, it a. Proportions
is crucial to understand what they mea-
A proportion may be calculated to show,
sure.
for example, how many females there are
in a population compared to the total
c. Variance population. Suppose we wished to com-
pute the proportion of females in a com-
The third measure of dispersion is vari- munity with a total of 58,520 persons:
ance. Similar to other measures of dis- Number females
persion, it provides a sense of the vari- Proportion female
Total persons
ability in the data. Similar to the standard
31,216
deviation, variance reects the average Proportion female
58,520
amount of deviation from the average
value in the distribution. Indeed, compu- Proportion female 0.533
tationally it is simply the standard devia-
tion squared. The term variance will be The females represent 0.53 of the pop-
encountered in later chapters dealing ulation. In other words, there are 0.53
with explained variance and in analysis of women and girls for every one person in
variance, a statistical test discussed in the community.
Chapter 12. The computation for vari-
ance is: b. Percentages
_
(X X )2 If we wished to represent a proportion as a
Variance sd 2
N1 percentage, we would simply multiply the
Table 11.7 Computation of Standard Deviation, Beths Grades

Step 1 Step _2 Step 3_
Subject Grade XX (X X)2
Sociology 66 66 82 16 256

Psychology 72 72 82 10 100
Nursing 88 88 82 6 36
Anthropology 90 90 82 8 64
Philosophy 94 94 82 12 144
Mean 82.0 Total 600
Step 1: Compute the mean for the scores by summing the scores and dividing by the number of observations.
Step 2: Calculate the deviation from the mean for each observation.
Step 3: Square each deviation score and sum them.
Step 4: Using the formula below, calculate the standard deviation.
_
(X X )2
sd
N1
sd
600

4
sd 12.25
_
Some notational conventions that are used in this text are: is a summation symbol; X with a bar over it indicates that
it is the mean value; N number of cases; sd standard deviation.
proportion by 100. In the example just We report percentages to adjust for the
given, we note that females constitute 53.3 fact that there are different numbers of stu-
percent of the population. For a more com- dents involved in each of the categories.
plex example, Table 11.9 presents the rela- By computing the percentages, we are able
tionship between size of home community to say that for every 100 rural students,
and whether the respondent plans to at- 52.3 are planning to attend a university
tend university. It is not adequate simply compared with 48.9 of every 100 students
to say that 69 rural students, 44 students from the small town category and 73.9 of
from small towns, and 102 students from every 100 high school students from towns
larger towns plan on attending university. with populations of more than 5000.
Table 11.8 Summary Statistics for Single Variables*

Level of Measures of Measures of
Measurement Central Tendency Dispersion SPSS Procedure
Nominal Mode FREQUENCIES
Ordinal Mode Range FREQUENCIES
Median
Ratio Mode Range DESCRIPTIVES
Median Standard deviation
Mean Variance
*Bold indicates the statistic used under normal circumstances.

Table 11.9 Plans to Attend University by Size of Home Community*

Town up to Town over 5000
Rural 5000 Population Population Total
University Plans? N % N % N % N %
Plans 69 52.3 44 48.9 102 73.9 215 59.7
No plans 63 47.7 46 51.1 36 26.1 145 40.3
Total 132 100.0 90 100.0 138 100.0 360 100.0
*If appropriate, test of signicance values are entered.
c. Percentage Change Divide the above total by the number

at time 1.
Often nurse researchers compare num-
bers at one time to those at another time.
Multiply the above total by 100.
For example, they may want to measure
the percentage increase in the number of d. Rates
teen pregnancies from one period to the
next. Table 11.10 illustrates a problem in The incidence of a social phenomenon is
which nurse scientists may wish to calcu- often presented in the form of a rate. A
late the percentage change between 1990 rate indicates the frequency of some
and 2000 in the number of teen pregnan- phenomenon for a standard-sized unit
cies. The general form of the equation for (such as incidence per 1000 or per
calculating percentage change would be: 100,000). This allows us to easily com-
pare the incidence of a phenomenon in
Percentage Change units of different size. To know, for ex-
Time 2 numberTime 1 number 100
ample, that there were 27 suicides in a
Time 1 number
city of 250,000 (Middle City) in one year
and 13 suicides in another city of 110,000
To calculate the percentage change
(Small City) does not allow quick com-
use the following steps:
parison unless we compute a suicide
Using the general equation, subtract the rate. A suicide rate may be computed in
time 1 number from the time 2 number. the following manner:
Table 11.10 Calculating Percentage Change: Percentage Change in Number of Teen Pregnancies
between 1990 and 2000, Canada and United States
Number per 1000 Teen Number per 1000 Teen Percent Change,
Country Women, 1990 (Time 1) Women, 2000 (Time 2) 1990 to 2000
Canada 41.6 39.5 5.05
United States 85.0 82.1 3.41
Note the method of calculating percent change:

Time 2 number Time 1 number
Percent change 100
Time 1 number
In the case of the United States in the above table, the calculation would be:
82.1 85.0
Percent change 100
85.0
Percent change 3.41
Suicide rate e. Ratios

Number suicides per year
Ratios are used to compare rates or other
Mid-year population
measures across categories. For exam-
100,000 ple, suppose one wished to compare the
suicide rate of black male teens and white
When calculated, we nd that the sui- male teens in the United States. Whereas
cide rate for Middle City is 10.8; for Small the suicide rate in the United States is
City, the rate is 11.10. This means that 12.2 in black males 15 to 19 years old (per
Middle City has 10.8 suicides in the year 100,000), the comparable rate in young
for every 100,000 people in the city; in white males is 19.1. The American white-
Small City the rate is 11.8 per 100,000. In to-black male suicide ratio for 15 to 19
this case, we see that Small City has a years old could be represented as:
slightly higher suicide rate. Rates can U.S. white/black male suicide ratio
also be computed for specic age cate-
gories or on other bases; the only adjust- White male suicide rate

ment required is that we use the number Black male suicide rate
of suicides in the category compared 19.1
U.S. white/black male suicide ratio
with its total size. Rates are computed for 12.2
many things, including births, marriages, U.S. white/black male suicide ratio 1.56
divorces, deaths, and crime.
Table 11.11 presents suicide rates for This ratio suggests that the white male
ve ctitious cities. Note that rates permit youth suicide rate is 1.56 times higher
easy comparisons between units of un- than the comparable black male youth
equal size. The data are standardized, per- rate. Many ratios can be computed that,
mitting comparison between the units. If like rates, facilitate comparison between
the absolute number of suicides were pre- categories.
sented, this would create the impression Table 11.12 presents the ratio between
that suicide is a much more serious prob- male and female suicide rates for black and
lem in the larger jurisdictions. This may white ethnic groups in the United States.
not be the case. The rates adjust for the Note that, nationally, men commit suicide
differences in size. 5.79 times more frequently than do women.
Table 11.11 Suicide Rates for Five Cities

Number of Suicides Population Suicide Rate*
City A 49 581,257 8.43
City B 16 134,567 11.89
City C 97 933,590 10.39
City D 1263 7287,940 17.33
City E 1083 10,939,394 9.90
*In the case of City A, the computation would be as follows:

Number of suicides in the year
Suicide rate 100,000-
Mid-year population
49
Suicide rate 100,000
581,252
Suicide rate 8.43

Table 11.12 Male and Female Suicide Ratios, by Group, Youth 15 to 19 Years of Age, United States*
Group Suicide Rate, Males Suicide Rate, Females Ratio, Male/Female Suicides
Black 12.2 1.9 6.42
White 19.1 4.2 4.54
United States 22 3.8 5.79
*In this table, the male/female suicide ratio is computed for each group by dividing the male suicide rate by the female
rate. The result indicates how many male suicides there are for every female suicide.
SOURCE: Statistical Abstract of the United States, 1994. Washington, D.C.: U.S. Government Printing Ofce, p. 101.
In some cases, ratios are reported so It forms a symmetrical, bell-shaped

that they are standardized to a base of 100. curve.
For example, if we had a community with It has the same mean, mode, and me-
27,304 males and 31,216 females, we might dian values, with half the cases falling
compute a sex ratio. The ratio for the com- below the mean and the other half
munity could be calculated as follows: above the mean.
Number of males
It becomes smoother as the number of
Gender ratio 100 observations and the number of mea-
Number of females
surement units become ner.
27,304
Gender ratio 100 If a plot is made showing the distribu-
31,216
tion, for example, of the weight of male
Gender ratio 87.5 university students, it would approxi-
mate a bell-shaped curve. There will be
This gender ratio would indicate that few cases on the extremesthe very light
there are 87.5 males in the community for and the very heavy; most of the cases will
every 100 females. Such a ratio allows us be found clustered toward the middle of
to quickly compare the sex ratios of com- the distribution.
munities, nations, age groups, or any Another example of a normal distribu-
other category. tion is a plot of the outcomes of a series of
10 coin ips. Suppose we ip a coin 10
5. The Normal Distribution times, record the number of heads, repeat
this operation 1024 times, and then plot
The normal distribution is another key
concept used by researchers. Many of
the observations we make on individual
or group characteristics approximate
what is referred to as a normal distribu-
tion. What does this mean? Many phe-
nomena in the social and physical world
are normally distributed. A normal distri-
bution is a theoretical representation of
the frequency of scores on a variable.
When the scores are plotted, values clus-
ter around the midpoint with a falling off 2 sd 1 sd X 1 sd 2 sd
toward the extremes, resulting in a bell- 68.28
shaped curve (see Fig. 11.3). A variable 95.46
that is normally distributed has the fol-
Figure 11.3 Normal curve distribution.
lowing characteristics when plotted:
the number of times we got 0, 1, 2, . . . 10 68.28 percent of the observations will

heads in the trials. The outcome will ap- be divided equally between the mean
proximate that shown in Figure 11.4, and one standard deviation to the
which is a graph of the theoretical proba- right of the mean (34.14) and one stan-
bilities of getting each of the 11 possible dard deviation to the left of the mean
outcomes (i.e., 0 through 10 heads). The (34.14).
result approximates that of a normal dis- 95.46 percent of the observations will
tribution. fall two standard deviations from
An interesting characteristic of the the mean.
normal distribution is its connection to 95 percent of the cases fall 1.96 stan-
the standard deviation. By denition, a dard deviation units from the mean.
xed proportion of cases in a normal dis- 99 percent of the cases fall 2.58 stan-
tribution will fall within a given number of dard deviation units from the mean.
standard deviation units of the mean (see
Fig. 11.3). About two thirds of the cases
will fall within one standard deviation of a. Z Scores
the mean; just over 95 percent of the
cases will fall within two standard devia- Z scores measure the distance, in stan-
tions of the mean. More precisely: dard deviation units, of any value in a dis-
300
200
100
Frequency
Std. Dev = 1.58

Mean = 5
0
0 2 4 6 8 10
No. Heads in 10 Flips, N = 1,024

Figure 11.4 Distribution of number of heads flipped in 10 tries.
tribution from the mean. Thus, if some- Such standardization can be accom-
ones income has a Z score of 1.43, it plished easily in SPSS.
would indicate that the income is 1.43 To perform this analysis in SPSS, use the
standard deviation units above the mean DESCRIPTIVES procedure; see Appendix A.
of the distribution. Suppose that the mean Combining indicators to create an in-
income is $65,000 and the standard devia- dex is a major use of Z scores. Suppose,
tion $22,000; the Z score 1.43 would infor example, that we have measures on
dicate an income of $96,460. How is this income and years of education and that
value computed? The formula for Z scores we wish to combine them to form a so-
is as follows: cioeconomic index. It would not make
_ sense simply to add a respondents years
XX of education to his or her annual income;
Z sd
the reason is that incomes may vary from
Where: $5,000 to $500,000 but years of education
X is the observation; may vary from 0 to 20. By adding them to-
_
X is the mean of the distribution; gether, the income component would to-
tally dominate the index. Someone earn-
sd is the standard deviation of the distribution.
ing $50,000 with 8 years of education
would have a score of 50,008, but a per-
By plugging the values into the equa-
son with a BA and $40,000 income would
tion and solving it, the value $96,460 is
end up with a score of 40,016. Somehow
obtained, as in:
we need to weight the components so
X 65,000 that income and education will equally in-
1.43 uence the outcome. Using Z scores is an
22,000
easy way to do this.
X (1.43 22,000) 65,000
Table 11.13 shows the computation of
X 96,460 socioeconomic index scores using Z
scores. Notice how either a lower-than-
One of the consequences of being able average income or a lower-than-average
to report a value in terms of its Z score is education leads to a reduction in the to-
that we now have a powerful comparative tal socioeconomic score. These indexes
tool. Suppose we wanted to compare indi- can be calculated quickly by a computer
viduals relative income positions in two and the components weighted in any way
countries; we could simply report the in- the researcher likes. The Z scores, for ex-
comes in Z score terms, and this would tell ample, may be added together, resulting
us where each individual stands in terms in a value that can be taken to represent
of his or her countrys income distribu- the relative socioeconomic position of
tion. This would permit us to compare a the various respondents, with income
British familys income of 24,000 to an and education making equal contribu-
American familys income of $90,000. tions to the nal score. Such computa-
Students should recognize that com- tions can be done rapidly within SPSS.
puting Z scores can be used to standard-
ize variables, and the resulting distribu-
tions have means of 0 and standard b. Areas Under the Normal
deviations of 1. Thus, instead of just hav- Distribution
ing variables with income scores, educa-
tional levels, and occupational prestige, Another useful property of the normal
we might have standardized variables distribution is that it is possible, with the
containing the Z scores for each variable. help of Table 11.14, to calculate the pro-
Table 11.13 Computing an Index Score Using Z Scores

Income Years of Education
Given population values
Mean 65,000 11
Standard deviation 22,000 4
Suppose ve individuals
A 55,000 7
B 41,000 12
C 30,000 8
D 64,000 16
E 86,000 9
Compute an index equally weighting income and years of education. The general
equation is:
_
XX
Z
sd
Case A. Income: (55,00065,000) 22,000 0.45
Education: (7 11) 4 1.00
Socioeconomic index score 1.45
Case B. Income: (41,00065,000) 22,000 1.09

Education: (12 11) 4 0.25
Case C. Income: (30,00065,000) 22,000 1.59

Education: (8 11) 4 0.75
Case D. Income (64,00065,000) 22,000 0.05

Education: (16 11) 4 1.25
Case E. Income: (86,00065,000) 22,000 0.95

Education: (9 11) 4 0.050
portion of cases that will fall between two point, shade the curve to the right of
values, or above or below a given value. the $100,000 mark.
To illustrate, suppose we wished to Step 2. Calculate the Z score to determine
know what percentage of incomes would how many standard deviation units
fall above $100,000, given a population $100,000 is above the mean, as in:
standard deviation of $22,000, and a _
mean of $65,000. The steps followed to X X
Z
solve this problem would be as follows: sd
100,000 65,000
Step 1. Draw a normal curve, marking be- Z
22,000
low it the mean and standard deviation
values, and drawing a line through the Z 1.59
curve at the point where you expect
$100,000 to fall. Because the question Step 3. Look up the value 1.59 in Table
asks about the percentage above this 11.14. Move down the Z score column
Table 11.14 Areas Under the Normal Curve

Fractional parts of the total area (10,000) under the normal curve, correspnding to distances between
the mean and ordinates which are Z standard-deviation units from the mean.
Z .00 .01 .02 .03 .04 .05 .06 .07 .08 .09
0.0 0000 0040 0080 0120 0159 0199 0239 0279 0319 0359
0.1 0398 0438 0478 0517 0557 0596 0636 0675 0714 0753
0.2 0793 0832 0871 0910 0948 0987 1026 1064 1103 1141
0.3 1179 1217 1255 1293 1331 1368 1406 1443 1480 1517
0.4 1554 1591 1628 1664 1700 1736 1772 1808 1844 1879
0.5 1915 1950 1985 2019 2054 2088 2123 2157 2190 2224
0.6 2257 2291 2324 2357 2389 2422 2454 2486 2518 2549
0.7 2580 2612 2642 2673 2704 2734 2764 2794 2823 2852
0.8 2881 2910 2939 2967 2995 3023 3051 3078 3106 3133
0.9 3159 3186 3212 3238 3264 3289 3315 3340 3365 3389
1.0 3413 3438 3461 3485 3508 3531 3554 3577 3599 3621
1.1 3643 3665 3686 3718 3729 3749 3770 3790 3810 3830
1.2 3849 3869 3888 3907 3925 3944 3962 3980 3997 4015
1.3 4032 4049 4066 4083 4099 4115 4131 4147 4162 4177
1.4 4192 4207 4222 4236 4251 4265 4279 4292 4306 4319
1.5 4332 4345 4357 4370 4382 4394 4406 4418 4430 4441
1.6 4452 4463 4474 4483 4495 4505 4515 4525 4535 4545
1.7 4554 4564 4573 4582 4591 4599 4608 4616 4625 4635
1.8 4641 4649 4656 4664 4671 4678 4686 4693 4699 4706
1.9 4713 4719 4726 4732 4738 4744 4750 4758 4762 4767
2.0 4773 4778 4783 4788 4793 4798 4803 4808 4812 4817
2.1 4821 4826 4830 4834 4838 4842 4846 4850 4854 4857
2.2 4861 4865 4868 4871 4875 4878 4881 4884 4887 4890
2.3 4893 4896 4898 4901 4904 4906 4909 4911 4913 4916
2.4 4918 4920 4922 4925 4927 4929 4931 4932 4934 4936
2.5 4938 4940 4941 4943 4945 4946 4948 4949 4951 4952
2.6 4953 4955 4956 4957 4959 4960 4961 4962 4963 4964
2.7 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974
2.8 4974 4975 4976 4977 4977 4978 4979 4980 4980 4981
2.9 4981 4982 4983 4984 4984 4984 4985 4985 4986 4986
3.0 4986.5 4987 4987 4988 4988 4988 4989 4989 4989 4990
3.1 4990.5 4991 4991 4991 4992 4992 4992 4992 4993 4993
3.2 4993.129
3.3 4995.166
3.4 4996.631
(continued on next page)

Table 11.14 Areas Under the Normal Curve (Continued )
Z .00 .01 .02 .03 .04 .05 .06 .07 .08 .09
3.5 4997.674
3.6 4998.409
3.7 4998.922
3.8 4999.277
3.9 4999.519
4.0 4999.683
4.5 4999.966
5.0 4999.997133
SOURCE: Rugg, H.O. (1945). Statistical Methods Applied to Education. Boston: Houghton Mifin. (Original work published
1917). Cited with permission.
until you come to the value 1.5, then should follow procedures similar to
read across to the column headed by those used in the case above. This time,
0.09, (this gives you the value for the 9 however, the diagram will show a shaded
in the value 1.59) and read the value. area between two points on either side of
You should have found the number the mean. This time two Z scores will
4441. This number should be under- need to be computed, the values looked
stood as 0.4441, a proportion. up in Table 11.14, and the proportions be-
Step 4. By denition, we know that one tween the mean and each cutpoint will
half of the cases will fall above the need to be determined, then added to-
mean. Expressed as a proportion, this gether to get the nal answer. The com-
would indicate that 0.5 of the cases will putations may be done as follows:
fall above the mean. The question we
Step 1. Proportion between the mean and
are trying to answer is what propor-
$70,000:
tion of the cases fall above $100,000. _
Looking at the diagram we made in XX
Z
Step 1, we realize that if the right side sd
of the curve contains 0.5 of all the 70,000 65,000
cases, and, if the value $100,000 is Z
22,000
0.4441 above the mean, then the cases
above $100,000 would have to be: Z 0.23
Proportion of normal curve included in Z score of

0.5000 0.4441 0.0559 0.23 0.0910.
Step 5. As a proportion, 0.0559 of the cases

Step 2. Proportion between $40,000 and
will fall above $100,000. Or, another
the mean:
way of expressing the same idea, is to _
X X
say that 5.6 percent of the cases will fall Z
sd
above $100,000 (multiply the propor-
tion 0.0559 by 100 to get 5.6 percent). 40,000 65,000
Z
22,000
Suppose we wish to determine the pro-
portion of cases that will fall between Z 1.14
$40,000 and $70,000, given the same pop- Proportion of normal curve included in Z score of
ulation mean and standard deviation. We 1.14 0.3729.
Step 3. Adding the proportions together: ment), median (ordinal measurement),

or mode (nominal measurement).
0.3729 0.0910 0.4639 A frequency distribution is used to re-
ect dispersion with nominal vari-
The computation indicates that just ables; the range is used as a measure
fewer than one half of all the cases, 46.4 of dispersion with ordinal variables;
percent, fall between the incomes of the standard deviation and variance
$40,000 and $70,000. The proportion be- are used to reect dispersion of ratio
tween the mean and the respective Z level variables.
scores is shown in Figure 11.5. In this To perform this analysis in SPSS, use the
case, the values are added together to de- DESCRIPTIVES for ratio variables and the
termine the proportion of cases that fall FREQUENCIES procedures for nominal and
between $40,000 and $70,000. ordinal variables; see Appendix A for de-
There are other types of normal curve tails.
problems that can be solved. Just keep
the previous examples in mind, draw a di-
agram shading in the area you need to de- C. DESCRIBING
termine, and remember that each side of RELATIONSHIPS BETWEEN
the normal curve contains one half, or VARIABLES
0.5, of the cases. With these things in
mind, it should be possible to solve most Researchers are concerned with describ-
normal curve problems. ing the relationships among variables. In
As a review of this rst section and be- dealing with two-variable relationships,
fore we go on to consider how we analyze one of the variables will be treated as a de-
relationships between variables, recall pendent variable and the other will be
that before we begin to analyze any vari- treated as an independent variable. (There
able, we need to appreciate its level of mea- may not necessarily be evidence to sup-
surement so that appropriate analyses port designating one of them as dependent
may be made. The level of measurement and the other independent, but to do the
attained determines the best ways of sum- analysis we are required to identify one as
marizing a variable. Note the following: dependent.) If a control variable is used, it
To provide a measure of central ten- will be an intervening source of spurious-
dency use the mean (ratio measure- ness, or conditional variable.
This section describes some of the ma-
jor procedures used to examine relation-
ships between variables. Three basic
.3729 steps must be taken to begin analyzing
any relationship:
.0910 The rst step is to decide which vari-
able is to be treated as the dependent
variable and which one as the inde-
pendent variable.
The second step is to determine the
level of measurement for each variable.
1.14 X .23 The third step is to decide on the ap-
.5 .5 propriate procedure for examining the
relationship.
Figure 11.5 Areas under the normal curve.
Finally, the analysis must be performed.
Because the researcher typically is 2. Cross-tabulation Tables

trying to understand what causes varia-
tions in a dependent variable, common Cross-tabulation tables present informa-
sense alone can generally determine tion so that the relationship between a
which variable should be designated as nominal level dependent variable can be
the dependent variable. However, in related to an independent variable. Re-
other cases, it is not as obvious. For ex- call that Table 11.9 presented ndings on
ample, suppose you were to examine the the relationship between the educational
relationship between healthy lifestyle plans and the size of the home commu-
choices and perceived health status. In nity for 360 high school students. Lets
this case, it would not be obvious if pos- examine this table again, as it typies the
itive perception of health status leads to cross-tabulation table.
healthy lifestyle choices or if the direc- A cross-tabulation table classies cases
tion of the relationship is reversed. on two or more variables. In this example,
Hence, it is not clear which variable the data are rst sorted into categories
should be designated as the dependent representing community size; next, each
variable. In such cases, try to decide of these categories is sorted into whether
which variable occurs last in a temporal the person does or does not plan to attend
sequence. It is entirely possible that the college. This sorting allows us to see if
two variables mutually inuence one an- those from rural areas are more likely to
other. If this is the case, one will nonethe- have college plans than their urban coun-
less have to be designated as the depen- terparts.
dent variable.
Having decided which variable is to be
treated as the dependent variable, one a. Rules for Constructing Cross-
must next identify the level of measure- tabulation Tables
ment of each of the variables. Now, using
Let us now examine some of the rules for
the information provided in Table 11.15,
constructing and for interpreting cross-
identify which procedure would be most
tabulation tables
appropriate for the analysis. Let us ex-
plore each of the procedures identied in Rule 1. In table titles, name the depen-
Table 11.15. dent variable rst. Tables must be
Table 11.15 SPSS Procedures for Bivariate Analysis by Levels of Measurement

Independent Variable
Dependent Nominal Ordinal Ratio
Nominal CROSSTABS CROSSTABS CROSSTABS
MEANS*
Ordinal CROSSTABS CROSSTABS CROSSTABS
NONPAR CORRELATION NONPAR CORRELATION
Ratio MEANS MEANS CORRELATION
ANOVA ANOVA GRAPH
t Test t Test REGRESSION
*In SPSS for this case, run the independent variable as though it were the dependent variable ( i.e., name it rst); the
interpretation of the test of signicance would be standard.
numbered and be given a title. In pro- If the total is 99.9 or 100.1 percent, re-
viding a title for a table, the dependent port it as such.
variable is named rst, followed by the Rule 7. Report statistical test results be-
independent variable, followed by any low table. Any special information and
control variables. In Table 11.9, for ex- the results of statistical tests should
ample, note that the table title is Plans be reported below the line under the
to Attend University by Size of Home table. (Tests of signicance are dis-
Community. cussed in Chapter 12.) The preferred
Rule 2. Place dependent variable on method of presenting the probability
vertical plane. Label the categories of or signicance level is to report the ex-
the dependent variable and arrange act value (e.g., P 0.0037).
these categories on the left-hand side Rule 8. Interpret the table by comparing
of the table. If the categories involve categories of the independent vari-
some cut points, these should be spec- able. Because we are attempting to as-
ied (e.g., for income, the rst two cat- sess the impact of the independent
egories might be Under 10,000 and variable on the dependent variable ( in
10,000 to 19,999.) this case, size of community on educa-
Rule 3. Place independent variable on tional plans), we are interested in the
horizontal plane. Label the categories percentage of positive planners for
on the independent variable and ar- each category of the independent vari-
range them across the top of the table. able. Although about one half of the
Again, if there are cut points, be care- rural and small town students (52.3
ful to specify these. percent and 48.9 percent, respec-
Rule 4. Use variable labels that are tively) plan on attending university,
clear. Avoid the use of computer vari- some 73.9 percent of those from com-
able labels that have been designed to munities over 5000 have such plans.
meet the space requirements of the In short, compare percentages in each
statistical program. For example, FAED column. Usually it is sufcient to use
may have been used to refer to the one row ( in this case, just the row for
variable, fathers education. Use clear, those planning to attend college).
easily understood labels, as in Fa- Rule 9. Collapse data into fewer cate-
thers Education. gories. When dealing with small sam-
Rule 5. Run percentages toward the ples of fewer than 100 and with vari-
independent variable. Percentages ables having multiple categories, it may
should be computed so that each col- be necessary to regroup data (i.e., code
umn will total 100 percent. A percent- into fewer categories) before cross-
age is computed by dividing the col- tabular tables are produced. This pro-
umn total into the cell frequency. In the cedure is necessary to avoid having too
rst cell, for example, the computation few cases in the cells of the table. This
involves: recoding may involve both the indepen-
dent and the dependent variable. For
Cell total example, if you had a dependent vari-
Cell percentage 100
Column total able dealing with smoking behavior
69 among adolescents (e.g., smoker vs.
Cell percentage 100 52.3
132 nonsmoker) and you wished to relate it
to a ve category variable measuring
Rule 6. Report percentages to one deci- self-esteem (e.g., very low, low, medium,
mal point. Percentages should be re- high, very high), with a small sample,
ported to one signicant decimal point. you would need to collapse the inde-
pendent variable categories into two Table 11.16 Number of Cases in which
Pregnancy Is Possible
categories (e.g., high and low).
Rule 10. Minimize categories in a table Able to Become Pregnant? Cases, n
with a control variable. When control Yes 74
variables are used, it is necessary to No 66
minimize the number of categories in
Total 140
the independent and in the control
variables. Generally, there should be
no more than two or three categories
within these variables. There are two person could become pregnant, the best
major reasons for this limitation. First, strategy would be to guess the category
the number of cases in each cell will with the most cases. Each time a case is
become too small if there are many presented, our best guess is to say, Yes,
categories in either the independent can get pregnant. (There are more in the
or the control variable; second, the in- sample who can become pregnant than
terpretation of the table is very dif- those who cannot become pregnant.) If
cult if simplicity is not maintained. For we went through all the cases, guessing
example, if you were examining the re- yes each time, we would be right 74
lationship between gender and smok- times and in error 66 times.
ing behavior in adolescents control- Table 11.17 supplements the material
ling for the effect of age, you would contained in Table 11.16 by adding infor-
collapse the age variable (12 to 18 mation on gender. With this additional in-
years) into two categories (probably formation, would we be able to make
at the midpoint of the distribution). fewer errors?
Chapter 17 discusses interpretations This time, instead of using the Total
of three-variable cross-tabular tables. column and always guessing yes, we will
use the gender information as a basis for
b. Lambda our guess: If the case considered is male,
we will guess no, cannot become preg-
Researchers using cross-tabular analysis nant. If we do this, we will make zero er-
are interested in how closely two vari- rors; if the case is female, we will always
ables are related. One simple measure is guess yes. By following this procedure,
lambda. This statistic measures the pro- we will make a total of 6 errorsthe cases
portionate reduction in error that oc- of females who are not able to conceive.
curs in estimating a dependent variable, Given the additional information on gen-
given knowledge of the independent vari- der, we will now only make a total of 6 er-
able. If two variables are strongly associ- rors (0 6 6) in estimating whether a
ated, then errors in predicting variations respondent could become pregnant.
in the dependent variable will be consid- Lambda is based on how much error
erably reduced if information on the in-
dependent variable is taken into account.
The following example shows how Table 11.17 Relationship between Ability to
lambda is calculated. Suppose we wish to become Pregnant and Gender
measure the strength of the association Able to Become Gender
between gender and the ability to be- Pregnant? Male Female Cases
come pregnant. Table 11.16 reports the Yes 0 74 74
result of the appropriate medical exami- No 60 6 66
nations on 140 individuals.
Total 60 80 140
If we were asked to guess whether a
reduction occurs with the additional in- the information on gender. Taking into
formation provided by the independent account the information on gender, we
variable ( in this case, gender). Recall that make six errors60 fewer than we made
we made 66 errors when we did not have without the gender information.
Errors not knowing gender Errors knowing gender

Lambda
Errors not knowing gender
66 6
Lambda
66
Lambda 0.909
In this case, we have reduced the er- the values and its values fall in the range
rors in our estimate by 0.909 (propor- from 1 to 1. (Later you will nd that the
tion) or 90.9 percent. Lambda varies from correlation coefcient also varies from 1
0 to 1. The higher the value of lambda, the to 1.) A negative value indicates that the
more closely two variables are associ- variables are inversely related; the greater
ated. A high value on lambda indicates X, the less Y. A positive value indicates a re-
that knowing the additional information lation of the form the greater X, the greater
about the independent variable ( in this Y. The higher the value, the stronger the
case, gender) greatly reduces the num- association between the variables.
ber of errors one would make in guessing To illustrate the calculation of gamma,
the value of the dependent variable ( in Table 11.18 presents data showing the re-
this case, the ability to become preg- lationship between alcohol and drug use.
nant). A value close to 0 would indicate Note that drug use is taken as the depen-
that the additional knowledge of the in- dent variable and alcohol use is taken as
dependent variable leads to a slight pro- the independent variable.
portionate reduction in error. In most Is there an association between fre-
cases, the reduction in error will not be quency of alcohol use and frequency of
as dramatic as that found in the example drug use? The data seem to be contrary to
used above. our expectations when we look at the cat-
egory of people who use alcohol more fre-
c. Gamma quently. Only 2 of the 19 frequent alcohol
users use drugs once a month or more. In-
Gamma is a measure of the strength of as- deed, of the 16 people who report rarely
sociation used in cross-tabulation analy- or never using alcohol, 10 of them indi-
ses of ordinal level variables. This statistic cate frequent drug use. If anything, then,
takes advantage of the numerical order of there appears to be an inverse relation-
Table 11.18 Frequency of Drug Use by Frequency of Alcohol Use

Frequency of Alcohol Use
Frequency of Once per Week to
Drug Use Rarely or Never Once per Month Every Few Days
Never 3 8 3
Less than once a month 3 25 14
Monthly or more often 10 15 2
Total 16 48 19
ship between alcohol and drug use. Note that the value is negative. The value indicates
Gamma may be used to measure the a modest inverse relation between alcohol use and
strength of this association. Gamma is drug use.
based on two measures: (1) a measure of
a positive trend in the data and (2) a mea- To perform this analysis using SPSS, see
sure of an inverse trend in the data. the CROSS-TABS procedure in Appendix A.
A positive trend is one that shows that
as one variable increases the other vari- 2. Comparing Means
able increases; an inverse measure shows
that as one variable increases, the other A nurse researcher may be interested in
variable decreases. The positive trend is comparing the responses of two groups
reected by multiplying the number of of patients to a treatment. For example,
cases in each cell by a sum of all the cases you may want to know if touch therapy
that are both below and to the right of the reduces heart rate in elderly patients
cell. In the case of Table 11.18, this calcu- with arrhythmias. One way to determine
lation would be done as follows: this is to do a comparison of the mean
heart rate for those exposed to touch
Positive trend measure:
therapy with the mean heart rate of those
3(25 15 14 2) 168 receiving routine nursing care. Any time
one has a ratio level dependent variable
8(14 2) 128
( in this case, mean heart rate) and either
3(15 2) 51 a nominal or ordinal independent vari-
able, then it is appropriate to compute
25(2) 50
the mean values of the dependent vari-
Total positive trend 397 able for each category of the independent
variable. Table 11.19 presents data that
The inverse trend is reected by mul- would be appropriate for this kind of
tiplying the number of cases in cells by analysis. Note that whereas the depen-
the sum of the cases that are both below dent variable (heart rate) is measured at
and to the left of the cell. The calculation the ratio level, the independent variable
would be done as follows: is nominal (treatment group).
Inverse trend measure:
3(25 15 3 10) 159 a. The MEANS Procedure (SPSS)
8(3 10) 104 In cases in which there are many cate-
14(15 10) 350 gories in the independent variable, they
25(10) 250
will have to be regrouped into two or
three categories before the analysis is
Total inverse trend 863 run. The number of categories created
Gamma is calculated according to the following will depend on the following criteria:
formula: A reasonable number of cases will ap-
Positive Inverse pear in each category (often we try to
Gamma
Positive Inverse have roughly equal numbers in the
various categories).
397 863 The categories used must make theo-
Gamma
397 863 retical sense. We have to exercise cau-
tion to ensure that the categories re-
Gamma 0.3698 main as coherent as possible; thus, if we
Table 11.19 Mean Heart Rate by Treatment Group

Treatment Group Mean Heart Rate Number of Cases
Touch therapy 74.6 78
Routine treatment 77.1 77
Combined mean 75.8 155
If appropriate, test of signicance values are entered here (e.g., F 3.514; df 2,153; P 0.05).
were recoding religious afliation from are compared or when the means for two
eight categories to three, we would per- groups are compared at two or more
haps want to do the grouping so as to re- points in time in a single analysis (e.g., a
ect the degree to which the religious prepost experimental design), analysis of
categories we create either reect or re- variance is the procedure required to test
ject mainstream societal values. for statistically signicant differences in
the mean scores. The procedures involved
In interpreting the outcome of an
in doing an analysis of variance (ANOVA)
analysis, the mean values should be com-
require a measurement of two kinds of
pared. In Table 11.19, for example, the av-
variation: differences in scores within a
erage heart rate of patients in the touch
group (e.g., differences within the drug
therapy group are compared with those
group A) and differences of scores be-
in the routine therapy group.
tween the groups (differences that show
To perform this analysis using SPSS, see
up between the drug groups A, B, and C ).
the MEANS procedure in Appendix A.
An analysis of variance involves comput-
ing a ratio that compares these two kinds
b. Comparing Means Using the t Test of variabilitywithin-group and between-
group variability. In an experimental de-
When you are using samples under 30 and sign, the treatment effect would be tapped
you wish to compare two groups on a ra- in the measure of the between-groups vari-
tio level dependent variable, the t test is ability.
frequently used. The t test is used to de- On examining the data within each col-
termine if the differences in the means umn of Table 11.20 (on mean diastolic
may be regarded as statistically signi- blood pressure by drug type), note that
cant. Statistical signicance indicates only the variation within each group cannot be
whether observed differences in the mean explained by a connection with the treat-
scores are caused by real differences ment variable. (In the rst column, the
rather than chance differences. This is dis- drug A data are reported and because all
cussed in greater detail in Chapter 12. the cases are administered drug A, drug
To perform this analysis in SPSS, use the intake cannot explain variations within
Independent Samples t TEST procedure; this column. A similar situation exists
see Appendix A for details. within the columns for drugs B and C.) Dif-
ferences between the three groups, how-
c. Analysis of Variance ever, may be associated with the treat-
ment variable (type of drug taken) as each
Although used by many researchers, group received a different drug.
analysis of variance procedures are partic- In Chapter 12, the computations are pre-
ularly important to experimenters. The t sented for doing a one-way analysis of vari-
test is used to compare the means of two ance. In this particular case, the estimates
groups. When three or more group means of variance indicate more variation within
Table 11.20 Mean Diastolic Blood Pressure by correlational techniques. But the cost of
Drug Type
using these powerful statistical tools is
Drug A Drug B Drug C Total that more attention must be paid to mea-
84 84 80 surement. Correlational techniques as-
87 87 80
sume measurement at the ratio level. Al-
though this assumption may be relaxed,
69 80 78
the cost of doing so is that the strength of
75 71 68 the relationships between variables tend
86 83 78 to be underestimated (see Chapter 13 for
80 78 84
more on this point).
Given the importance of correlational
82 80 90
techniques, it is crucial that beginning re-
95 62 92 searchers understand the fundamentals
78 82 80 of these procedures. After they are un-
81 84 86
derstood, the more sophisticated proce-
dures are extensions of the simple ones.
Sums 817 791 818 2426
We have two basic concerns:
Means 817 791 818 808
What is the equation that describes the
Cases, n 10 10 10 30
relationship between the variables?
What is the strength of the relation-
ship between the two variables?
the groups than between them. Details for
An attempt is made to show how each
the computation and interpretation of
may be visually estimated; in addition, a
ANOVAs are presented in Chapter 12.
simple, intuitively obvious approach to
To perform this analysis in SPSS, use the
each computation is presented in Boxes
ONE-WAY ANOVA procedure; see Appen-
11.1 and 11.2.
dix A for details.
a. The Linear Equation: A Visual

3. Correlational Analysis Estimation Procedure
Correlational analysis is a procedure for Our rst concern is to determine the
measuring how closely two variables are equation that describes the relationship
related to each other. It is used with vari- between two variables. The general form
ables that are measured at the ratio level. of the equation is:
After the fundamentals of this statistical
technique are understood, the beginning Y a bX
researcher is in a position to grasp a
whole family of related advanced tech- The components of the equation are Y,
niques, including partial correlations, the dependent variable (starting salary to
multiple correlations, multiple regres- the nearest 10,000 dollars) and X, the in-
sion, factor analysis, path analysis, and dependent variable (years of postsec-
canonical correlations. ondary education); a is a constant that
A major advantage of using correla- identies the point at which the regres-
tional techniques is that many variables sion line crosses the Y axis; b refers to the
can be analyzed simultaneously. Multivari- slope of the regression line that de-
ate (many variables) analysis, whose com- scribes the relationship between the vari-
putations have been made easier through ables. The terms Y axis and regression
the use of computers, relies heavily on line are discussed below.
BOX 11.1 Calculating a Linear Equation
Having estimated the equation describing the re- b 19 26

lation between the variables, let us now compute
the actual equation. The table below presents b .73
the data and the computations necessary to de- Step 8: Inspect the regression line; if it slopes
termine the equation. The following steps are re- upward (highest on the right side), then the
quired: sign of the b will be positive (); if it slopes
downward (lowest on the right side), then
Step 1: Determine the mean value for the X and the b value will be negative ().
Y variables. This can be done by summing the Step 9: To determine the a value, apply the
values and dividing by the number of cases. formula:
Step 2: Subtract each value of X from the mean _ _
a Y bX
of X.
Step 3: Square the values determined in the a 5 .73(5)
previous step. a 1.35
Step 4: Subtract each value of Y from the mean
of Y. Step 10: The values may now be applied and
Step 5: Multiply the value determined in Step the final equation determined. The calculated
2 by those determined in Step 4. equation is:
Step 6: Sum all columns. Y 1.35 .73X
Step 7: To determine the b value: divide the
column total determined in Step 5 by the col- Recall that the visual estimation of the formula
umn total in Step 3. As in: was:
_ _ Y 1.33 .79X
(X X ) (Y_ Y )
b
(X X )2 Computing a Linear Equation
Step 1 Step _2 Step 3_ Step_4 Step 5_ _

X Y XX (X X )2 YY (X X ) (Y Y )
2 3 3 9 2 6
3 4 2 4 1 2
5 4 0 0 1 0
7 6 2 4 1 2
8 8 3 9 3 9
Total 0 26 0 19
_
X_ 5
Y5
General equation: Y a b(X ), where:
_ _
(X X )(Y_ Y ) 19
b .73
(X X )2 26
a 5 0.73 (5)
a 1.35
Hence, Y 1.35 0.73(X )
Note: The sign of the b coefficient is determined by inspection; if the slope of the regression line is positive (highest on
the right side), then the b coefficient is positive (); if it is negative (lowest on the right side), then the b coefficient is
negative.
BOX 11.2 Calculating a Correlation: Correlation Coefficient:

A Simple Computational Procedure
The following table presents the information dure outlined above. The steps are simple and
necessary to hand compute a correlation using can be quickly performed if there are only a few
a method which parallels the estimation proce- observations.
_
X Y (Y Y )2 Yp Y Yp (Y Yp)2
2 3 4 2.81* 0.19 0.0361
3 4 1 3.54 9.46 0.2116
4 5 1 5.00 1.00 1.0000
7 6 1 6.46 0.46 0.2116
8 8 9 7.19 0.81 0.6561
Totals 16 2.1154
*The Yp value is computed by substituting each value of X into the equation determined in Table 11.1. In
the rst observation, the computation would be Yp 1.35 0.73(2) 2.81.
Step 1: The first step is to determine the varia- Step 4: The previous three steps provide us
tion around the regression line. For each ob- with a measure of the variations around the
servation of X, we will need to compute the regression line. To get an estimate of the de-
predicted value for Y. To do this, we simply go viations around the mean of Y, we need only
to the equation determined in Section c, plug look at the sum for the column:
in the value for X, and solve. The first obser- _
vation would be done as follows: (Y Y )2
Yp 1.35 0.73(X) Step 5: We are now able to plug the values

into the formula:
Yp 1.35 0.73(2)
Variations around regression
r2 1
Yp 2.81 Variations around mean of Y
(Y Y_p)2 N
The predicted values for Y are determined for r2 1
each case in the manner described for the first (Y Y )2 N
observation. .423
r2 1
Step 2: The second step is to compute how 3.2
much each observation deviates from its
predicted value. (Y Yp). r2 1 0.13
Step 3: The third step is to square the results
r2 0.87
of the previous step. After this is completed,
this column should be summed. (Y Yp)2. r 0.93
Note: If the regression line is highest on the right side, the r value will be positive; if it is lowest on the right side of
the plot, the r value will be negative.
For purposes of illustration, we will Y

use the data shown in Table 11.21. 8
Step 1. The rst step in visually estimating 7

the equation that describes the rela-
tionship between the variables would 6
be to plot the relationship on graph pa- 5
per. To have fairly accurate estimates,
it is necessary to plot carefully and to 4
ensure that units of measurement of
the same size are used on both dimen- 3
sions of the graph. Figure 11.6 shows
2
what such a graph would look like.
Note that the dependent variable (Y ) is 1
plotted on the vertical axis and the in-
dependent variable (X ) is plotted on X
the horizontal axis. 0 1 2 3 4 5 6 7 8
Step 2. Insert a straight regression line so Figure 11.6 Scatterplot of sample data.
that the vertical deviations of the
points above the line are equal to the
vertical deviations below the line.
sents the constant, or the a value, in
There need not be the same number of
the regression equation. Note that in
points above and below the line, nor
Figure 11.7, we have estimated that it
need any of the points necessarily fall
crosses the Y axis at 1.33.
right on the line. The regression line of-
Step 4. Draw a line parallel to the X axis
fers the best linear description of the re-
and one parallel to the Y axis to form a
lationship between the two variables.
right-angled triangle with the regres-
From the regression line, one can esti-
sion line, similar to that shown in Figure
mate how much one has to change the
11.8. Measure the lines in millimeters.
independent variable in order to pro-
(In Figure 11.8, the vertical measures 72
duce a unit of change in the dependent
variable. The following is a hint to lo-
cate where the regression line should Y
be drawn. Turn a ruler on its edge; then 8
move the ruler to achieve both minimal
7
deviations from it and equal deviations
on both sides of the ruler (Fig. 11.7). 6
Step 3. Observe where the regression line
crosses the Y axis; this point repre- 5
4
Table 11.21 Sample Data Set
3
X Y
2 3 2
3 4
1
5 4
7 6 X
0 1 2 3 4 5 6 7 8
8 8 Figure 11.7 Inserting the regression line.
Y can compare the results with those we got

8 using the visual estimation procedure.
Note that we have come fairly close to
7
the computed gures. In a research proj-
6
ect, we would have the computer generate
the a and the b values using the REGRES-
5 SION procedure. Estimating equations is a
72 mm
good exercise to become familiar with the
4 different elements involved in regression
analysis.
3
2 the REGRESSION procedure in Appendix A.
91 mm In some cases, the a value will turn out
1 to be negative; this simply means that the
regression line crosses the Y axis below
X the X axis. It should be noted as well that
0 1 2 3 4 5 6 7 8
as the b value increases, the regression
Figure 11.8 Estimating the equation.
line is steeper. In this case, smaller incre-
ments in the X variable lead to increments
in the Y variable. A negative value on the
mm and the horizontal measures 91 b indicates a negative slope, a situation in
mm.) Divide the horizontal distance which the data indicate a relationship in
into the vertical distance; this compu- which the greater X, the less Y.
tation will provide our estimated b Beginning researchers should recog-
value. (In our gure, 72 91 0.79.) nize that with a linear equation, it is pos-
Step 5. If the slope of the regression line sible to predict the value of a dependent
is such that it is lower on the right- variable given a value of the independent
hand side, the b coefcient is negative, variable. When nurse researchers speak
meaning the more X, the less Y. If the of prediction, this is often the sense in
slope is negative, use a minus sign in which they are using the term. Figure 11.9
your equation.
Step 6. The visual estimation of the equa- Y
tion describing the relationship be- 8
tween the variables is determined by
7
simply adding the a and b values to the
general equation: 6
Y a bX 5
In our illustration, the values would be 4

as follows: 3
Y 1.33 0.79(X ) 2
The above formula is our visually esti- 1

mated equation of the relationship be-
tween the two variables. Box 11.2 pre- X
0 1 2 3 4 5 6 7 8
sents a calculation of the actual equation.
Figure 11.9 Visually Estimating Y from X.
After the calculations have been made, we
shows how one could visually estimate gression line increases. Plot 5 shows a
the predicted value of Y (income), given a curvilinear relationship in which the plot
value of X (years of postsecondary educa- goes in one direction and then switches to
tion). The procedure simply involves lo- another one. Plot 6 shows a case in which
cating the X value on the X axis, moving the linear correlation is zero but in which
vertically to the regression line, and then there is a fairly strong association between
moving horizontally to the Y axis. The the variables. Plots 7 and 8 show situations
point at which the Y axis is intersected to be wary ofnamely, those where a few
represents the visual estimate of the Y deviant cases can radically shift the slope
variable. of the regression line. The change of two
A predicted value is computed using points in the two plots shifts the sign of the
an equation in which values of the inde- correlation from a positive one (plot 7) to
pendent variable (or variables) is (are) a negative one (plot 8).
plugged into the equation. Suppose, for One reason that it is important to plot
example, that we attempted to predict the out relationships is to permit a visual in-
values of Y given X values of 1, 4, and 6. To spection of the results. If there are extreme
solve the problem, we would simply use values or if the plot indicates a nonlinear
the equation computed above and then relationship, then a linear correlation anal-
determine the predicted values of Y, as in: ysis would be inappropriate.
The correlation coefcient (r) is a mea-
Computed equation: Y 1.35 0.73(X )
sure of the strength of association be-
Yp
tween two variables. The correlation may
with X value of 1: Y 1.35 0.73(1) 2.08 vary from 1 to 1. Perfect correlations
are rare, except when a variable is corre-
with X value of 4: Y 1.35 0.73(4) 4.27
lated with itself; therefore, almost all of the
with X value of 6: Y 1.35 0.73(6) 5.73 correlations are represented by values
preceded by a decimal point, as in 0.98,
0.37, or 0.56. Negative correlations mean
We use the same procedures in situa-
that there is a negative slope in the rela-
tions in which there are multiple inde-
tionship, meaning the more X the less Y.
pendent variables determining the pre-
Let us now develop an intuitively sim-
dicted values of a dependent variable,
ple way of estimating the strength of the
except that there are more values to be
relationship between two variables. Ex-
plugged into the equation. amine the rst four plots shown in Figure
11.10. Note that the closer the plotted
b. Correlation Coefficient: A Visual points are to the regression line, the
Estimation Procedure higher the correlation. Conversely, the
more points diverge from the regression
In learning to visually estimate a correla- line, the lower the correlation. In estimat-
tion, it is important to develop some sense ing the correlation coefcient, there are
of what correlations of different magnitude two kinds of variability that we have to be
look like. Figure 11.10 presents graphs of concerned with: (1) variations around
eight relationships. In the rst four, the the regression line and (2) variations
correlation coefcients vary from 0.99 to around the mean of Y.
0.36. Note that if the correlation dropped We can determine the ratio between
below the 0.36 level, it would become dif- these two types of variability. In essence,
cult to determine where the regression line the correlation coefcient (r) reects this
should be drawn. At the other end of the ratio so that the higher the ratio, the
continuum, note that correlations drop higher the correlation. Indeed, we can
fairly slowly as the scatter around the re- represent the relationship as follows:
Plot 1 Plot 2
Y Y
r = 0.99 r = 0.85
X X
Plot 3 Plot 4
Y Y
r = 0.53 r = 0.36
X X
Plot 5 Plot 6
Y Y
r = 0.0 r = 0.0
X X
Plot 7 Plot 8
Y Y
r = 0.70 r = -0.70
X X
Figure 11.10 Eight linear correlations.

r2 1
Variations around mean of Y
As an exercise in trying to visually es- Y

8 8mm
timate the strength of a correlation, the
following steps may be taken: 7
44mm
Step 1. Plot the data on graph paper and
5
draw in an estimated regression line.
Again, remember that the same units 4
of measurement must be used on both 3
dimensions of your graph.
Step 2. Draw in a line parallel to the X axis 2
that will cut through the estimated 1

mean value of Y.
X
Step 3. To estimate the deviations around 1 2 3 4 5 6 7 8
the regression line, draw in an addi-
tional regression line parallel to the Figure 11.11 Estimating a correlation.
original one for the points on or above
the existing regression line. (You may
want to cover the points below the re- Your estimation is based on the idea
gression line to avoid confusion.) Now that the correlation reects the ratio of
draw yet another regression line paral- variations around the regression line to
lel to the other two for the points below variations around the mean of Y. As the
the original regression line. Measure variations around the regression line be-
and record the perpendicular distance come relatively smaller, the correlation
between the two new regression lines. rises. Conversely, as the two measures of
Step 4. To estimate the deviations around variation approach equality, the correla-
the mean of Y, two additional lines par- tion approaches zero. Although visually
allel to the mean of Y line must be estimated correlations are never re-
drawnthe rst for those points above ported because they are not exact, the ex-
the line and the second for those points ercise is an excellent one for becoming fa-
below the line. Again, the perpendicu- miliar with the meaning of the correlation
lar distance between these new lines coefcient. If you have few cases, it is
should be measured and recorded. easy to hand compute the exact correla-
Step 5. At this point, you should have tion using the steps outlined in Box 11.2.
drawn a graph similar to the one However, with more than 15 or 20 cases,
shown in Figure 11.11. you should probably have the computa-
To estimate the correlation, simply en- tions done by a computer.
ter the values from your graph into the Note that the computations in Box 11.2
following equation: have led to results roughly similar to
those achieved using the visual estima-
r2 1 tion procedures. (Usually the results will
not be so close!) Having done a few visual
8 estimations and a few hand calculations
r2 1
44 of correlations, you should have a good
understanding of simple correlations.
r2 .82
There are many statistical techniques
r .91 that are extensions of correlational tech-
niques. After the basics are understood, 6. Computing Partial Correlations

then computations may be turned over to
a computer. A partial correlation is a special type of
correlation that may be used with ratio
level variables. It measures the strength
4. Plotting the Data of association between two variables
while simultaneously controlling for the
It is a good idea to produce a scatterplot effects of one or more additional vari-
of important relationships in your study. ables. For example, if a nurse researcher
A plot will alert you to problems such as: is interested in studying the relationship
between perceived health status and the
A few extreme cases that may be inu- engagement in healthy lifestyles, he or
encing the correlation between the vari- she would probably want to control for
ables (see Fig. 11.10, plots 7 and 8) the socioeconomic status of the partici-
A strong relationship exists, but it is pants. Providing the variables are mea-
not linear and the correlation does not sured at the ratio level, partial correla-
reect the true strength of the relations would provide an analytical solution
tionship (see Fig. 11.10, plots 5 and 6) to the problem.
There are a lot of data points with the In partial correlations, we adjust the
same value; this lack of variation may values of the dependent ( in this case, en-
alert you to a problem in the measure- gagement in healthy lifestyles) and the
ment of one of the ratio variables in- independent variable ( in this case, per-
volved in the plot ceived health status) in order to take into
Many computer programs, including account the inuence of the control vari-
SPSS, can produce such plots quickly. able ( in this case, socioeconomic status).
The advantage of partial correlations
over cross-tabular table analysis is that:
5. Computing Spearman We make use of all of the data (by not
Correlations recoding variables into two or three
categories as would be done in a cross-
A nurse researcher may wish to test a re-
tabular analysis (CROSSTABS).
lationship between variables that are
measured at the ordinal level. For exam-
We can work with fewer cases without
running into cell-size problems as hap-
ple, the nurse may want to explore the
pens frequently with cross-tabular table
relationship between distance to hospi-
analysis.
tal (under 1 mile, 1 to 5 miles, 5 miles and
over) and frequency of visits to emer- Similar to ordinary correlations, par-
gency departments (low frequency, me- tial correlations take on values from 1.0
dium, and high frequency) to identify through to 1.0. Partial correlations con-
one of the factors related to emergency trol one or more independent variables.
department use. Spearman correlations The number of controls determines the
are the appropriate measurement of as- order of the partial. A correlation with
sociation when the variables are mea- one control variable is a rst-order partial
sured at the ordinal level. The details of correlation; one with two controls is a sec-
such computations may be checked in ond-order partial correlation, and so forth.
any elementary statistics text and are Incidentally, ordinary correlations are
not presented here. Think of them as be- sometimes referred to as zero-order cor-
ing similar to the correlation procedures relations, which simply means that there
just discussed. are no control variables in the analysis.
The strategy involved in partial corre- the numbers of the two major variables
lations is that regression equations can and these are separated from the num-
be used to express the relationship be- bers of the control variables by a . as in:
tween each pair of variables in the equa-
tion. For any value of an independent r12.3 .56
variable, it is possible to predict the value
of the dependent variable while adjusting In this case, we have a rst-order partial
for the inuence of the control variables. correlation reported, with a value of 0.56.
The idea of residuals is also useful in This value represents a measure of the
understanding partial correlations. In the strength of association between variables
three-variable case, if the possible combi- 1 and 2, controlling for variable 3. A third-
nations (XY, ZY, and XZ ) are plotted order partial correlation simply designates
and a regression line is entered for the three control variables, as in:
XY relation, we could argue that the de-
viations from the line are the result of the r12.345 .28
inuence of factor Z plus that of other
known and unknown factors. These devi- Partial correlations will be used in test-
ations are residuals. They arise when we ing causal models in which the variables
allow one variable to explain all the varia- involved are measured at the ratio level
tion that it can in another variable; what is of measurement (see Chapter 17).
left unexplained (deviations from the re- Because it is easy to hand-compute a
gression line) are the residuals. By corre- rst-order partial, the formula is pre-
lating residuals, we can get a measure of sented here. It may be used if the re-
the amount of inuence a third variable searcher has the zero-order correlation
has on the rst relationship (XY ), inde- matrix.
pendent of the second relationship (XZ).
The notational convention that we use r12(r13)(r23)
r12.3
in referring to partial correlations shows
1 3 1
r12 r2
3
2
E X E R C I S E S
1. As an exercise in learning to identify Nursing Student Grades
different levels of measurement, go for a Mid-term Test
through a sample questionnaire and
88 69
identify the level of measurement
achieved for each question. 72 93
93 92
77 74
94 85
2. Calculate the mode, median, mean,
range, standard deviations, and vari- 79 87
ance for the nursing grades in the ac- 85 82
companying table. 66 71
79 74
83 86
3. Using the mean and standard devia- tween the respondents number of vis-
tion calculated above, use Z scores to its to the doctors ofce and the num-
estimate the percentage of nursing stu- ber of visits to the doctors ofce by
dents you would expect to score 70 the mother.
percent or below. What percentage Sample Data for Visits to the Doctors Office
would you expect to score over 75 per-
Respondent Father Mother Sister
cent? What percentage would you ex-
pect to score between 80 and 90 per- 1 3 1 4
cent? 2 3 4 3
3 4 3 5
9 7 6 6
4. Equally weighting income and educa- 10 10 10 8

tion, use Z scores to calculate each 12 11 9 14
persons socioeconomic status from 14 13 15 8
the following data set:
5 4 7 2
6 7 5 8
Standard
Given the following: Mean Deviation 8 8 10 12
Education 14.0 4.0

Income $47,000 $13,000
Person A Education: 12.0 years
Income: $88,000 6. Compute the r and the equation. How
Person B Education: 16.0 years close were your estimates to the ac-
Income: $105,000 tual value?
Person C Education: 11.0 years
Income: $44,900
Person D Education: 18.0 years
Income: $92,500 7. Using the formula computed in ques-
Person E Education: 13 years tion 6, what would you predict the
Income: $42,500 respondents number of visits to the
doctors ofce to be if the mother
made 1, 6, or 15 yearly visits?
8. Using visual estimation procedures,

5. Using the data given here on a respon-
estimate the correlation between mo-
dents, fathers, mothers, and sisters
thers and sisters number of visits.
visits to a doctors ofce in the past
12 months, plot the data (treating the
respondents number of visits to a
doctors ofce as dependent) and go
9. Using visual estimation procedures,
through the procedures to visually es-
estimate the linear equation that de-
timate the correlation and the equa-
scribes the relationship between re-
tion that describe the relationship be-
spondents (dependent) and the fa- 10. Is a batting average an actual aver-
thers number of visits to the doctors age? From what you have learned in
ofce. this chapter, how might batting aver-
ages be described better?
Mitchell. M., and Jolley, J. (2000). Research De- Norwood, S.L. (2000). Research Strategies for
sign Explained. Orlando: Harcourt Brace Jo- Advanced Practice Nurses. Upper Saddle
vanovitch College Publishers. This text River, NJ: Prentice Hall Health. Three
prepared by two psychologists is highly clearly written chapters on the most fre-
readable and even entertaining at times. quently used statistical procedures in
Munro, B.H. (2001). Statistical Methods for nursing research.
Health Care Research. Philadelphia: J.B. Polit, D. (1996). Data Analysis and Statistics for
Lippincott. Although primarily authored Nursing Research. Stamford, CT: Appleton
by Munro, this book also contains addi- & Lange. This very practical text that ex-
tional articles on statistical methods used amines the why, when, and how of statisti-
in health care by other authors. cal analysis.
Chapter 12
A Primer for Inferential

Statistics
CHAPTER OUTLINE
A. Key Elements in Tests of Statistical 1. Elements in a Chi-Square Test

Significance 2. Learning How to Hand Compute Chi-
1. The Research Hypothesis and the Null Squares
Hypothesis C. The t Test of Statistical Significance
2. The Sampling Distribution 1. Between-Subjects t Test
3. One- and Two-Tailed Tests of 2. A Within-Subject t Test
Signicance
D. The F Test of Statistical Significance
4. Signicance Levels
5. Probability Levels 1. One-Way Analysis of Variance
6. Type I and Type II Errors 2. Two-Way Analysis of Variance
7. Degrees of Freedom 3. Repeated Measures Analysis of Variance
8. Six Steps in Tests of Signicance E. When Tests of Statistical Significance
B. The Chi-Square Test of Statistical Are Not Appropriate
Significance (2)
KEY TERMS
Analysis of variance One-way analysis of variance t Test

Between-subjects Probability level Test of significance
Chi-square test (2) Null hypothesis Two-tailed test
Clinical significance Repeated measures analysis Two-way analysis of variance
Critical value of variance Type I error
Degrees of freedom Research hypothesis Type II error
F distribution Significance level Within-subject
Inferential statistics Standard error of the means
One-tailed test Statistical significance
389
I
nferential statistics are a group of sta- Other such tests are based on similar
tistics that do two basic things. First, principles and, therefore, need not be ex-
they enable us to make judgments plored in detail by beginning researchers.
about the accuracy of a given sample in re- What each of these statistics has in com-
ecting the characteristics of the popula- mon is that each of them allows us to de-
tion from which it was drawn. For exam- termine if the differences we are observ-
ple, in a random sample of 300 men age 25 ing can be characterized as statistically
to 45 years, 30 percent of the men were signicant; that is, the differences we are
smokers. Inferential statistics allow us to observing in the data are probably true dif-
generalize from this limited sample to the ferences and not the result of chance uc-
larger population and draw conclusions tuations in sampling. Because tests of sig-
about the percentage of male smokers in nicance are often misunderstood, this
the population based on the sample data. chapter also reviews the conditions un-
Given the sample size, we can be 95 per- der which such tests are not appropriate.
cent certain that the true population mean Chapter 18 explores several more ad-
will be 30 percent plus or minus 4 percent. vanced techniques.
In other words, we can be 95 percent cer- As a precautionary note, beginning re-
tain that the true population mean falls be- searchers should be aware that although
tween 26 and 34 percent the ndings of a study may be statistically
Second, inferential statistics are used signicant, they may not be clinically sig-
in testing hypotheses, which is the ability nicant. To be clinically signicant, the
of statistics to help us make decisions ndings must have meaning for patient
about which study outcomes reect uke care in the presence or absence of statis-
differences between groups and which tical signicance (Fain, 1999). Statistical
ones reect true differences. signicance indicates that whereas the
Suppose that you, as a graduate student ndings are unlikely to result from chance,
in nursing, wish to determine if listening to clinical signicance requires the nurse to
certain types of music has a calming im- interpret the ndings in terms of their
pact (as measured by pulse rate and blood value to nursing. For example, a nurse
pressure) on patients after surgery. Sup- researcher may determine that two
pose that you wished to compare those groups of clients are signicantly different
placed in a group to hear the music with a in terms of the number of hours of sleep
control group who watched television reported for each group. One group has
instead. After collecting your data, you received an average of 13 minutes more of
would somehow have to determine if your sleep per 24-hour period than the other.
music therapy was making a difference in Although the groups are statistically sig-
calming the patients. Could any measured nicantly different, the ndings have no
differences simply be the result of chance clinical signicance because the magni-
factors or is that unlikely? Inferential sta- tude of the difference is not great enough
tistics to the rescue! to change practice.
Chapter 2 discussed deriving testable On the other hand, suppose we nd
hypotheses from theoretical proposi- that although pain reduction scores us-
tions. Such hypotheses are tested using a ing a new therapy are not statistically sig-
variety of inferential statistics. Inferential nicant, a review of the literature and our
statistics, therefore, are a key element in own research indicate that the new ther-
testing theories of human behavior. This apy does tend to reduce pain. In this case,
chapter explores the three most widely because there appears to be an enhance-
used tests of statistical signicance: the ment of patient comfort using the new
chi-square test, the T test, and the F test. therapy and the trends appear to be pos-
A PRIMER FOR INFERENTIAL STATISTICS 391
itive, we would take the new therapy as earnings of male and female nurse ad-
clinically signicant. ministrators . . .
It is the null hypothesis that is tested

the proposition that there is no relation-
A. KEY ELEMENTS IN TESTS OF ship between the variables. The test will
STATISTICAL SIGNIFICANCE lead us either to accept or to reject the
null hypothesis. If the null hypothesis is
A number of values are reported beneath accepted, we conclude that the associa-
most statistical tables found in journal ar- tion or the difference may simply be the
ticles. Almost certainly one of these val- result of sampling uctuations and may
ues will indicate whether the differences not reect an association or difference in
reported in the table are statistically sig- the population being studied. The impli-
nicant. What does this mean? A test of cation is that if we accept the null hy-
signicance reports the probability that pothesis, then the research hypothesis is
an observed association or difference is false. If the null hypothesis is rejected,
the result of sampling uctuations and then we argue that there is an association
not reective of a real difference in the between the variables in the population
population from which the sample has and that this association is of a magni-
been taken. tude that probably has not occurred be-
cause of chance uctuations in sampling.
If the null hypothesis is rejected, the data
1. The Research Hypothesis and are examined to see if the association is
the Null Hypothesis in the predicted direction; if it is, then
this is one piece of evidence that is con-
Tests of signicance are used to test hy-
sistent with the research hypothesis.
potheses. These are set up in a research
(But note that the nding may be in the
and in a null form. The research hy-
opposite direction than the one pre-
pothesis (also commonly known as the al-
dicted, in which case the evidence is not
ternative hypothesis) is simply a predic-
consistent with the prediction.)
tion of the relationship between the
Nurse scientists typically avoid ex-
variables; the null hypothesis states that
pressions such as proof or conclu-
there is no relationship between the vari-
sively conclude. Instead, they prefer a
ables. The following statements illustrate
more tentative phrasing such as: The ev-
research and null hypotheses:
idence supports the view that . . . , or
Research hypothesis 1: The greater The data are consistent with the re-
the physical participation, the better search hypothesis. The reason for the
the health status. cautious wording is that empirical scien-
Null hypothesis 1: There is no rela- tists can never provide absolute proof of
tionship between levels of physical the accuracy of a research hypothesis.
participation and health status. Future research may well disconrm or
Research hypothesis 2: Controlling qualify in important ways the ndings of
for qualications, achievements, and any research. Therefore, when the re-
experience, male nurse administrators sults of a study lead to the rejection of a
are paid more than their female coun- null hypothesis, it only means that there
terparts. is probably a relationship between the
Null hypothesis 2: Controlling for variables under examination. Simply put,
qualications, achievements, and ex- the possibility that the research hypoth-
perience, there is no difference in the esis is correct has not been ruled out.
When a null hypothesis is accepted, tudes toward the use of birth control. Sim-
however, it is taken to mean that the pre- ilarly, it would be important if we found
dicted relationship in the research hy- that there is no signicant difference be-
pothesis is false. And even this conclusion tween men and women in their grief re-
must remain tentative because of the pos- sponse to a diagnosis of terminal illness.
sibility that luck or error may have played
a role in the results. When the evidence 2. The Sampling Distribution
leads to the acceptance of the null hypoth-
esis, it indicates to the researcher that the A key to understanding tests of signi-
research hypothesis may well be in error. cance is to recognize that in drawing one
(Box 12.1) sample (e.g., containing 50 individuals),
Often new researchers think that they we are getting a unique collection of re-
have somehow failed if they do a project spondents who are assumed to represent
and nd out that they have to accept the the larger population from which the sam-
null hypothesis. Such should not be the ple was selected. If another sample were
case. To accept the null hypothesis is (1) drawn from that same population, we
a research nding and (2) it may well be would have another unique collection of
just as important to nd out that two vari- individuals that is slightly different from
ables are not associated as it is to nd out the rst sample. If 1000 such samples
that they are. It might be extremely im- were drawn and the means of the same
portant to discover, for example, that variable for each of the samples were plot-
among people younger than age 25 years, ted, a normal distribution curve would re-
there is no signicant difference between sult, albeit a peaked, or leptokurtic, one
Catholics and Protestants in their atti- (see Chapter 15 for more details).
Suppose, for example, the means of the
BOX 12.1 Terminology weights of respondents are plotted. Al-
Conundrum: Do You though these mean weights might range
Accept a Null from 159 to 169 lb for the men, the major-
Hypothesis or Do You ity of the samples would cluster around
Fail to Reject a Null the true mean of 164 lb. Note that we are
Hypothesis? not plotting the individual weights of the
respondents; rather, we are plotting the
There appears to be some variation in us- mean weight of each of the 1000 samples
age here. The common practice among drawn. The distribution is quite peaked
some disciplines such as psychology is to because we are plotting the mean weight
use the fail to reject terminology. Al-
for each sample. To measure the disper-
though this is probably the most accurate
sion or range of variability of the means of
phrasing, it does carry with it an unfortu-
nate, unintended implication. It conveys the these samples, we use a statistic known as
idea that the researcher has somehow the standard error of the means. If the
failed if a statistically significant association standard deviation of the population was
is not found. 14 lb, with repeated samples of size 50,
For pedagogical reasons, we prefer the the standard error of the means would be
accept the null hypothesis terminol- given by the following formula:
ogy that does not carry with it the idea of
sd population
failing if no significant association is found. Standard error of means
Thus it, is easier to convince students that N
it is okay to have findings that are not sta- 14
Standard error of means
tistically significant. 50

Standard error of means 2
When any one sample is drawn, the if a variable in the population is not nor-
sample mean will fall somewhere within mally distributed, the means will be nor-
the sampling distribution curve. In doing mally distributed when repeated sam-
tests of signicance, we are assessing ple means are plotted. For example,
whether the results of our one sample fall weights of a population of males and fe-
within the null hypothesis acceptance males are bimodal, but if we did re-
zone (usually 95 percent of the distribu- peated samples, the means of sample
tion) or outside this zone in one of the weights (in which the males and females
tails of the distribution, in which case we are combined) will be normally distrib-
will reject the null hypothesis. uted. By using larger sample sizes and
Although beginning researchers are plotting mean values of the samples, the
not likely to be doing methodological re- distribution of means will be normally
search and selecting repeated samples, it distributed because the large sample
is necessary to understand such sampling size and the use of mean values aver-
distributions in order to understand sam- ages out the variability in the data.
pling and related tests of signicance.
There are four key points that can be
made about probability sampling proce- 3. One- and Two-Tailed Tests of
dures where repeated samples are taken: Significance
1. Plotting the means of repeated sam- Recall that a sampling distribution has a
ples will produce a normal distribu- peak and two tails. If our research hy-
tion. Note that this distribution, how- pothesis predicts the tail into which our
ever, will be more peaked than that ndings will fall if the null hypothesis is to
achieved when raw data are plotted be rejected, then we will use a one-tailed
(Fig. 12.1). test; if the tail is not predicted, then we do
2. The larger the sample sizes, the more a two-tailed test. Thus, we speak of a one-
peaked the distribution and the closer tailed test because we are predicting
the means of the samples to the popu- which particular tail of the sampling dis-
lation mean, as shown in Figure 12.2. tribution curve the result will fall into if
3. The greater the variability in the popu- the null hypothesis is to be rejected. If no
lation, the greater the variation in the prediction about the direction of the re-
samples. lationship is made, then a two-tailed test
4. When sample sizes are above 100, even is in order. For example:
(i) Raw data (ii) Means of samples
Figure 12.1 Distribution of raw data versus means of samples.

(i) Sample of 20 (ii) Sample of 100
Figure 12.2 Sample size and the normal distribution.
A one-tailed research hypothesis: Fe- the sampling distribution curve that the
males are less approving of physical difference will fall into, then we have made
conict than males. (Note that the a one-tailed prediction; if we simply pre-
prediction indicates which gender will dict a difference without specifying into
be less approving.) which tail the difference will fall, then we
A two-tailed research hypothesis: would do a two-tailed test.
There is a difference between males It should be noted that it is easier to re-
and females in their approval of physi- ject a null hypothesis if a one-tailed test is
cal conict. (Note that there is no pre- conducted. The more conservative strat-
diction here about which gender will egy is to do two-tailed tests. However, a
be more approving.) strong argument could be advanced to en-
courage more use of one-tailed tests. Es-
A test of signicance measures the like- pecially in situations in which a clear pre-
lihood that an observed difference (e.g., diction of the direction of the relation can
the difference in approval of conict be made or derived axiomatically from a
scores between males and females) falls theory, then one-tailed tests should be fa-
within normal sampling uctuations and vored. Given measurement problems in
therefore does not reect a real difference many of our projects, we should not un-
between the males and the females in the duly stack the deck against the research
population. Think of the sampling dis- hypothesis (see Chapter 13).
tribution curve and imagine we have plot-
ted difference scores between men and
womenon the right hand side, we have 4. Signicance Levels
situations in which the females have
higher tolerance for violence scores; on Figure 12.3 shows two sampling distribu-
the left side, we show the scores for less tion curves; the rst one has the 5 per-
tolerance of violence. In our study, if there cent rejection area split between the two
is less than a 5 percent chance that the tailsthis is a two-tailed testand the
magnitude of the observed relationship is second one has the 5 percent rejection
the result of sampling uctuations, then area all in one tail, indicating a one-tailed
the null hypothesis is rejected. We are con- test. The same principle applies to tests
cluding that there is probably a real differ- at the 1 percent level, only now the dif-
ence between the male and female toler- ference between the males and the fe-
ance of conict. If we predict the tail of males would have to be greater in order
Two-tailed Test One-tailed Test
2.5% 2.5% 5.0%
Rejection area Rejection area Rejection area
Figure 12.3 Five percent probability rejection area: One- and two-tailed tests.
to fall into the null hypothesis rejection analysis; therefore, in contemporary re-
area. These tails represent the signi- porting, the exact P value is typically
cance levels of our statistical tests. If the noted in the form P 0.037. Note that if a
results of the test fall into the tail of the probability level were indicated at 0.43, it
sampling distribution, we can be 95 per- would mean that 43 percent of the time
cent condent that the result is not the observed differences could be ex-
caused by chance uctuations in the sam- pected to occur on a chance basis.
ple we have studied.
6. Type I and Type II Errors

5. Probability Levels
In doing tests of signicance, we have to
When calculated, the test statistic will balance the chances of making wrong
provide a probability that the result is conclusions. If we are terribly conserva-
caused by sampling uctuations. This is tive and set standards that are tough to
reported in various ways, normally la- meet, we are more prone to what are
beled as the p value. Researchers should called type II errorsaccepting a null
understand that this value needs to be hypothesis when it should be rejected. If
less than 0.05 ( if using the 5 percent sig- we reject a null hypothesis that should be
nicance level) or 0.01 ( if using the 1 per- accepted, we have made what is known
cent level). Normally, probability levels, as a type I error. There is a trade-off be-
along with the test statistic raw score, are tween the two types of errorif we use
reported in any publication of the re- the 0.001 level of signicance, we will be
search. In cases in which the test has more prone to type II errors. The proba-
been manually calculated, the researcher bility of making a type I error is the same
typically reports the probability as P as the signicance level of the test. Recall
0.05 ( in the case in which the probability that if we use the 0.05 level of signi-
is less than 0.05; alternatively, if the null cance, we will reject the null hypothesis
hypothesis is to be accepted, the note only if there is a less than 5 percent
would be P 0.05). Researchers using chance that the difference we are observ-
various software packages such as Statis- ing may be caused by chance sampling
tical Package for the Social Sciences factors. But there may well be consistent
(SPSS) will have the exact probability of differences that are present in the popu-
the result occurring because of chance lation that are rather slight and hence
provided for them on the output of the rarely will prove to be statistically signif-
icant. As a researcher, you have to decide B. THE CHI-SQUARE TEST OF

what signicance level to use in order to STATISTICAL SIGNIFICANCE
balance off the risk of type I versus type II (2)
errors. Most researchers in the behav-
ioral sciences use the 0.05 level. The chi-square test is a test of signi-
Incidentally, if you have trouble re- cance used primarily in cross-tabular
membering the difference between type I table analysis and is used when:
and type II errors, here is a little memory
jogger that might help: RAARReject
The dependent variable is a nominal
one.
when you should Accept; Accept when
you should Rejectthe rst two letters
You wish to determine if frequencies
vary across categories.
give you type I, the second two letters,
type II. Remember, too, that the reference
The expected frequencies are above 5
in most cells of the table.
is always to the null hypothesis.
The variables are normally distributed.
The measures of the variables are in-
dependent of one another.
7. Degrees of Freedom
Essentially, the chi-square compares
The test statistic has to take into account the observed frequencies with the ex-
the complexity of the relationship being pected frequencies that would occur if
analyzed. To do this, degrees of freedom there is no difference between the cate-
are calculated. The number of variables, gories in the population ( i.e., if the null
the number of categories within vari- hypothesis is correct). We begin by pre-
ables, or the total number of cases may senting the elements in a chi-square test
inuence the degrees of freedom in an followed by three illustrations showing
analysis. In the chi-square test, for exam- different applications of the test.
ple, the degrees of freedom increases
with the number of cells there are in the
cross-tabular table. Each test of signi-
1. Elements in a Chi-Square Test
cance has a method of calculating de- a. Observed and Expected
grees of freedom, and the method for cal- Frequencies
culating them is shown in the various
boxes below. In cross-tabular analysis, the number of
cases that would be expected to fall into
any one cell, assuming that there is no dif-
8. Six Steps in Tests of ference between the categories of the in-
Significance dependent variable (i.e., assuming the null
hypothesis is correct) are referred to as
1. State the research and null hypothesis. expected frequencies. For example, if you
2. Determine where the outcome will fall compared having had suicide thoughts
in the sampling distribution if the null of male versus female respondents, the ex-
hypothesis is to be rejected. pected frequencies would be that an equal
3. Indicate the chosen signicance level proportion of males and females would
to be used in the test (usually 0.05). have had such thoughts. Observed fre-
4. Compute the test statistic. quencies are the counts of the behavior
5. Note whether the test indicates if you (e.g., suicide thoughts) recorded in each
should accept or reject the null hy- cell of the table. The chi-square provides a
pothesis. calculation of the probability of getting the
6. Interpret the ndings. gap that is observed between the observed
and the expected frequencies on the basis the results of computer tests of statistical
of sampling uctuations. There are several signicance with ease.
ways of calculating expected frequencies;
they are illustrated in this chapters boxes.
a. A One-Sample Test: Are There
Significantly More Red Balls in the
b. The Chi-Square Formula Container?
The chi-square test (also known as 2) is Suppose that we have a huge container
dened by the following equation: lled with a mixture containing thousands
of red and white balls. Now suppose that
(fo fe)2 we draw a sample of 200 balls from the
x2
fe container and that we get 114 red ones and
where fo is the frequency observed and fe is the fre-
86 white ones. Can we safely conclude that
quency expected. Note that the formula reflects the there are more red balls in the container
amount of deviation from the expected values in re- than white ones? Perhaps if we drew an-
lation to the magnitude of the expected values. other sample, we would nd more white
than red balls. We will use the chi-square
c. Critical Value test to try to answer this question.
We will follow the six steps in con-
When hand calculating tests of signi- ducting tests of signicance:
cance, it is necessary to determine if your
1. Research hypothesis: There will be a
calculated value exceeds the critical value.
signicant difference in the number of
The critical value is determined by looking
red versus white balls in the container
up the value on a statistical table. There is
(note that this implies a two-tailed test).
a table for each test, and it requires you to
Null hypothesis: There is no difference
know the appropriate degrees of freedom
in the number of red versus white balls
and whether you are using a one- or two-
in the container.
tailed test. In order for a null hypothesis to
2. If the null hypothesis is to be rejected,
be rejected, the critical value must be
the result will fall into either tail of the
equaled or exceeded by the computed test
sampling distribution. This is a two-
value. (When a computer program is used,
tailed test.
these values are not necessary because
3. We will use the 0.05 signicance level
the exact probabilities are provided on
in the test, meaning that if the result
the computer output.)
falls into either the 2.5 percent left tail
or the 2.5 percent right tail of the dis-
2. Learning How to Hand tribution, we can be 95 percent certain
Compute Chi-Squares that there is a signicant difference in
the number of red versus white balls in
In order to fully understand the compu- the container.
tation of chi-squares, three examples are 4. Box 12.2 presents a hand calculation of
presented, each following the six steps of the test statistic.
conducting tests of signicance. Work 5. The results reported in Box 12.2 would
through each of the boxes and at the end lead us to reject the null hypothesis be-
of the exercise you should have an un- cause the difference observed would
derstanding of how the calculations are occur by chance less than 5 percent of
done. Each of the three represents a dif- the time. The calculated chi-square
ferent use of the test. After you under- was 3.92 and because this value is
stand them, you will be able to interpret greater than 3.84 (the critical value we
BOX 12.2 A Sample Chi-Square Computation: Red Ball/White

Ball Example
Step 1 Step 2 Step 3 Step 4 Step 5

Color fo fe fo fe (fo fe)2 (fo fe)2

fe
Red 114 100 14 196 1.96
White 86 100 14 196 1.96
Total 3.92
Chi-square 3.92; df 1 signicant at the 0.05 level; required critical value to reject the null hypothesis is
3.84 (two-tailed test, 1 degree of freedom, at 0.05 level); see Table 12.1. Decision: Because the chi-square
value is 3.92 and this exceeds the critical value, the null hypothesis is rejected.
The steps in the calculation are as follows: the remaining cells could be determined by
subtraction, given that the total expected
Step 1: Create a calculation table that contains must equal the total observed. In the red
the headings as shown above. Enter the ob- ball/white ball example, after we determine
servations that have been made for each of that we would expect 100 red balls (50 per-
the cells. In cross-tabular tables, the conven- cent of all balls selected is implied by the null
tion for naming cells is to label them by start- hypothesis) and because there were 200 balls
ing in the top left corner of the table, naming selected in all, then 100 would be expected to
them a, b, c, and so on across the table, then be white). Because we only had to calculate
continuing at the beginning of the next line. one of the cells before the other could be de-
In the current example, we have two cells, a termined by subtraction, we have 1 degree of
(red) and b (white). freedom.
Step 2: Compute the frequency expected Step 8: Decide the significance level. Usu-
(fe) for each of the cells. In this case, we sim- ally you will choose either the 0.01 or 0.05
ply theoretically determine what the ex- level. This means that you are willing to ac-
pected frequency would be if the null hy- cept that a difference is statistically signifi-
pothesis is correct (in the red/white ball cant with a 1 or 5 percent chance of being
example, the theoretical expectation would wrong. We will use the 0.05 level for this ex-
be that one half would be red and the other ample.
half white). Therefore, in our total sample of Step 9: Determine whether you are doing a
200, we would expect to get 100 red balls one- or a two-tailed test of significance. Re-
and 100 white ones. member that if the direction of the relation
Step 3: Subtract the expected from the fre- is predicted, then the appropriate test will be
quency observed (fo) for each of the cells. one-tailed because we are predicting which
Step 4: Square the values determined in Step 3. particular tail of the sampling distribution the
Step 5: Divide the result of Step 4 by the fre- result will fall into if the null hypothesis is to
quency expected. be rejected. If no prediction about the direc-
Step 6: Sum the column of values determined tion of the relationship is made, then a two-
in Step 5. This value is the raw chi-square. tailed test is required. In our example, the re-
Step 7: Determine the degrees of freedom. search hypothesis did not indicate which
The number of cells in which the expected color would occur most frequently, so we
frequency would have to be computed before will do a two-tailed test.
BOX 12.2 A Sample Chi-Square Computation: Red Ball/White

Ball Example (Continued )
Step 10: Look up the critical chi-square value cally significant, and at what significance level
found in Table 12.1 to determine if the dif- (we are using 0.05).
ference between the expected and the ob- Step 12: In the case of one-tailed tests when the
served frequencies is statistically significant. null hypothesis is rejected, the data must then
For a two-tailed test at the 0.05 level with 1 be inspected to see if the data are consistent
degree of freedom, the critical value is 3.84. with the research hypothesis. (It is possible, of
To be statistically significant, our computed course, that the difference could be statisti-
chi-square value (3.92) must equal or exceed cally significant, but in the opposite direction
the critical value. Because 3.92 exceeds the predicted by the research hypothesis.) If the
3.84 critical value, we are able to reject the difference is not statistically significant, then
null hypothesis. the null hypothesis is accepted. In the case of
Step 11: Below the table report the raw chi- our example, we reject the null hypothesis and
square value, degrees of freedom, and conclude that there are more red balls than
whether the difference observed is statisti- white ones in the container.
look up in Table 12.1), our decision is draw. Another way to think about it is
to reject the null hypothesis. to say that we can be condent that we
6. We can then conclude that there are will make the correct decision 19 out of
more red balls than white balls in the 20 times (95 percent).
container. The risk we take in making
that conclusion is that we may, by In attempting to more clearly under-
chance, just happen to have overse- stand tests of signicance, one might note
lected red balls. A difference of that that if repeated samples were taken and
magnitude, however, would occur in the results plotted (e.g., the percentage of
fewer than 5 percent of the samples we red balls), the plot would resemble a nor-
Table 12.1 Critical Values for the Chi-Square Distribution, One- and Two-Tailed Tests*
0.05 Level of Significance 0.01 Level of Significance
Degrees of Freedom One-Tailed Two-Tailed One-Tailed Two-Tailed
1 2.71 3.84 5.41 6.63
2 4.61 5.99 7.82 9.21
3 6.25 7.82 9.84 11.34
4 7.78 9.49 11.67 13.28
5 9.24 11.07 13.39 15.09
6 10.64 12.59 15.03 16.81
7 12.02 14.07 16.62 18.48
8 13.36 15.51 18.17 20.09
9 14.68 16.92 19.68 21.67
10 15.99 18.31 21.16 23.21
*Please consult the appendix of any statistics text if you require additional degrees of freedom or other levels of
signicance.
mal distribution curve. As sample sizes be- We will follow the six steps in con-
come larger, a more peaked (leptokurtic) ducting tests of signicance:
distribution would result, which means
1. Research hypothesis: Alternative ther-
that larger samples require smaller differ-
apy clinic users overrepresent the
ences in order to reject the null hypothe-
higher family incomes from the com-
sis. Most times we draw one sample and
munity (note that this implies a one-
have no way of knowing if it is a truly rep-
tailed test). Null hypothesis: There is
resentative sample. Perhaps our sample is
no difference in the family incomes of
atypicalwe just happened to draw one
alternative therapy clinic users and
with a preponderance of red balls and we
the family incomes in the community.
might, therefore, wrongly conclude that
there are more red balls in the container.
the result will fall into the tail of the sam-
But because it would take too long to
pling distribution representing high in-
count the thousands of balls in the con-
comes. This is a one-tailed test.
tainer, we simply take the chance of being
3. We will use the 0.05 signicance level
wrong. We can reduce that risk by chang-
in the test, meaning that if the result
ing the signicance level from 0.05 to 0.01.
falls into the 5 percent right tail, we
That would reduce, but not eliminate, the
can be 95 percent certain that there is
chance of coming to an incorrect conclu-
a signicant overrepresentation of the
sion.
wealthier families using the alterna-
tive therapy clinics.
b. One Sample Compared with a 4. Box 12.3 presents a hand calculation of
Regional Distribution: Is There a the test statistic.
Relationship Between Family 5. The results reported in Box 12.3 would
Income and the Use of Alternative lead us to reject the null hypothesis be-
Therapy Clinics? cause the difference observed would
occur by chance less than 5 percent of
There are situations in which one wishes the time. The critical value that must be
to compare a cohort of respondents with exceeded in order to reject the null hy-
a known distribution. Table 12.2 reports pothesis is 4.61 (see Table 12.1; note
family incomes of clients attending an al- that we look up the value for two de-
ternative therapy clinic that is privately grees of freedom using the one-tailed
funded along with the income distribution test column because we are predicting
of all families in the vicinity of the clinic. Is the direction of the relationship). Be-
there a signicant difference between the cause the calculated chi-square value is
income distribution of the general popu- 45.61 and this exceeds the critical value
lation in the area and the income distri- of 4.61, we reject the null hypothesis.
bution of users of the alternative therapy 6. We can then conclude that there is a
clinic? statistically signicant overrepresen-
Table 12.2 Alternative Therapy Clinic Users and General Population Family Incomes
Income, dollars Clinic Users, n Sample, % General Population, %
Over 100,000 30 15.0 7.8
40,000 to 99,999 160 80.0 68.9
Under 40,000 100 5.0 23.3
Total 200 100.0 100.0
BOX 12.3 Computing Chi-Square for Alternative Therapy Clinic

Users Family Income Data
Income Category, dollars fo fe fo fe (fo fe)2 (fo fe)2

fe
Over 100,000 30 15.6* 14.4 207.36 13.29
40,000 to 99,999 160 137.8 22.2 492.84 3.57
Under 40,000 10 46.6 -36.6 1339.56 28.75
Totals 200 200 45.61
Chi-square 45.61; df 2; probability 0.05.

*The expected frequencies are computed on the presumption that if there is no difference between the
general population and the families of clinic users, then the income distribution would have similar propor-
tions in the various categories. In the rst cell, where there were 30 clinic users, we would have expected 15.6
(200 .078 15.6). In the general population, 7.8 percent of the families had incomes above $100,000.
Degrees of freedom 2. Two cells would degrees of freedom, at the 0.05 level); see
have to be computed before the final one Table 12.1.
could be calculated by subtraction, as in fe cell Decision: Because the chi-square value is 45.61
c [200 (15.6 137.8)]. and this exceeds the critical value, the null hy-
Required critical chi-square value to reject pothesis is rejected.
the null hypothesis is 4.61 (one-tailed test, 2
tation of the wealthier families using procedures are presented for the compu-
the alternative therapy clinic. The risk tation of the chi-square.
we take in making that conclusion is We will follow the six steps in conduct-
that we may, by chance, just happen ing tests of signicance:
to have overselected wealthier fami-
1. Research hypothesis: There is a differ-
lies. A difference of that magnitude,
ence in drug use by gender (note that
however, would occur in fewer than 5
this implies a two-tailed test). Null hy-
percent of the samples we draw. An-
pothesis: There is no difference in
other way to think about it is to say
drug use by gender.
that we can be condent that we will
make the correct decision 19 out of 20
the result will fall into either tail of the
times.
sampling distribution, this being a two-
tailed test.
c. Calculating a Chi-Square for a 3. We will use the 0.05 signicance level
Typical Cross-Tabular Analysis: in the test, meaning that if the result
Frequency of Drug Use by Gender falls into the 2.5 percent right tail or
the 2.5 percent left tail, we can be 95
Box 12.4 presents a two-variable cross- percent certain that there is a signi-
tabular table showing the relationship cant difference in drug use by gender.
between frequency of drug use by gen- 4. Box 12.4 presents a hand calculation of
der. Beneath the table, the computational the test statistic.
BOX 12.4 Frequency of Drug Use by Gender
Gender
Male Female Total
Frequency of Drug Use in Lifetime N % N % N %
No experience 47 34.8 63 49.2 110 41.8
Once or twice 51 37.8 39 30.5 90 34.2
Three or more times 37 27.4 26 20.3 63 24.0
Totals 135 100.0 128 100.0 263 100.0
Chi-square 5.689; df 2; probability 0.05.
The chi-square could be hand computed in

the following manner:
Cell fo fe (fo fe) (fo fe)2 (fo fe)2/fe

a 47 56.5 9.5 90.25 1.597
b 63 53.5 9.5 90.25 1.687
c 51 46.2 4.8 23.06 0.499
d 39 43.8 -4.8 23.06 0.526
e 37 32.3 4.7 21.72 0.673
f 26 30.7 -4.7 21.72 0.707

5.689
Required critical value to reject the null hypothesis is 5.99 (two-tailed test, 2 degrees of freedom, at the 0.05
level); see Table 12.1.
Decision: Because the chi-square value is 5.689 and this does not equal or exceed the critical value (5.99),
the null hypothesis is accepted.
Further Notes: ing expected frequencies can be computed in a

Determining expected frequencies: The eas- similar manner or determined by subtracting
iest way to calculate the expected frequencies the calculated cells from the marginal total; cell
in a cell of a standard cross-tabular table is to e could be determined by adding together the
multiply the row marginal total by the column computed expected frequencies for cells a and
marginal total and divide this result by the total c and this total could be subtracted from the
number of cases in the table. (Exclude missing column marginal total, as in 56.5 46.2
cases from the table before calculations are 102.7, then 135 102.7 32.3.
made.) For cell a, the calculation would be (110 Determining degrees of freedom: The
135)/263 56.5. For cell c, the calculation easiest method for determining degrees of free-
would be (90 135)/263 46.2. The remain- dom in standard cross-tabular tables (such as
BOX 12.4 Frequency of Drug Use by Gender (Continued )
the one used in this example) is to count the Degrees of freedom 2

number of rows and columns in the table and
We could also use the slower method of
use the following formula:
figuring out how many cells we would have to
Degrees of freedom (# rows 1) hand calculate the expected frequencies be-
(# columns 1) fore we could get the rest by subtraction. The
Degrees of freedom (3 1) (2 1) answer is 2.
5. The results reported in Box 12.4 would C. THE t TEST OF

lead us to accept the null hypothesis be- STATISTICAL SIGNIFICANCE
cause the difference observed would
occur by chance more than 5 percent
The t test is a popular test of statistical sig-
of the time. The critical value that
nicance most commonly used in studies
must be exceeded in order to reject
requiring the comparison between two
the null hypothesis is 5.99 (see Table
groups on a measure or a comparison of
12.1; note that we look up the value for
measures taken at two different times. Sup-
two degrees of freedom using the two-
pose that an experimental control group
tailed test column because we are not
design has been used to measure the ac-
predicting the direction of the rela-
quisition of psychomotor injection admin-
tionship). Because the calculated chi-
istration skills (measured at the ratio level)
square value is 5.689 and this does not
and that the study involves the compari-
equal or exceed the critical value of
son of 20 nursing students, 10 taught in the
5.99, we accept the null hypothesis.
conventional manner and 10 using a new
6. We can then conclude that there is not
interactive computer methodology. What
a statistically signicant relationship
test would be appropriate here? We could
between drug use and gender. The dif-
collapse the psychomotor skill scores into
ferences between the genders ob-
high and low scores and then run a chi-
served in drug use may well be the re-
square. But this would not make full use of
sult of random sampling uctuations;
our data. In effect, we would be throwing
we cannot be 95 percent certain the
out all the variability within the two grade
differences observed reect a real
categories. Instead, it would make more
difference in drug use. Note, however,
sense to compare the means of the psy-
that had we been doing a one-tailed
chomotor skills scores of students ex-
test, we would reject the null hypothe-
posed to the different teaching techniques.
sis. In this case, the decision to do a
And because we have a small sample, a t
two-tailed test has had a crucial im-
test would do what we want.
pact on the interpretation of the result.
The t test is used most often in cases in
To perform this analysis using SPSS, see which:
the CROSSTABS procedure in Appendix A.
The next two sections introduce tests Sample sizes are small (typically under
that involve the comparison of means be- 30).
tween categories. Each test assumes that The dependent variable is measured at
the dependent variable is measured at the ratio level.
the ratio level. We will begin with two ver- Assignment to groups has been done
sions of the t test. independently and randomly.
The treatment variable has two levels, in the test, meaning that if the result
presence and absence. falls into the 2.5 percent right tail or
The population from which the sample the 2.5 percent left tail, we can be 95
was drawn is normally distributed; percent certain that there is a signi-
therefore, the distribution of sample cant change in blood pressure levels
means would be normally distributed. by type of touch.
The researcher wants to nd out if 4. Box 12.5 presents a hand calculation of
there are statistically signicant differ- the test statistic.
ences between the groups. 5. The results reported in Box 12.5 would
lead us to accept the null hypothesis be-
Two commonly used variants of the t cause the difference observed would
test are presented in this section. The rst occur by chance more than 5 percent of
is appropriate for a between-subjects de- the time. The critical value that must be
sign and the second is appropriate for a exceeded in order to reject the null hy-
within-subject design. The t test repre- pothesis is 2.10 (In Table 12.3; note that
sents the ratio between the difference in we look up the value for 18 degrees of
means between two groups and the stan- freedom, using the two-tailed test col-
dard error of the difference. The standard umn because we are not predicting the
error of the difference reects the average direction of the relationship). Because
amount of deviation in the two comparison the calculated t test value is 1.46 and
groups (the formula for it is provided in this does not equal or exceed the criti-
Box 12.5) The t test formula is as follows: cal value of 2.10, we accept the null hy-
Difference between means pothesis.
t 6. We conclude that there is not a statis-
Standard error of the difference
tically signicant relationship between
1. Between-Subjects t Test change in blood pressure readings and
type of touch. The differences in out-
Box 12.5 shows how one could compute comes by type of touch may well be
the t score in a between-subjects design the result of random sampling uctua-
for differences in blood pressure read- tionswe cannot be 95 percent cer-
ings resulting from the administration of tain the changes observed reect a
touch therapy by nursing students to real difference in the consequences
women subjected to articially induced of the type of touch therapy applied.
stress in a laboratory setting. This was re-
Although the average change in dias-
ported in Table 4.2.
tolic blood pressure readings for subjects
We will follow the six steps in con-
in the therapeutic touch group is slightly
ducting tests of signicance:
higher, there was a lot of variation within
1. Research hypothesis: There is a differ- the two groups. The test of signicance
ence in the change in blood pressure leads us to accept the null hypothesis:
readings for those receiving therapeutic there is simply not enough between-group
touch therapy versus those receiving variation compared to within-group varia-
casual touch (note that this implies a tion to reject the null hypothesis.
two-tailed test). Null hypothesis: There
is no difference in the change in blood
pressure by type of touch therapy. 2. A Within-Subject t Test
the result will fall into either tail of the A variation of the t test is available for de-
sampling distribution, this being a two- pendent samples (in which the compari-
tailed test. son groups are not selected indepen-
3. We will use the 0.05 signicance level dently, as in repeated measures designs in
BOX 12.5 Between-Subjects t Test for Equal-Sized Groups (From

Therapeutic Touch and Blood Pressure Data, Chapter 4)
In Chapter 4, data were reported comparing using each type of touch. This was a between-
therapeutic touch with casual touch on blood subjects design with random assignment to
pressure (BP) readings. The means and stan- groups and was conducted by student nurses.
dard deviations were reported for diastolic BP The data were as follows:

I. Student nurses:
Between-subjects design;
random assignment to groups Kathleen 10 Paula 12
Li 9 Marlies 7
Danielle 11 Vanny 6
Kim 8 Paula 6
Kevin 12 Sandra 11
Mary 4 Marius 1
Yvonne 5 Chris 2
Tara 7 Andrea 4
Ursula 8 Tony 7
Holly 10 Carol 8
Mean 8.40 6.40
Using the following formula we can quickly number of subjects in group 1; N2 refers to the
compute the t test: number of subjects in group 2.
_ _
X1 X 2
t Step 1: Subtract the mean change in BP for the
Standard error of the difference
therapeutic touch group from the mean BP
The standard error of the difference may be change for the casual touch group (8.406.40
computed using the following formula: 2.00). This is the value we will use in the nu-
merator (the value above the line in the equa-

sd sd 2 2
1 2 tion).
N N
1 2 Step 2: For the denominator (below the line in
where sd12 is the variance for group 1 and sd22 the equation), we require the variance for
is the variance for group 2. N1 refers to the each type of touch therapy and because we
BOX 12.5 Between-Subjects t Test for Equal-Sized Groups (From Therapeutic

Touch and Blood Pressure Data, Chapter 4) (Continued )
have the standard deviations reported in the

Standard

table, all we have to do is square the standard 6.487 12.264
deviations to get the variances. Hence: error of
10 10
difference
Variance therapeutic touch: Standard error of difference 1.370
sd2 2.5472 6.487
Variance casual touch: Step 5: Plug the values into the equation for
sd2 3.5022 12.264 the t test, as in:
2.00
t test score
Step 3: The n (number of cases) for therapeu- 1.37
tic touch is 10; the n for casual touch is also t test score 1.46
10.
Step 4: Calculate the values for the denomina- Step 6 Determine the degrees of freedom by
tor by using the values calculated in Step 2, subtracting number of groups from number
using the equation given above. of cases (20 2 18).
Step 7: Decision: Because the t value does not
exceed the one shown in Table 12.3, we ac-
Standard

sd 2
sd 2 cept the null hypothesis. The differences in
error of
1 2
N N the BP readings may simply be the result of
difference 1 2
sampling fluctuations.
which we collect data from the same sub- in the test, meaning that if the result
jects run under each of the treatment con- falls into the 2.5 percent right tail or
ditions; these are known as paired sample the 2.5 percent left tail, we can be 95
designs). The steps for hand computing a percent certain that there is a signi-
dependent t test are reported for the cant change in blood pressure levels
within-subject design touch therapy data by type of touch.
originally reported in Chapter 4 (Box 12.6). 4. Box 12.6 presents a hand calculation of
We will follow the six steps in con- the test statistic.
ducting tests of signicance: 5. The results reported in Box 12.6 would
lead us to reject the null hypothesis be-
1. Research hypothesis: There is a differ- cause the difference observed would
ence in the change in blood pressure occur by chance less than 5 percent of
readings for those receiving therapeutic the time. The critical value that must
touch therapy versus those receiving be exceeded in order to reject the null
casual touch (note that this implies a hypothesis is 2.26 (see Table 12.3; note
two-tailed test). Null hypothesis: There that we look up the value for 9 degrees
is no difference in the change in blood of freedom using the two-tailed test col-
pressure by type of touch therapy. umn because we are not predicting the
2. If the null hypothesis is to be rejected, direction of the relationship). Because
the result will fall into either tail of the the observed t test value is 6.71 and
sampling distribution, this being a two- this exceeds the critical value of 2.26,
tailed test. we reject the null hypothesis.
3. We will use the 0.05 signicance level 6. We conclude that there is a statisti-
Table 12.3 Critical Values for the T Distribution: One- and Two-Tailed Tests*
0.05 Level of Significance 0.01 Level of Significance
Degrees of Freedom One-Tailed Two-Tailed One-Tailed Two-Tailed
1 6.31 12.71 31.82 63.66
2 2.92 4.30 6.96 9.92
3 2.35 3.18 4.54 5.84
4 2.13 2.78 3.75 4.60
5 2.02 2.57 3.36 4.03
6 1.94 2.45 3.14 3.71
7 1.90 2.36 3.00 3.50
8 1.86 2.31 2.90 3.34
9 1.83 2.26 2.82 3.25
10 1.81 2.23 2.76 3.17
11 1.80 2.20 2.72 3.10
12 1.78 2.18 2.68 3.06
13 1.77 2.16 2.65 3.01
14 1.76 2.14 2.62 2.98
15 1.75 2.13 2.60 2.95
16 1.75 2.12 2.58 2.92
17 1.74 2.11 2.57 2.90
18 1.73 2.10 2.55 2.88
19 1.73 2.09 2.54 2.86
20 1.72 2.09 2.53 2.84
25 1.71 2.06 2.48 2.79
30 1.70 2.04 2.46 2.75
40 1.68 2.02 2.42 2.70
60 1.67 2.00 2.39 2.66
120 1.66 1.98 2.36 2.62
1.64 1.96 2.33 2.58
signicance.
cally signicant relationship between differences in the mean change in di-

change in blood pressure readings and astolic blood pressure achieved with
type of touch. The differences in out- the two different types of touch is
comes by type of touch are not likely slight (1.00 U ), yet the difference in
to be caused by random sampling uc- this case is found to be statistically sig-
tuationswe can be 95 percent cer- nicant (we reject the null hypothe-
tain the changes observed reect a sis). Compared with the student
real difference as a result of the type nurses (see Box 12.5), the expert nurse
of touch therapy applied. Note that the practitioners using the two touch ther-
apies produced much less variation in analysis of variance techniques. These

blood pressure readings (standard de- are introduced next.
viations of 0.816 and 0.823 compared
with 2.547 and 3.502, respectively, for
the student nurses).
When a test of signicance is re- D. THE F TEST OF STATISTICAL
quired to test the signicance of differ- SIGNFICANCE
ences between means for a study with a
more complex experimental design, A further test of signicance associated
such as those with several independent with comparing two or more group means
variables or more than two different involves an analysis of variance. The dis-
levels in the treatment or control vari- tribution associated with this test is the F
able, the researcher typically uses distribution and is used to test whether
BOX 12.6 Testing Touch Therapies: A Within-Subject Design (From

Therapeutic Touch and Blood Pressure Data, Chapter 4)

III. Expert nurses:
within-subject design Jamie 12 Jamie 10
Petra 11 Petra 10
Margo 11 Margo 11
Jill 13 Jill 12
Sandra 13 Sandra 12
Dale 12 Dale 11
Paula 11 Paula 10
Teresa 11 Teresa 10
Linda 11 Linda 10
Melanie 12 Melanie 11
Mean 11.70 10.70

BOX 12.6 Testing Touch Therapies: A Within-Subject Design (From Therapeutic

Touch and Blood Pressure Data, Chapter 4) (Continued)
Calculating a Within-Subject (Dependent Groups) t Test
Difference
Time 1 Time 2 Between
Therapeutic Casual Time 1 and Average
Touch Touch Time 2 Difference
(D) (AD) AD D (AD D)2
Subject Step 1 Step 2 Step 3 Step 4
Jamie 12 10 2 1 1 1
Petra 11 10 1 1 0 0
Margo 11 11 0 1 1 1
Jill 13 12 1 1 0 0
Sandra 13 12 1 1 0 0
Dale 12 11 1 1 0 0
Paula 11 10 1 1 0 0
Teresa 11 10 1 1 0 0
Linda 11 10 1 1 0 0
Melanie 12 11 1 1 0 0
Totals 10 2
Average Difference 10 10 1.0

Variances of differences 1 9 .111
Step 1: Subtract time 2 measure from time 1 Step 4: Square each value calculated in Step
measure to get the difference score (D) (in 3.
the case of Jamie, 12 10 2). Step 5: Sum the squared values in Step 4 and
Step 2: Calculate the average difference score divide by the number of subjects minus 1
(AD): Sum the values for the 10 subjects (2 (10 1 9). The measure of the variance
101111111 of the differences is 2 9 0.222
10); divide sum by number of subjects (10 Step 6: To get the standard deviation of the dif-
10 1). ferences, calculate the square root of the
Step 3: Subtract the difference score (D), cal- variance of the differences, as in:
culated in Step 1, from the average difference
score (AD) calculated in Step 2: (in the case Standard deviation of differences
of Jamie, 2 1 1). 0 .2
22 0.471
BOX 12.6 Testing Touch Therapies: A Within-Subject Design (From Therapeutic

Touch and Blood Pressure Data, Chapter 4) (Continued)
Step 7: Calculate the standard error of the dif- Step 9. Determine the degrees of freedom by
ference by dividing the standard deviation of subtracting 1 from the number of pairs of
the differences by the square root of the scores (10 1 9). Look up the t value in
number of pairs of scores, as in: Table 12.3. The two-tailed value with 9 df
2.26.
Standard error of the differences
Step 10. Decision: Because the calculated
0.471 1 0 0.149
t value (6.71) exceeds the critical value
Step 8: Compute the t ratio by dividing the av- (2.26) in Table 12.3, we reject the null hy-
erage difference (AD) by the standard error pothesis.
of the difference, as in:
t 1 0.149
t 6.71
there is a signicant difference in the To illustrate this procedure, we will exam-

means of various categories. Some of this ine data on identity awareness scores in
material will not seem to be new because adolescents. The researcher wishes to
similar ideas were presented in Chapter examine whether these scores vary signi-
11 when discussing both the chi-square cantly for early, middle, and late adole-
test and correlation analyses. Analysis of scence. Identity is the central developmen-
variance is typically used when: tal task of adolescence. Identity awareness
was measured using six items that reect
The dependent variable is measured at how adolescents make choices by explor-
the ratio level. ing alternatives and committing to roles.
The treatment variable has two or The sample involves a total of 182 respon-
more levels. dents from the early (age 10 to 13 years),
In cases in which two or more treat- middle (age 14 to 17 years), and late ado-
ments are used simultaneously. lescent (age 18 to 20 years) age groups.
It is assumed that the population from Our research hypothesis is that identity
which the sample was drawn is nor- awareness scores will vary by age group.
mally distributed and, therefore, that The null hypothesis is that there is no dif-
the distribution of sample means ference in identity awareness scores
would be normally distributed. across age groups.
The researcher wants to nd out if Box 12.7 presents the analysis of our
there are statistically signicant differ- study of identity awareness in adoles-
ences between the groups. cents. Various computations are made
The researcher wants to check to see and the resulting test of signicance, the
if the various treatments may be inter- F test, indicates that the null hypothesis
acting with one another. should be rejected. The differences in the
mean scores on identity awareness (27.35
1. One-Way Analysis of Variance in early adolescence, 30.60 in middle ado-
lescence, and 34.35 in late adolescence)
A one-way analysis of variance provides a are more than one would expect from
test of signicance for the effect of a inde- normal sample uctuations. We conclude,
pendent variable on a dependent variable. therefore, that there is a statistically sig-
BOX 12.7 One-Way Analysis of Variance of Identity Awareness by

Adolescent Age Group Identity Awareness Scores
Early Adolescence Middle Adolescence Late Adolescence Total

34 36 37
28 49 34
24 38 33
31 33 24
30 27 31
30 34 43
... ... ... ... ...
Sums 1989 1559 2061 5609
Means 30.60 27.35 30.82
Cases, n 65 57 60 182
Steps in Computing a One-Way Analysis of Step 4: The between variation, which is the vari-
Variance ability resulting from the different treatments
Three measures of variation are required to (age groups), may be computed by squaring
compute the analysis of variance. These are total the column totals, dividing this value by the
variation, between variation, and within variation: number of cases, then summing the result,
and finally subtract the value in step 2 above,
Total variation Between variation Within as in:
variation
((19892 65) (15592 57)
The total variation, which is all the variability in (20612 60)) 172862 1437
the data, may be computed by summing the
squares of all the identity awareness scores and Step 5: The within variation, which is the vari-
subtracting from that result the square of the ability resulting from differences among
sum divided by the total number of cases, as in: members within each age group, may be de-
termined by simply subtracting the between
Step 1: 342 282 . . . 432 186677 variation from the total variation, as in:
minus
Step 2: 56092 182 172862 13815 1437 12378
equals
Step 3: 186677 172862 13815 Step 6: An F ratio table may be computed as
Total variation 13815 follows:
Sums of Degrees of Mean
Squares Freedom Squares F F Prob.
Total variation 13815 N 1 181
Between variation 1437 k12 718.4 10.39 .0001
Within Variation 12378 N k 179 69.2
where k number of categories in the independent variable.

BOX 12.7 One-Way Analysis of Variance of Identity Awareness by Adolescent

Age Group Identity Awareness Scores (Continued )
Step 7: To estimate the mean squares (popu- across the top of the table (in the sample case
lation variance) for the between and within df 2); the within variation df is arranged along
rows, divide the sums of squares for the row the vertical column (in our table, df 179; see
by the degrees of freedom for the row, as in: table in Step 6 above for these values). We will
test the hypothesis at the 0.05 level. The crit-
i) Between estimate of mean squares: ical value given in the table is 3.07. The F ratio
1437 2 718.4 must equal or exceed this value in order to re-
ii) Within estimate of mean squares: ject the null hypothesis.
12378 179 69.2 Step 10: Because the F ratio for the sample
data is 10.39 (exceeding the critical value),
Step 8: To calculate the F ratio divide the within the null hypothesis is rejected. Variability of
variation (69.2) into the between variation the magnitude reported in the identity
(718.4): awareness scores is regarded as statistically
significant. If the differences between the age
718.4 69.2 10.39 groups had been smaller and we were led to
accept the null hypothesis, we would argue
Step 9: If the probability is not calculated for you that the differences observed simply reflect
(in cases in which the values are hand calcu- sampling fluctuations and we cannot reason-
lated or when using some software package), ably conclude that there is a real difference
you will need to determine the value that has in the identity awareness scores of the re-
to be exceeded in order to reject the null hy- spondents from each of the three age groups.
pothesis by looking up the value on the F dis- (In most software packages, the exact prob-
tribution table (see Table 12.4). The F table abilities are provided for the researcher,
requires the use of two values for degrees of which allows the researcher to skip steps 9
freedom: the between variation df is placed and 10.)
nicant difference in identity awareness certain that there is a signicant change

scores of samples drawn from the three in identity awareness score.
adolescent age groups. 4. Box 12.7 presents a hand calculation of
We will follow the six steps in con- the test statistic.
ducting tests of signicance: 5. The results reported in Box 12.7 would
lead us to reject the null hypothesis be-
1. Research hypothesis: Identity aware- cause the difference observed would
ness scores will vary by age group. The occur by chance less than 5 percent of
null hypothesis is that there is no dif- the time. The critical value that must
ference in identity awareness scores be exceeded in order to reject the null
across age groups hypothesis is 3.07 (Table 12.4; note
2. If the null hypothesis is to be rejected, that we look up the value for 2 and 179
the result will fall into either tail of the degrees of freedom). Because the ob-
sampling distribution, this being a two- served F value is 10.39 and this ex-
tailed test. ceeds the critical value of 3.07, we re-
3. We will use the 0.05 signicance level in ject the null hypothesis.
the test, meaning that if the result falls 6. We conclude that there is a statistically
into the 2.5 percent right tail or the 2.5 signicant relationship between iden-
percent left tail, we can be 95 percent tify awareness scores and age of youth.
Table 12.4 Critical Values for the F Distribution: One-Tailed Tests, 0.05 Level of Significance*
df for the df for the Numerator
Denominator 1 2 3 4 5
2 18.51 19.00 19.16 19.25 19.30
3 10.13 9.55 9.28 9.12 9.01
4 7.71 6.94 6.59 6.39 6.26
5 6.61 5.79 5.41 5.19 5.05
6 5.99 5.14 4.76 4.53 4.39
7 5.59 4.74 4.35 4.12 3.97
8 5.32 4.46 4.07 3.84 3.69
9 5.12 4.26 3.86 3.63 3.48
10 4.96 4.10 3.71 3.48 3.33
11 4.84 3.98 3.59 3.36 3.20
12 4.75 3.88 3.49 3.26 3.11
13 4.67 3.80 3.41 3.18 3.02
14 4.60 3.74 3.34 3.11 2.96
15 4.54 3.68 3.29 3.06 2.90
16 4.49 3.63 3.24 3.01 2.85
17 4.45 3.59 3.20 2.96 2.81
18 4.41 3.55 3.16 2.93 2.77
19 4.38 3.52 3.13 2.90 2.74
20 4.35 3.49 3.10 2.87 2.71
25 4.42 3.38 2.99 2.76 2.60
30 4.17 3.32 2.92 2.69 2.53
40 4.08 3.23 2.84 2.61 2.45
60 4.00 3.15 2.76 2.52 2.37
120 3.92 3.07 2.68 2.45 2.29
3.84 3.00 2.60 2.37 2.21
signicance.
The differences in outcomes by age awareness scores. What the test is doing is
group are not likely to be caused by assessing the chance, if the true difference
random sampling uctuationswe can between the age groups in identity aware-
be 95 percent certain the identity ness is zero, of getting a sample uctuation
awareness differences observed reect of the magnitude shown in the sample
a real difference across the three age data. In the case under examination, the
group categories. differences range from a low of 27.35 in
early adolescence to a high of 34.35 in late
The test of signicance is based on the adolescence on the identity awareness in-
idea that, when a sample is drawn, there dex. The test of signicance reveals that
will be uctuations in the mean identity the observed uctuation is greater than a
uctuation that would probably be caused between the columns than within them; as
by normal sample variability. In other a result, the F ratio is more than one. After
words, if there were no real identity the F ratio is computed and the critical
awareness difference between the age value is looked up in Table 12.4, the deci-
groups, a sample would reveal a difference sion is made to reject the null hypothesis
7.0 points less than 5 percent of the time because the critical value is exceeded.
simply because of sampling uctuations.
On the other hand, if the observed differ-
ence could occur more than 5 percent of 2. Two-Way Analysis of Variance
the time, we would accept the null hypoth-
esis and conclude that there are no statis- Experimental researchers also use a two-
tically signicant differences in identity way analysis of variance, or a multifactor
awareness in the populations studied. analysis of variance, when their design
To repeatand to emphasize: A statis- has two or more independent variables.
tically signicant relationship is one in Theoretically, any number of indepen-
which the observed difference would oc- dent variables is possible, but in practice
cur by chance less than 5 percent of the rarely more than three or four are used
time. If the test were set at the 1 percent because of the large sample size that
level, it would simply be more difcult to would be required to run the analysis.
reject the null hypothesis; here, the dif- The inclusion of additional complexity is
ference would be statistically signicant relatively simple and provides an oppor-
only if it could occur by chance less than tunity to explore the interaction of treat-
1 percent of the time. ment variables (in which effects change
Similar to the computational proce- for different combinations of treatment
dure involved in determining the correla- variables) in inuencing the dependent
tion between two variables, the proce- variable. Note that the variations, which
dures for an analysis of variance require a are compared in an analysis of variance,
measurement of two kinds of variation: are composed of two elements: random
variations within a column (e.g., differ- error and possible treatment effects. The
ences within each age group), and varia- test that is done compares the ratio of the
tions between columns (e.g., differences differences in the following way:
that show up between early, middle, and Random error Possible treatment effects
late adolescence). An analysis of variance F
Random error
involves computing a ratio that compares
these two kinds of variabilitywithin-col- The interpretation of the results of an
umn and between-column variability. analysis of variance test examines:
On examining the data within column The main effect of treatment variable A
one (early adolescence data) of Box 12.7, The main effect of treatment variable B
note that age group cannot explain varia- An interaction of the treatment vari-
tions within this column (all cases are in ables (A B)
the early adolescence age group). Differ- An error term (within groups)
ences in identity awareness scores be-
tween the columns, however, may be ex- Box 12.8 presents the results of a two-
plained by the association with the way analysis of variance test. To inter-
independent variable (age group). pret such a table, one normally begins
Box 12.7 goes through the computation with an examination of the main effects; if
of a one-way analysis of variance for the there are signicant main effects, one
sample data. In this particular case, the es- would then examine whether there are
timates of variance indicate more variation signicant interaction effects.
BOX 12.8 Sample ANOVA Results
Suppose we had run an experiment with one form a two-way analysis of variance, with a test
dependent variable and two treatments and we for interaction.
wish to test for possible interactions between The output might be summarized in the fol-
the treatments. In such a case, we would per- lowing way:
Source of Variance Sums of Squares df Mean Square F Sig. of F

Main effect (A) 1449.393 2 724.697 10.421 0.000
Main effect (B) 12.652 1 12.652 0.182 0.607
Interaction (A B) 125.990 2 62.995 0.906 0.406
Error term-within group 12239.591 176 69.543
Total 13815.016 181 76.326
1. Understanding the Values in the Sum- if there is any interaction between the
mary Table treatments.
i) The degrees of freedom are calculated a. Main Effects
as follows: To decide if a treatment was statistically
significant, you will need to compare the F
Main effect (A): df 1 less than num- value to the values found in Table 12.4 us-
ber of levels in that factor. [In this
case, 3 1 2] ing the appropriate degrees of freedom. If
the value you obtained is larger than the
Main effect (B): df l less than num- one reported in Table 12.4, you have a sta-
ber of levels in that factor. [In this
case, 2 1 1] tistically significant effect. To use Table
12.4, you will need values for the required
Interaction effect (A B): df prod- degrees of freedom; use the smaller value
uct of the dfs making up the interac-
tion. [In this case, 2 1 2] across the top of the table and the larger
one on the vertical axis.
Error term: df total number of b. Interaction Effects
cases minus the product of number of
levels of A and B. [In this case, 182 If you have a significant main effect, you
(3 2) 176.] will then want to inspect your results for
any interactions between the effects. Your
Total df N 1 [In this case, 182 job is simple if there is no significant inter-
1 181]
ii) To compute the mean square for an ef- action effect; you simply report that none
fect, divide the sum of squares for the ef- was present. It is more complicated with
fect by its df. an interaction; in this case, the effect of a
iii) To compute the F value for each effect, treatment is not independent of the other
divide the mean square by the mean treatment; it may be that under the high
square error term-within groups. condition of treatment A, we find that
2. Interpreting the Results treatment B does not enhance the impact
In inspecting your results, you should ex- on the dependent variable.
amine two issues as you begin to interpret See Figures 18.1, 18.2, and 18.3 in Chapter 18
your results. First, you will want to know if for an illustration of how plots can be used to
your treatment variables had a significant show relationships with and without significant
effect; if either or both did, you should see interactions.
3. Repeated Measures Analysis of tion effect would be examined to determine

Variance if there was a differential treatment effect
on exercise frequency at different points in
In a number of nursing situations, a re- time. Repeated measures analysis of vari-
searcher may be interested in studying ance is a useful tool for the nurse re-
the effects of an intervention over several searcher given that nurse researchers of-
points in time or a researcher may exten record data for subjects at several
pose the same subject to two or more points in time.
treatment conditions when it is reason- Analysis of variance is a major analyti-
able to assume that one treatment cannot cal tool, used particularly by experimen-
contaminate a second treatment condi- tal researchers. Normally, computers will
tion. In such situations, repeated mea- make all such computations. You need
sures analysis of variance may be used not concern yourself with memorizing
to test for differences among the mean the calculation formula included in the
scores of the treatment conditions. boxes in this chapter. However, to fully
For example, if you are interested in understand what inferential statistics can
studying adherence to a regular exercise do for you, you need to be aware of what
regime after exposure to a 6-week work the statistical test is doing with your data
site health promotion physical activity to answer your research question or test
program for exployees, the researcher your hypotheses. The researcher will en-
would use a pre-postexperimental con- counter F distributions in using the
trol group design in which the subjects are MEANS, ANOVA, MANOVA, and REGRES-
randomly assigned to the experimental SION procedures in SPSS.
(health promotion program) and to the To perform this analysis using SPSS, see
control group (regular work site routine). the ANOVA procedure in Appendix A.
The researcher would collect data on
the frequency of exercise before initia-
tion of the health promotion program E. WHEN TESTS OF
(time 1), at the midpoint of the program STATISTICAL SIGNIFICANCE
(time 2), 2 weeks after completion of the ARE NOT APPROPRIATE
program (time 3), and 3 months after
completion of the program (time 4). An F At the beginning of this chapter, readers
statistic would be calculated to deter- were cautioned that although the ndings
mine differences in the mean scores of ex- of a study may be statistically signicant,
ercise frequency for both the experimen- they may not be clinically signicant. To
tal and control groups. This would be be clinically signicant, the ndings must
referred to as the between-subjects effect. have meaning for patient care in the pres-
This statistic would determine if, across ence or absence of statistical signicance.
all time periods, there was a difference in Whereas statistical signicance indicates
the frequency of exercise for the experi- that the ndings are unlikely to result from
mental and the control groups. A second F chance, clinical signicance requires the
statistic would be calculated to determine nurse to interpret the ndings in terms of
differences in exercise frequencies across their value to nursing. Therefore, statisti-
all time periods within each group. This cal signicance is not the only criterion of
would be known as the within-subjects ef- importance, and the practitioners would
fect. This statistic would determine if, do well not to be overly impressed by its
across both the experimental and control presence or absence. And sometimes
groups, the frequency of exercise differed tests of signicance are simply inappro-
at times 1, 2, 3, and 4. Finally, the interac- priate. Yet they are widely used, often im-
properly (Munro and Page, 2001; Polit, one sample out of a universe of possi-
1996). Why is this the case? ble samples that might be taken. The
Performing tests of signicance is, no problem with this argument is that not
doubt, often motivated by the desire to be all hospitals were given an equal
scienticor appear to be scientic. If chance to be included; second, if you
ones data turn out to be statistically sig- wish to make the argument that this is
nicant, then that is taken to demon- but one case among many, then you
strate the importance of the nding and to have shifted the unit of analysis to the
confer scientic legitimacy on ones work. institutional level (a hospital) and now
Such tests help to create the impression have one case represented by one dif-
that the God of sciencethat indepen- ference. Here you would be left with a
dent, unbiased arbiter of truthhas difference of $1500, but with no knowl-
blessed ones research with approval. A edge of what differences are present in
second motivation has to do with the fact other institutions.
that such tests are routinelyeven if in- Rule 2. Because tests of signicance only
appropriatelyreported in the literature. provide a measure of the probability of
Hence, to produce a report that meets the a given differences being the result of
standards of the discipline, tests of signif- sampling uctuations, assuming that
icance are expected. A third possibility is probability sampling procedures have
that some researchers poorly understand been used, such tests are not appro-
tests of signicance and these tests are in- priate if those methods are not used. If
appropriately used because of such mis- an experimenter asked people to vol-
understandings. For a variety of reasons, unteer to be in a treatment group and
many studies should not use tests of sig- then wished to compare these individ-
nicance. The following rules indicate uals with others who agreed to serve
when such tests are inappropriate. as controls, any test of signicance on
the differences between the groups
Rule 1. Tests of signicance are not ap- would be meaningless.
plicable when total populations are In experimental designs in which vol-
studied. If a study is being done on the unteers have been sought for participa-
wage differences between male and fe- tion (convenience sample) but in which
male nursing administrators in one in- subjects have been assigned to treat-
stitution and data relating to all the ment and control groups using random
nursing administrators are analyzed, sampling techniques, it is appropriate
then a test of signicance would not be to use tests of signicance to compare
in order. If a $1500 difference is ob- the treatment and control groups. How-
served after the appropriate controls ever, in such instances, the population
are introduced, this difference is ab- should be regarded as those who vol-
solute. The researcher must decide unteered for participation.
whether the difference is to be charac- Rule 3. Tests of signicance are to be con-
terized as substantial, modest, or triv- sidered suspect when there is a sub-
ial. To say that it is statistically signi- stantial nonparticipation rate. If sub-
cant is simply incorrect! Remember, stantial numbers of respondents have
tests of signicance assume that sam- refused to participate (for arguments
pling procedures have been followed sake, lets say 40 percent), then tests of
they are not applicable when total signicance become problematic be-
populations have been studied. cause it is difcult to assume that non-
Arguments have been advanced that participants are similar to those who
a study like the one above represents agree to participate in the study. There-
fore, the extent to which such data can collection of data for the study. Re-
be considered representative can be searchers routinely collect data on
called into question. many variables; it is not unusual in a
Rule 4. In nonexperimental research, if one survey, for example, to collect data on
is exploring a relationship that has been 100 or more variables. If we run every
shown to be statistically signicant, it is variable against every other one, we will
inappropriate to use a test of signi- not only generate enormous piles of
cance once again when controls have computer output, but we will also gen-
been applied to test for an intervening erate results for many hundreds of theo-
variable. The issue in evaluating models retically meaningless relationships. We
is to assess the impact of the control on will also encounter many statistically
the original relationship and to observe signicant relationships. Indeed, one
whether the differences have remained would expect that on a simple chance
the same or if they have increased, de- basis, one in 20 tables will prove to be
creased, or disappeared. Frequently, statistically signicant, providing one
they may change from being statisti- is using the 0.05 signicance level for
cally signicant to being statistically in- ones tests. Unfortunately, researchers
signicant, but this change may simply rarely report how many relationships
represent the fact that the data have were analyzed, what confounding data
been partitioned in order to do the have been discarded, and which causal
analysis (perhaps cut in half ). There- models were not fully developed before
fore, the switch from signicance to data collection.
nonsignicance may simply reect the An exception to the above rule has
fact that fewer cases are being used. to be made for those analyzing sec-
Researchers should note how much ondary data. In such cases, it is impos-
the original relationship has shifted, sible to formulate hypotheses before
not simply note whether the relation- data collection; nonetheless, hypothe-
ship is still statistically signicant. To ses should be formulated before data
avoid confusion, do not use tests of analysis. Moreover, it should be noted
signicance when causal models are that some researchers nd it useful to
being evaluated beyond the stage of run tests of signicance, even in inap-
the initial relationship. For example, if propriate situations, to help shed light
the relationship between X and Y is be- on the relationships explored even
ing explored and if sampling and other though they do not report these tests
conditions are appropriate, then a test in their articles.
of signicance for the relationship be- Researchers must use tests of signif-
tween X and Y is legitimate. However, icance only with great care. Above all,
having established that there is a sta- beginning researchers must appreciate
tistically signicant relationship be- what such tests really measure. All too
tween X and Y, a further test of signi- often it is not recognized that a nding
cance would not be appropriate when may be equally important whether the
one is testing to ensure that the rela- relationship is statistically signicant
tionship is not spurious because of its or not. Nurse researchers are in the
connection to some third variable. business of trying to understand and
(This point is discussed in detail in describe phenomena of concern to
Chapter 17.) nursing; therefore, to discover that
Rule 5. Tests of signicance may not be there is no relationship between A and
applied to relationships that were not B is as important as it is to discover
formulated as hypotheses before the that there is a relation between A and
B. Futhermore, if one is testing a the- do not despair if your study does not
ory, it is especially important to nd yield statistically signicant results.
that a predicted relationship does not Science proceeds through disconrma-
hold true because this will cast doubt tion, ruling out alternatives, rejecting,
on the theory and, perhaps, lead to a modifying, and continually rethinking
renement or refutation of it. In short, theoretical formulations.
E X E R C I S E S
1. Test the null hypothesis that there is heparin lock in place, without changing
no difference in the rate of complica- it, for 48 hours versus 96 hours. Test
tions for clients with AIDS who are re- the hypothesis at the 0.05 level. The
ceiving intravenous antibiotics with a data are shown in the following table.
Rate of Complications in Patients with AIDS by Heparin Lock Replacement Time
Heparin Replacement Time Group
Complication 48 hours, 96 hours, Total,
rate N% N% N%
Had complications 31 47.0 21 34.4 52 40.9
No complications 35 53.0 40 65.5 75 59.1
Total 66 100.0 61 99.9 127 100.0
2. You have examined the salaries of all tistically signicant difference between
the nursing staff working in a large the salaries of men and women? (Cau-
teaching hospital. What is the appro- tion: This is a trick question!) The data
priate test to determine if there is a sta- are shown in the following table.
Income List by Gender, Large Teaching Hospital
Men, dollars Women, dollars
46,200 46,600
47,300 46,800
38,800 49,100
50,200 48,700
49,600 37,700
48,800 48,600
47,900 57,700
51,100 50,400
Cobb, G.W. (1999). Introduction to Design and research examples that explains analysis
Analysis of Experiments. New York: Springer. of variance well.
This is a ne treatment of the application of Munro, B. (2001). Statistical Methods for Health
analysis of variance in experimental de- Care Research (4th ed). Philadelphia, PA:
signs. J.B. Lippincott. This book provides a com-
Howell, D.C. (1997). Statistical Methods for Psy- prehensive overview of the statistical tech-
chology (4th ed). Belmont, CA: Duxbury niques most frequently encountered in the
Press, Wadsworth Publishing Co. This text health professional research literature.
presents a comprehensive discussion of Computer printouts help demonstrate spe-
the use of inferential statistics in experi- cic techniques.
mental design. Polit, D. (1996). Data Analysis and Statistics for
Mitchell, M., and Jolley J. (1997). Research De- Nursing Research. Stamford, CT: Appleton
sign Explained (3rd ed). Orlando: Harcourt & Lange. This very practical text examines
Brace Jovanovitch College Publishers. This the whys, whens, and hows of statistical
is an exceptionally ne text with excellent analysis.
Part 5
ISSUES IN
DATA
COLLECTION
T
he chapters in Part 5 deal with thinking about measurement
inssues (Chapter 13), designing questionnaires (Chapter 14),
and how to go about sampling (Chapter 15). How relevant
each of these chapters is will depend upon the type of re-
search design that you are working with; if you are doing a content
analysis, for example, you will be able to skip Chapter 14. If you are
doing a hospital-based participant observation study, Chapters 14
and 15 will not be of much relevance to your present study although
the information may be of interest to you for future research designs.
421

Chapter 13
Data Collection and

Measurement
CHAPTER OUTLINE
A. Levels of a Research Project F. Procedures for Data Collection

B. Linkages Between Levels 1. Indexes
1. Validity 2. Likert-Based Indexes
2. Reliability 3. Semantic Differential Procedures
C. Measurement Error 4. Magnitude Estimation Procedures
D. Levels of Measurement 5. Visual Analogue Scales
6. Delphi Technique
1. Nominal Measurement
7. Physiological Measures
2. Ordinal Measurement
8. Observational Measurement
3. Ratio Measurement
9. Interviews
E. The Effects of Reduced Levels of 10. Questionnaires
Measurement
G. Development of an Instrument
KEY TERMS
Concept Internal consistency Questionnaire

Conceptual hypothesis Internal validity Ratio measurement
Concurrent validity Kuder-Richardson coefficient Reliability
Construct Likert-based index Response set
Construct validity Magnitude estimation Scale
Content validity procedures Semantic differential
Credibility Measurement Split-half method
Criterion validity Measurement error Stability
Cronbachs alpha Nominal measurement Summated rating index
Data collection Observational measurement Theoretical level
Elite bias Operationalization Theoretical substruction
Equivalence Operational level True value
External validity Ordinal measurement Validity
Face validity Pilot study Visual analogue scale
Index Predictive validity
423
424 ISSUES IN DATA COLLECTION
Y
ou have completed the steps out- levels of aggression in psychiatric clients
lined in Chapter 3 to get your proj- who are frustrated as part of an experi-
ect started. You have identied ment, to classifying television soap op-
the research problem, reviewed the eras according to whether they stereo-
existing literature, selected a theoretical type health professionals by gender, to
framework to guide the study, and articu- having individuals complete question-
lated the research problem as an inter- naires or undergo personal interviews.
rogative question, a declarative state- We measure things because we wish to
ment of purpose, or a hypothesis. Now accurately describe objects. We wish, for
you are ready to translate the research example, to know how popular a new
concepts in your problem statement into smoking cessation program is now com-
observable and measurable phenomena. pared with when it was originally intro-
This will enable you to measure the vari- duced; to describe the programs popular-
ables of interest in the study. Measure- ity in various parts of the country; to see if
ment refers to the process of assigning the programs popularity varies with the
values to observations so as to reect gender of the person doing the rating, the
variables. Then you can proceed to data length of time the person smoked, or the
collection, the process of gathering infor- number of smoking cessation programs at-
mation from identied participants to an- tended previously by the participant; or to
swer the research question. Because the understand what factors explain gender
quality of a study is often determined by differences in the program ratings. If we
the data collection and measurement pose such questions, we will need to mea-
methods selected, this chapter focuses sure the appropriate variables and analyze
on these two vital steps. General mea- them. We could guess about the answer,
surement principles and concepts are but we would not have much condence in
discussed; these include the levels of a our answer unless the question is tackled
research project, validity and reliability seriously and the data to answer the ques-
issues, measurement levels, measure- tion are collected objectively and system-
ment error, and steps that can be taken to atically. Objectively implies the individual
reduce measurement error. In addition, who collects it does not inuence the data
various means available to nurse rein any way. Systematically refers to a con-
searchers for collecting data and measur- sistent method for collecting the data so
ing nursing phenomena are explored and that everyone involved follows the same
the process of instrument development procedures for data collection (LoBiondo-
discussed. Wood and Haber, 1998).
Measurement is the process of linking
abstract concepts to their empirical indi-
cators. Typically, researchers work from
A. LEVELS OF A RESEARCH the general to the specic: for each gen-
PROJECT eral concept, an indicator is identied. To
look for a way of reecting the idea of so-
Most methods of data collection involve cial support is to ask a measurement ques-
an attempt to measure variables. Mea- tion. How might we best measure, or indi-
surement takes many forms: from unob- cate, a persons level of social support?
trusively observing nursing clients and We refer to concepts we intend to mea-
recording which of them use the recre- sure as variables.
ation facilities available in the hospital Figure 13.1 presents the levels of a re-
sunroom, to getting information on sui- search project. To move down the gure is
cides from public statistics, to observing to move from the general to the specic
DATA COLLECTION AND MEASUREMENT 425
Proposition: >A >B THEORETICAL LEVEL

>B >C
Axiomatic derivation
>A >C
Replacement of terms
And given: A = A1 = A2
Derived hypothesis
C = C1
Derivation: >A1 >C1
>A1 >C1 CONCEPTUAL LEVEL

Concept Concept
>A1 >C1 OPERATIONAL LEVEL

Measure Measure
Figure 13.1 Levels in research design.
from the theoretical level to the opera- tual variables or constructs. An example
tional level. At the theoretical level there would be: the greater the integration into
are a number of interconnected proposi- campus life, the greater the students so-
tions, assumptions, and statements of cial support during their college years.
relationship between concepts. By em- Having stated the conceptual hypothesis,
ploying axiomatic derivations and replace- the researcher would proceed to the oper-
ment of terms (see Chapter 2), it is possi- ational level, deciding how each of the con-
ble to derive conceptual hypotheses. Such structs (integration into campus life and
hypotheses may also be identied by re- social support) would be measured, and
viewing the research literature, or by an in- what procedures would be used to collect
sightful analysis of the problem. A con- and analyze the information.
ceptual hypothesis is a statement of the Theoretical substruction is the dyna-
relationship between two or more concep- mic thinking process that the researcher
engages in to move from the theoretical The suggested theory states that identity
level of a quantitative study to the opera- is predicted by the strength of maternal at-
tional or measurement level (Wolf and tachment and family coping ability. At the
Heinzer, 1999). Through substruction, the next level, constructs reective of the con-
researcher identies the foundational ele- cepts are selected. For example, the con-
ments of a study, determines the relation- cept of maternal attachment is reected in
ships among the elements, and depicts the construct relationship to mother; the
this in a diagram (Dulock and Holzemar, concept of family coping is reected in
1991). This process brings clarity to the coping with normative life events and cop-
research process and enables the re- ing with nonnormative life events; and the
searcher to better understand the rela- concept of identity is reected in three
tionship among the theoretical, concep- constructs: self-worth, life satisfaction,
tual, and operational level of the study. It and general affect. The nal step involves
also illustrates the hierarchical order selecting an instrument or measurement
among the major constituents of the study. device to operationalize the constructs.
Figure 13.2 presents a substruction that il- Operationalization refers to translating
lustrates the connections from the re- the constructs of interest into observable
search question through to the measure- and measurable phenomena. In our exam-
ment and data analysis process. In Figure ple, we see that relationship to mother is
13.2, the theoretical level demonstrates measured by the Maternal Scale of the In-
the major concepts of a proposed theory ventory of Parent and Peer Attachment
on adolescent identity in cancer survivors (IPPA) instrument, coping with normative
and the relationships among the concepts. and nonnormative life events is measured
Predictor Outcome
CONCEPT
Maternal Family Identity
attachment coping
Self worth General

Relation Coping Coping with
affect
CONSTRUCT ship to with non-
mother normative normative
life events life events
Life satisfaction
VARIABLE IPPA* Family Rosenberg Index of

Maternal Coping Self-esteem Well-Being
Scale Inventory Scale
Figure 13.2 Theoretical substructure: Identity in adolescent cancer survivors after childhood
diagnosis and its relationship to maternal attachment and family coping. A possible model.
by the Family Coping Inventory (FCI), so- world. Making a concept potentially ob-
cial worth is measured by the Rosenberg servable facilitates testing of it. Some au-
Self-Esteem Scale, and general affect and thors distinguish construct from con-
life satisfaction are measured by the Index cept by stating that constructs are
of Well-Being. The selection of the data col- constructed or invented by the re-
lection instruments ultimately directs the searcher. A construct refers to an ob-
choice of analytical methods. servable characteristic of an individual,
In some qualitative approaches, the group, or nation. Constructs help us to
process is reversed and, after doing some organize our thinking about the world.
observations, the researcher may ask, Nurse scientists use constructs such as
Well, what concept does this reect? In health status, job satisfaction, compli-
this case, the researcher begins with ob- ance, and self-care ability. There are
servations and then tries to link them to hundreds of such constructs. It is im-
more general ideas. This is the approach portant that the researcher define pre-
recommended by the grounded theory cisely what is meant by each construct
method (Strauss and Corbin, 1997). This used. Precision helps to make clear what
approach involves a process of discovery is included in the idea and also provides
in which the researcher begins with obser- a guide as to how it should be measured.
vations and tries to make sense of them, to The operational level of research refers
identify the concepts the data seem to re- to the indicators used to reflect the con-
ect. As observations are made, the constructs as well as to the procedures used
cepts are continuously identied and re- to collect and analyze data.
ned. Concepts become the basic units of The review of the literature is helpful in
analysis: showing how other researchers conceptu-
ally dene and measure variables of inter-
In grounded theory, representativeness
est to the researcher. After variables are
of concepts, not of persons, is crucial.
conceptually and operationally dened,
The aim is ultimately to build a theo-
the researcher searches the literature
retical explanation by specifying phe-
again to identify instruments that may be
nomena in terms of conditions that give
used as is or modied for use in the study.
rise to them, how they are expressed
When instruments are available, the re-
through action/interaction, the conse-
searcher must rst seek permission from
quences that result from them, and vari-
the author before implementing their use
ations of these qualiers . . . For in-
in a study (LoBionda-Wood, 1998).
stance, one might want to know how
representative comfort work is of the
total amount of work that nurses do . . .
Do nurses engage in it all of the time or B. LINKAGES BETWEEN LEVELS
some of the time? What are the condi-
tions that enable them to do it or pre- Measurement, in essence, refers to the
vent their doing it? (Corbin and Strauss, linkage between the conceptual and the
1990, p. 9). operational levels of a research project.
There are two key issues in this linkage:
Measurement refers to the process by
validity and reliability.
which categories or numbers are used to
reflect or indicate concepts and con-
structs. A concept is a general idea not 1. Validity
directly observable in the real world. A
construct is a concept specified in such In Chapter 1, validity was dened as the
a way that it is observable in the real extent to which a measure reects a con-
cept, reecting neither more nor less than concurrent validity would be demon-
what is implied by the conceptual deni- strated if two different tests were given
tion. Validity has to do, then, with the con- and the scores correlated highly with
gruence of concept and indicator. When each other. If a theoretically derived hy-
an instrument is considered valid, it mea- pothesis turns out as predicted, it would
sures the concept it was intended to mea- constitute one piece of evidence for the
sure. There are ve types of validity: face, validity of the measures. This latter type
content, concurrent, predictive, and con- of validity is known as construct validity.
struct validity. An instrument may be Another way to think about construct va-
judged to possess one, all, or any combi- lidity is to recognize that it is based on in-
nation of these types of validity. Table ductive evidence. If one nds evidence to
13.1 describes the various types of valid- support a theoretically derived hypothe-
ity, describes how one would test for sis, then it would indicate that ones mea-
them, and provides examples of studies sures have construct validity.
that used tests for validity. For a full dis- Experimentalists, in particular, distin-
cussion of each test, interested readers guish between internal and external va-
should consult the reference citation. Un- lidity (see Chapter 4). Internal validity
like reliability, there are no simple statis- may be taken to mean that the researcher
tical tests to assess validity, yet beginning has demonstrated that the treatment, in
researchers should look for evidence that fact, produced the changes in the depen-
the data collection instrument is a valid dent variable. External validity has to do
measure of the research phenomenon. with the extent to which results may be
Beginning nurse researchers should, extrapolated from the particular study to
at minimum, be aware of the types of other groups in general.
validity outlined in Table 13.1 and con- When the concept to be measured is a
vinced that selected measures have face concrete one such as age or body weight,
validity. A measure has face validity if, on the issue of measurement is straightfor-
inspection, it appears to reect the con- ward. Many times, however, direct mea-
struct that you wish to measure. Content sures of concepts relevant to nursing
validity refers to the extent to which a research are not available. Often nurses
measure reects the dimension (or di- are interested in exploring abstract ideas
mensions) implied by the concept. Typi- such as health, lifestyle, self-esteem, social
cally, content validity requires a more status, coping, stress, fatigue, pain, or re-
formal assessment, including the use of a silience. Rarely are direct measures avail-
panel of experts, corroboration of the able for such concepts. This requires the
measurement with the concepts dimen- researcher to use indirect measures that
sions noted in the literature review, and are measures of the characteristics or at-
with the studys theoretical framework in tributes of the abstract concept. For exam-
order to assure that the content of the ple, if you are interested in measuring
measures reect the framework. There lifestyle patterns, indicators may include
are two types of criterion validity: con- nutrition practices, ability to deal with
current and predictive. If you were at- problems, stress management techniques,
tempting to develop a measure to predict rest and relaxation patterns, exercise lev-
success in nurse registration examina- els, attitudes, decision-making styles, so-
tions, you could assess the validity of cial support systems, coping patterns, and
your measure by correlating it with suc- self-actualization behaviors. It is a chal-
cess rates on registered nurse examina- lenge to get a valid measure of these
tions. A high correlation would indicate aspects through one measurement ap-
high predictive validity. In this example, proach. The researcher would rst need to
Table 13.1 Types of Validity and Their Tests

Type Definition Test For Validity Example
Face validity Upon inspection or Expert opinion. Use a Gillis, 1997
face value, the instru- panel of experts to
ment appears to be a determine if the tool
good indicator of the measures what it is
concept it is intended intended to measure.
to measure.
Content validity The instrument items Veried by other Ewen, 1993
reect the full range evidence such as
of the attributes of the literature review to
concept being determine what con-
measured. tent should be
included; opinion of
experts on scope and
representativeness of
content; use of studys
theoretical framework
to support the content
validity of the tool.
Criterion validity
1. Concurrent validity The correlation of one Test instrument with Norbeck et al., 1982
measure with another another instrument
measure of the same that measures the
phenomenon. same concept and is
known to be valid.
2. Predictive The instrument can Tested by using the Norbeck et al., 1982
accurately predict a instrument in a study
phenomenon. For and then comparing
example, GPAs can the results with a
predict NCLEX success future outcome.
rate or risk behavior
scales are often used
to predict the proba-
bility of risk-taking
acts in adolescents
and the agreement of
the scales with subse-
quent risk behaviors
gives a measure of
their predictive validity.
Construct validity The measure distin- Establishing construct
guishes participants validity is a complex
who differ on the process involving
construct being several approaches.
measured by the These include:
instrument. Hypothesis testing
approach
Convergent and Norbeck et al., 1982
divergent approaches
Contrasted group Bunn & OConnor,
approach 1996
Multitrait-multi-
method approach
Factor analysis Gillis, 1997
approach
Causal modeling Carruth, 1996
approach
clearly dene the conceptual denition of community with dedication for many
lifestyle so it is obvious what aspects of years has inherited her husbands sub-
lifestyle the research is dealing with and stantial estate. Although she has no
then operationalize the denition by se- salary, she probably has more prestige
lecting indicators congruent with the con- than anyone else in the community.
ceptual denition. Does the researcher
The problem here is that salary does
wish to dene lifestyle in terms of actual
not always capture the concept of social
health behaviors engaged in or attitudes
prestige adequately, and even if you could
and values toward health behaviors or a
get a measure of annual income rather
combination of behaviors and attitudes?
than salary, you would still not have a
The instrument or tool selected to mea-
valid measure of the concept as dened.
sure lifestyle would vary depending on the
(Think of your own home community. Can
researchers denition of lifestyle.
you think of cases in which annual income
might not reect prestige?)
a. The Idea of Validity in Quantitative When measuring socioeconomic sta-
Research tus, many nurse researchers rely on oc-
cupational rating scales that have been
To illustrate, let us examine one possible based on either:
conceptual denition of socioeconomic
status. If socioeconomic status is dened The four-factor index of social status
as a hierarchical continuum of respect proposed by Hollingshead (1975) that
and prestige and we then choose to op- includes education, occupation, gen-
erationalize the concept by measuring der, and marital status. This index pro-
the annual salary of each individual we vides a measure of social status reec-
study, we will most certainly have prob- tive of the social system of the United
lems convincing others of the validity of States.
our measure. The average levels of income and edu-
On inspection, it becomes clear that cation for people in any given occupa-
annual salary would not adequately re- tion in Canada (Blishen and McRoberts,
ect an individuals place on a continuum 1976).
of respect and prestige. Such inspections The subjective assessment of the pres-
are generally referred to as face validity. tige of different occupations (see Pi-
The following are reasons why you might neo and Porter, 1967.
doubt the validity of the proposed mea-
All of the above approaches lead to
sures:
quite similar rankings of different occu-
A local drug dealer is making a fortune pations. A major advantage of using such
but enjoys little respect in the commu- indexes is that it is quite easy to nd out
nity. a persons occupation and then simply
The best poker player in town always use the score supplied in one of the
seems to have lots of money, takes ex- above indexes to reect the socioeco-
pensive trips, and many people admire nomic status of the individual.
his worldliness even though his salary However, one of the problems of using
at the local pub is little above the min- prestige indexes is that such indexes may
imum wage. not reect the complexity of the organiza-
The Protestant minister is looked up to tion of the modern household. For exam-
by almost everyone even though his ple, such indexes traditionally did not
salary is a pittance. produce a prestige score for housewife
The widow of a doctor who served the (researchers would assign a missing value
code or a separate special code to house- may wish to contact prestigious commu-
wives), and researchers simply tended to nity members in the future to support a
use the fathers occupational prestige as a health promotion project for their com-
measure of the familys socioeconomic munity. The local drug dealer or poker
level. Given the participation of women in player who may receive high scores on
the labor force, this is no longer a satis- some indicators of socioeconomic status
factory solution for gauging household so- would not bode well as respectable and
cioeconomic status. To avoid the inherent prestigious exemplars of the new project.
sexism of using the fathers occupational An option would be for the nurse re-
prestige rating as a measure of a house- searcher to contact a panel of local infor-
holds socioeconomic status, in recent mants who could estimate the relative
years our students have been basing rat- prestige of known members of the com-
ings of a households socioeconomic sta- munity.
tus on the higher of the two occupational When choosing indicators, one should,
prestige scores in households in which ideally, not select causes, consequences,
parents are living together and both are or correlates of the variable but, instead,
employed outside the household. When get a direct measure of the phenomenon
only one of the parents is employed out- itself. As annual income may be one of
side the household, the occupational several important causes of variation in
prestige of that person is used. Simply prestige, one must not simply measure
adding the prestige scores together would prestige by income. To do so would be to
result in a prestige score that would not be choose a correlate as ones only measure.
in line with public perceptions. (For exam- Similarly, to measure the prestige of the
ple, the combined prestige scores of par- previously described widow by the suc-
ents who are both teachers would be cess of her children would be to look at a
about one and a half times that of a single- consequence of her position and may not
earner physician household.) It would be adequately reect her prestige itself.
possible to add the values together but Again, the resulting measure may be cor-
weight the lower of the two at a reduced related with her prestige but not reect it
level, perhaps using only one quarter of its adequately.
value or add them together and divide by Researchers must ensure that mea-
two, as suggested by the Hollingshead in- sures developed to reect an indepen-
dex. These calculations are easily done in dent variable are indeed independent of
the Statistical Package for the Social Sci- those used to reect the dependent vari-
ences (SPSS), but they are a little tricky able. If care is not taken, a researcher may
(because of the housewife problem). An end up with two different measures of the
example is shown in Appendix A. same phenomenon and may inappropri-
To perform this analysis using SPSS, ately conclude that there is a strong
see the Index Construction section of Ap- causal relationship between the two vari-
pendix A. ables; instead, there may only be a corre-
Clearly, however, a denition of socio- lation between two measures of the same
economic status that stresses the relative underlying phenomenon. For example,
respect and prestige of members of a com- suppose we measured the weight of hos-
munity begs for a measure that would get pitalized children in pounds; suppose we
at the extent to which different individuals then weighed the children again, this time
in the community are looked up to. In a in kilograms. We would then have two
community health nursing study, it may be measures, highly correlated, that reect
important to get a valid measure of this as- the same underlying reality. But we would
pect of socioeconomic status because you err if we concluded that the weight in
pounds causes weight in kilograms. These ences weight in kilograms. But in nursing
measures lack independencethey both research, things are not always so easy.
reect the same underlying variable (the Suppose that you were measuring fac-
weights of the children). tors that inuence healthy lifestyles in
The example is a simple one, and few adolescents and that you included the
researchers would be in danger of con- following among the measures of healthy
cluding that variation in pounds inu- lifestyles:
Independent Variables (Causes of Dependent Variable
Variation in Healthy Lifestyle) (Healthy Lifestyle)
Concept of health Social support
Perceived health status Nutrition practices
Family support Physical participation activities
Peer support Stress management
Safety practices
Clearly there is some overlap between of research. In fact, much research does
measures of the independent and depen- not follow such a simple path.
dent variables; note that social support is Before researchers offer conceptual
one of the measures used for dependent denitions of variables, they often have
variables and that family and peer support given thought to how they will measure
are two of the measures used for indepen- the variable. One way to increase the ap-
dent variables. You would need to review parent validity of a measure is to select a
these indicators to ensure that they are conceptual denition that is most congru-
measuring different variables and not sim- ent with the proposed measure. To go
ply reecting the same underlying reality. back to the earlier example of measuring
If you are not careful, you may inadver- socioeconomic status, socioeconomic sta-
tently inate the relationship and come to tus was dened as a hierarchical contin-
an inappropriate conclusion. Thus, we can uum of respect and prestige. If the design
formulate the following rule: Make cer- of the study was to be a door-to-door ques-
tain that the indicators of your inde- tionnaire in a community, then the re-
pendent and dependent variables do searcher might well reconsider the con-
not overlap. This means that researchers ceptual denition of socioeconomic status
must be careful to ensure that the strength because it would be difcult to measure it
of the association between the indepen- easily in this way using a questionnaire. It
dent and dependent variable is not in- is easier to ask about educational level, oc-
ated by failing to use measures that re- cupation, or even income. Knowing this,
ect different concepts. the researcher might decide that a deni-
Another issue, not typically discussed tion of socioeconomic status along the
in nursing research texts, is the problem lines of differential access to scarce re-
of inating the apparent validity of a mea- sources would be more appropriate; now
sure by choosing a conceptual denition the researcher can achieve greater face
after considering what the most conve- validity by using some combination of ed-
nient measure would be. What does this ucation, occupation, or income to reect
mean? In the discussion so far, we have socioeconomic status.
assumed that the researcher is working Ultimately, in quantitative research,
down from the theoretical to the concep- analysis proceeds by examining relation-
tual and, nally, to the operational level ships between indicators. And one must
be careful when evaluating such research that there are dangers in thinking that in-
because, even though two studies may formation previously collected can be sub-
examine socioeconomic status, the mea- sequently simply checked, corroborated,
sures used to determine socioeconomic or corrected. This may be valid for only
status may differ enormously. Is this a certain cases. Participants may, for exam-
shell game? To the extent that conceptual ple, nd that their recollections of the mo-
denitions are selected to enhance valid- ment when the original data were col-
ity, one could argue that something of a lected cannot be decontextualized to
slight of hand is going on. Thus, even with constitute a test of validity. Attempts to do
clearly dened constructs and an indica- this may cause as many problems as they
tion of the measures used, the research solve.
consumer must always be alert to tau- Sandelowski cautions about the dan-
tologies lurking beneath the surface of a ger of making a fetish of rigor. She warns
research report. that trying to model the quantitative stan-
dards of rigor, takes us too far from the
b. The Idea of Validity in Qualitative
. . . artfulness, versatility, and sensitivity
Research
to meaning and context that mark qual-
In qualitative research, the issue of exter- itative works of distinction. It is as if, in
nal validity may need to be thought about our quasi-militaristic zeal to neutralize
in a slightly different way. Given the small bias and to defend our projects against
number of cases typically studied in qual- threats to validity, we were more pre-
itative projects, the issue of validity is occupied with building fortications
perhaps better thought about in terms of against attack than with creating the
credibility. As Sandelowski (1986, 1993) evocative, true-to-life, and meaningful
has argued, a qualitative study is credible portraits, stories, and landscapes of hu-
when it presents accurate descriptions man experiences that constitute the best
or interpretations of a meaningful human test of rigor in qualitative work (Sande-
experience that the people having that lowski, 1993, p. 1).
experience would immediately recognize
In terms of external validity, the very
as reecting their own experiences. A
act of controlling so many extraneous
study is also credible when other re-
factors (as in an experiment) actually
searchers or readers can recognize the
serves to reduce the generalizability of
experience when confronted with it after
such studies; in qualitative studies, such
having only read about it in a study.
articiality is reduced given that studies
Traditionally in qualitative studies, the
are done in natural settings even though
degree to which a description rings true
the limited sample sizes mitigate against
to the participants of the study, other read-
extrapolations to other populations.
ers, or other researchers has been taken as
In distinguishing qualitative from
an indication of whether the researcher
quantitative research traditions, Sande-
has, in fact, measured what you wished to
lowski notes:
measure. Furthermore, it has been argued
that the closeness of the researcher and . . . the artistic approach to qualitative
the participant being studied should be en- inquiry emphasizes the irreplicability
couraged rather than discouraged. Only of the research process and product.
through such closeness can the re- Every human experience is viewed as
searcher truly penetrate and understand unique, and truth is viewed as relative.
the experiences of the participant. More The artistic integrity, rather than the sci-
recently, Sandelowski (1993) has argued entic objectivity, of the research is
achieved when the researcher commu- record how the respondent and the re-
nicates the richness and diversity of hu- searcher have mutually inuenced one
man experience in an engaging and another. If this is done, the report will
even poetic manner . . . qualitative recognize the reciprocal inuence of
methods such as historical inquiry may the respondent on the researcher.
employ the methods of science but the
presentation or reporting style of art The issue of generalizability (external
(Sandelowski, 1986, p. 29). validity) poses difculties in many re-
search designs, but it presents special
In order to maximize rigor in qualita-
challenges in ethnographic studies. Con-
tive research, several suggestions are rel-
ventional sampling procedures are usually
evant:
not relevant to ethnographic studies be-
Keep careful records. Researchers cause such studies depend so extensively
should keep detailed records of all de- on case studies. However, it may be possi-
cisions that have been made and how ble to argue that the circumstances stud-
they were made. This establishes an ied in a particular ethnographic case study
audit trail. Details on how the subject are comparable to other situations and
matter was selected, how data were that the conclusions of the study may
collected, what evidence was deemed therefore be applied to them. One solution
unimportant, and the ways in which is to carry out research on numerous sites.
categories were developed should all However, there are four factors that may
be noted. The objective here is to illus- inuence the credibility of such cross-
trate as clearly as possible the regroup comparisons:
searchers thought processes that led
to the conclusions of the study. 1. Selection effects. The researcher may
Avoid the holistic fallacy. This fallacy select sites in which some factors may
would make the results of the study not be present. A study based on such
look more patterned than they actu- a site may, therefore, not allow testing
ally are; the researcher should attempt of certain ideas.
to establish how typical the observa- 2. Setting effects. Studying a nursing situ-
tions of the study are. Be careful not to ation may itself inuence the results de-
report only events and behaviors that rived. The impact of the researchers in-
are patterned and consistent; report trusion may vary from setting to setting,
the exceptions as well. distorting the results more in some set-
Guard against an elite bias. Elite bias tings than in others. Hence, it is difcult
is a danger because informants are to compare the degree to which the re-
more likely to be drawn from the more searcher inuenced the results of differ-
articulate, high-status elements in a so- ent studies conducted at different sites.
ciety. Hence, unless care is taken, there 3. History effects. Each group studied is
is a tendency to overrepresent the subject to unique historical inuences.
views of the elite in ones research. When sites are studied at different
Be wary of being taken over by the re- times, history rather than the interac-
spondent. If the researcher identies tion of factors within the site may ex-
completely with the views of a respon- plain some of the variations between
dent, it may be difcult to maintain the sites.
a clear distinction between the re- 4. Construct effects. Concepts may be re-
searchers experiences and those of garded differently by both observers
the respondent. This may be a prob- in different settings and those being
lem unless the researcher attempts to observed.
2. Reliability liability coefcients of 0.70 or above for

new instruments and 0.80 or above for es-
Reliability was dened in Chapter 1 as tablished instruments are acceptable.
the extent to which, on repeated mea- Estimates of reliability need to be de-
sures, an instrument yields similar re- termined each time the instrument is used
sults. Terms such as dependability, con- with a different population or sample. Re-
sistency, stability, and accuracy are often liability estimates are population depen-
used interchangeably to refer to reliabil- dent, so one must never assume that an
ity. The rst three terms refer to the in- established instrument will yield the same
struments ability to produce the same high levels of reliability in a diverse popu-
results on repeated measures. For exam- lation. For example, an instrument de-
ple, if a scale were used to weigh a newly signed to measure healthy lifestyle in rural
admitted hospital client, one would ex- adolescents may yield reliable results
pect the scale to yield the same value if when administered to rural adolescents,
three consecutive readings were taken of but low correlations may result when the
that individuals weight. The fourth term, same instrument is administered to a
accuracy, reects a different dimension of group of urban adults. It is important to al-
reliability: the measurement tools ability ways test the reliability of instruments
to reect the true value being measured. used in a study before conducting other
True value is the value that would be statistical tests of the data.
obtained if the measurement tool were
perfect. In other words, the true value is
free of random measurement error. Few, if a. The Idea of Reliability in
any, instruments or measurement tools Quantitative Research
are totally free of measurement error. In
There are both simple and complex ways
summary, one would expect a reliable
of assessing reliability in quantitative in-
measurement tool to yield similar results
vestigations, yet no single correct ap-
on subsequent administrations of the in-
proach exists. The three attributes of re-
strument to measure a variable that is rel-
liability should be considered in the
atively constant. Reliability issues emerge
assessment of an instrument. These in-
in both single indicator questions and in
clude stability, internal consistency, and
those in which a number of indicators are
equivalence. Table 13.2 outlines the vari-
used to reect a variable.
ous types, tests for, and examples of re-
Reliability is a relative term. It exists in
search studies that test reliability.
degrees and is usually expressed as a form
of correlation coefcient, with 1.00 indi- ( i ) Stability
cating perfect reliability and 0.00 indicat- Stability is concerned with the consis-
ing an absence of reliability. The intended tency of the results with repeated mea-
application of the instrument determines sures. Perhaps the easiest way to assess
what is an acceptable reliability coef- stability is to repeat a question that has
cient. For example, if one is using an in- been posed. The idea of a retest procedure
strument to make a lifesaving decision re- is that if the same question is posed twice
lated to a medical treatment, one would and the respondent understands the ques-
demand a high reliability coefcient such tion identically on both occasions, the re-
as 0.90 or above. If, however, the response should be identical on both occa-
searcher is using the instrument for group sions assuming that the variable measured
comparison on a psychosocial phenome- remains the same at the two testing times.
non, a lower reliability coefcient is ac- This type of reliability is especially impor-
ceptable. Nunnally (1978) suggests that re- tant in studies in which an instrument is
Table 13.2 Types of Reliability and Their Tests

Type Definition Test For Reliability Example
Stability Consistent results are Testretest procedure. Gillis, 1997
obtained on repeated Administer a question-
administration of the naire to the same
measurement. person on two sepa-
rate occasions and
examine the consis-
tency of their scores.
A good correlation
suggests good test
retest reliability.
Parallel form. Same as Hoskins, 1988
testretest procedure
except you use two
comparable forms of
the same instrument
Internal consistency Items within the scale or Item-total correlations. Chalmers et al., 1997
instrument reect the This is the correlation
same variable. between each item
and the total scale.
Uncorrected item-
total correlations
above 0.25 are
acceptable.
Kuder-Richardson Sherman, 1996
(KD-20). This correla-
tion is used with in-
struments that have a
dichotomous response
format (yes/no).
Cronbachs alpha. A Gillis, 1997
coefcient measure of
reliability that com-
pares each item in a
Likert scale with each
other simultaneously.
Split-half procedure.
Divide the instrument
into two halves to
make a comparison.
Scores from the two
halves are correlated.
The Spearman-Brown
formula is used to
calculate the correlation
coefcient.
Equivalence Different observers or Interrater reliability. Oliver and Redfern,
different forms of an Two or more observ- 1991
instrument yield the ers rate a situation
same results. and scores are corre-
lated. It is expressed
as a correlation co-
efcient or as a
percentage of agree-
ment between scorers.
Panel form. See Stability.
administered at two or more points in dividual item scores to get a total score
time. For example, in a nursing interven- that reects the construct. Often nursing
tion study, a researcher may administer research deals with phenomena that are
the instrument at time 1, apply an inter- multidimensional such as job satisfaction,
vention to an experimental group, and re- health motivation, social support, satis-
administer the same instrument after the faction with nursing care, or health sta-
intervention at time 2 to measure differ- tus. In these situations, the nurse re-
ences in a dependent variable in both the searcher must establish separate internal
experimental and control group as a result consistency estimates for each dimension
of the intervention. of the construct. For example, Chalmers
Gillis (1997) tested the reliability of the and associates (1997) developed the Pri-
43-item Adolescent Lifestyle Question- mary Health Care Questionnaire (PHCQ),
naire (ALQ) by using the testretest pro- which includes three dimensions: knowl-
cedure. She administered the ALQ to 65 edge about PHC, attitudes toward PHC,
school-based adolescents at an interval of and an open-ended section that ad-
2 weeks. The adolescents ranged in age dresses PHC practices. Separate internal
from 12 to 19 years. The Pearson product- consistency reliability estimates were de-
moment correlation coefcient was 0.88 termined for the 35 knowledge items
for the total score and ranged from 0.80 to (Cronbachs alpha of 0.76) and the 34-item
0.88 for the seven subindexes. This high attitude index (Cronbachs alpha of 0.85).
stability coefcient indicated that scores Methods available to test the inter-
changed very little between the test and nal consistency of a measure include
retest time. Hence, the instrument demon- the Kuder-Richardson (KR-20) coef-
strates high stability and a respectable cient, item-total correlations, split-half relia-
level of reliability. bility, and Cronbachs alpha. A full discus-
A second simple approach, although sion of these methods is beyond the scope
often not feasible, is to verify the answers of this text. Interested readers are referred
independently. Occasionally, for exam- to Table 13.2 for examples of each.
ple, it may be possible to compare a stu- When you are assessing the reliability
dents self-reported grade to that actu- of the items being used to construct an in-
ally received by the student. Here the dex (an index represents the combination
issue is the extent to which students sys- of several items into a single score), you
tematically over- or underreport grades. can randomly split the items into two
groups, compute the indexes, and then
( ii ) Internal Consistency correlate the resulting two scores. Inter-
Internal consistency refers to the ability nal reliability is indicated by a high corre-
of the items in an instrument to measure lation between the two groups of items.
the same variable. Homogeneity is often This method is known as a split-half
used to refer to the internal consistency method and was one of the rst methods
of an instrument. A homogenous instru- available for testing reliability (homo-
ment contains items that are strongly geneity). The advantage of this procedure
correlated to each other. The higher the is that it allows researchers to assess
intercorrelations among the items, the testretest reliability without administer-
greater the internal consistency of the ining the instrument twice. The Spearman-
strument. Homogenous instruments are Brown correlation is usually used in this
unidimensional; that is, the items taken procedure.
together measure one construct. Because Another procedure to assess homo-
all the items reect the same underlying geneity is to compare an individual
construct, the researcher can sum the in- items correlation with the total index
score. If an item is consistent with the to- items. Using the 0.30 rule is recom-
tal score, it will correlate with it. This mended to provide greater uniformity in
technique, known as the internal consis- the development and assessment of in-
tency approach to reliability is discussed dexes. Table 13.3 indicates the alpha val-
further when index construction is de- ues that are associated with the number
scribed. The item-total correlations rep- of items and the average interitem corre-
resent a group of correlations between lation.
each item and the total score for the in- Gillis (1997) tested the internal consis-
strument. Item-total correlations are use- tencies of the total Adolescent Lifestyle
ful in instrument development. They are Questionnaire and each of the subindexes
helpful in determining which items to re- using Cronbachs alpha. The total instru-
tain and which ones to discard in creat- ment was found to have high internal con-
ing a nal index. The item-total correla- sistency, with an alpha coefcient of 0.91.
tions should be high enough to indicate Alpha coefcients for the seven sub-
internal consistency but low enough to indexes ranged from 0.60 to 0.88 (Table
indicate that they are not redundant. 13.4).
Nunnally (1978) suggests that uncor-
( iii ) Equivalence
rected item to total correlations above
Equivalence refers to the degree of agree-
0.25 and below 0.70 are typically accept-
ment among two or more different ob-
able for item retention when developing
servers using the same measurement tool
a new index.
or the agreement between two or more al-
The most common method used by
ternate forms of an instrument. Equiva-
nurse researchers for assessing internal
lence for parallel versions of an instru-
consistency is Cronbachs alpha. This
ment is determined by correlating the two
test is based on the strength of the inter-
scores with each other. Interrater reliabil-
correlations of all the items in the instru-
ity assesses the extent to which two peo-
ment as well as the number of items used
ple consistently agree or disagree on ob-
(see Appendix A). The alpha value
servations. The formula for calculating
ranges from 0 to 1, with 1 indicating per-
interrater reliability is provided in Chap-
fect internal consistency and 0 no inter-
ter 7 as part of the discussion of content
nal consistency. Because the alpha coef-
analysis. Refer to that section for a review
cient is dependent on both the number
of the calculations. It should be noted that
of items used in creating the index and
interrater reliability may be determined
the average interitem correlation, there
several times in a study because factors
is no easy answer as to how high a value
such as boredom, fatigue, or familiarity
is required. As a rule of thumb, you need
may interfere with the raters ability to ac-
an average interitem correlations above
curately rate observations.
0.30 and, with ve items in the index, that
No matter what indicators are used, re-
would yield an alpha of 0.682. Normally it
searchers always try to reect precisely
seems easiest to judge the adequacy of
the reality that is being described.
an index by the average interitem corre-
lation rather than the alpha value itself.
Because alpha values increase as a func- b. The Idea of Reliability in
tion of the number of items used, one can Qualitative Research
increase the alpha value simply by
adding items. An alpha of 0.700 means When a particular group is studied using
one thing if there are ve items in the in- qualitative methods, there are problems
dex, but quite another if there are 14 with replication because the circum-
Table 13.3 Cronbachs Alphas for Various Interitem Correlations and Number of Items*
Mean Interitem Correlation
Items, n 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90
2 0.182 0.333 0.462 0.571 0.667 0.750 0.824 0.889 0.947
3 0.250 0.429 0.562 0.667 0.750 0.818 0.875 0.923 0.964
4 0.308 0.500 0.632 0.727 0.800 0.857 0.903 0.941 0.973
5 0.357 0.556 0.682 0.769 0.833 0.882 0.921 0.952 0.978
6 0.400 0.600 0.720 0.800 0.857 0.900 0.933 0.960 0.982
7 0.438 0.636 0.750 0.824 0.875 0.913 0.942 0.966 0.984
8 0.471 0.667 0.774 0.842 0.889 0.923 0.949 0.970 0.986
9 0.500 0.692 0.794 0.857 0.900 0.931 0.955 0.973 0.988
10 0.526 0.714 0.811 0.870 0.909 0.937 0.959 0.976 0.989
11 0.550 0.733 0.825 0.880 0.917 0.943 0.962 0.978 0.990
12 0.571 0.750 0.837 0.889 0.923 0.947 0.966 0.980 0.991
13 0.591 0.765 0.848 0.897 0.929 0.951 0.968 0.981 0.992
14 0.609 0.778 0.857 0.903 0.933 0.955 0.970 0.982 0.992
15 0.625 0.789 0.865 0.909 0.938 0.957 0.972 0.984 0.993
16 0.640 0.800 0.873 0.914 0.941 0.960 0.974 0.985 0.993
*If we set 0.30 as the minimum average, interitem correlation for an acceptable index, the column headed by 0.30 would
constitute the minimum alphas required. Thus, the bolded values represent those that are acceptable. Unbolded values
are below the acceptable level. For example, if you had a 12-item index and the mean interitem correlation was 0.20, the
alpha value would be 0.750. Because this is in the nonbolded area, the index would not be sufciently reliable. However,
three-item index, with an alpha of 0.562 should be treated as an adequate index.
stances and the individuals can never be cating a study, LeCompte and Goetz
the same at some later time. (1982) have suggested the use of ve
strategies:
. . . what observers see and report is a
function of the position they occupy 1. Focus on verbatim reports and stick to
within participant groups, the status ac- the facts.
corded them, and the role behavior ex- 2. Use multiple researchers because it al-
pected of them. Direct observer effects lows the results of the researchers to
may occur when informants become be compared.
dependent on the ethnographer for sta- 3. Use participant researchers; this in-
tus enhancement . . . (LeCompte and volves training individuals in observa-
Goetz, 1982, p. 46). tional techniques.
4. Use peer examination. When careful de-
Social settings in which questions are scriptions have been made, researchers
asked can also be important. So unless can check their results against the ob-
the qualitative researcher fully reports servations and experiences of fellow re-
how and where the observations were searchers.
made, there is little chance of replicating 5. Use mechanical recording devices such
a study. To increase the chances of repli- as audiotapes and videos to allow oth-
to eliminate rival hypotheses face a

challenge because new hypotheses can
always be suggested. Qualitative re-
searchers have to test alternative hy-
potheses continually as observations
are being made. Because observations
Image/Text rights unavailable are made over time, there are opportu-
nities to explore such hypotheses as
the study unfolds.
C. MEASUREMENT ERROR
Typically, we begin with the assumption

that the manifestations of the phenome-
non we wish to measure have:
1. Two or more values inherent in them

( i.e., we are dealing with a variable, not
ers to check your observations inde- a constant).
pendently at a later date. 2. That any manifestation has a true value.
Avoiding incorrect conclusions, such
Measurement error always occurs be-
as identifying relationships incorrectly, is
cause our instruments are imperfect, our
a challenge for qualitative researchers.
research participants do not always pay
There are three issues that must be han-
sufcient attention to our instructions, or
dled:
we are not careful enough in coding data.
1. Qualitative researchers must establish It exists in both direct and indirect mea-
the ordering of events and phenome- sures of variables.
non; because qualitative researchers In essence, measurement error is any
make observations over time, estab- deviation from the true value. As de-
lishing the ordering of events is usually scribed previously, a true value is the un-
not problematic. Researchers are pres- derlying exact quantity of a variable at
ent on the research site over time and any given time. In measurement, we at-
are able to observe actions and reac- tempt to reect this true value as pre-
tions. cisely as we can. By specifying at any
2. Researchers must establish that the given time, we acknowledge that vari-
variables are related to one another ables change over time and that any mea-
(vary together); researchers depend, to sure will vary from day to day. Just like
some degree, on the occurrence of so- our weight, such measures vary slightly
cial events that permit her or him to see from one day to the next. And although
whether the variables are varying to- some variables, such as our gender or re-
gether. In contrast, experimentalists ligious afliation, remain quite stable, it is
create a situation that systematically nevertheless possible that a change may
varies the intensity of the treatment and occur. Sex change operations and reli-
observe the reaction to the treatment. gious conversions are not unknown.
3. Researchers must eliminate rival hy- Measures are made up from the fol-
potheses. All researchers who attempt lowing components:
MEASURE TV (SE RE) the true value and the measurement value is
without pattern. It may be high in one case and
The above equation contains four elements:
low in a subsequent case. Random error usually
MEASURE. This refers to the value that the re- results from a transient state in the study partic-
searcher assigns to the variable in the process of ipants or in the administration of the instrument.
recording the information. For example, anxiety on the part of study partic-
TV. The True value is the underlying exact quan- ipants can affect responses on the instrument.
tity of the variable. An example: Suppose a male respondent is asked
SE. Systematic error is nonrandom error repre- to indicate his weight on a questionnaire. The re-
senting systematic under- or overestimation of spondent writes in 70 kilo. Providing that the in-
the value. In systematic errors, the variation in formation is then correctly transcribed in entering
measurement values is usually in the same direc- the data into the computer, the MEASURE is
tion and is caused by a stable characteristic of the recorded as 70.0 kg; suppose the TV, or true
population studied or the measurement tool. For value, is 73.367132 kg (rounded!). People usually
example, a blood pressure machine that mea- underreport their weight by 2 kg; this is SE, the
sured diastolic blood pressure at four points systematic error; the remaining 1.367132 kilo-
higher than the true diastolic reading would ele- grams is random error (RE).
vate by four points the systematic error in each
persons reading.
MEASURE TV (SE RE)
RE. Random error is fluctuation around the true
value, in which higher or lower scores are equally or
likely (Carmines and Zeller, 1979). With random
errors, the direction of the difference between 70.0 73.367132 (2.0 1.367132)
The adequacy of a measure is the ex- in one direction. For example, if you tell
tent to which the indicator reects the respondents what you expect to nd,
true value of the variable. All of the listed they may bias their responses to conrm
components vary over time and through the expectation. When feasible, consider
the various stages of data collection in a using some of the following ways of re-
research project. Usually we think that the ducing error:
subject makes measurement errors, but
actually the situation is more complex. Tip 1. Take the average of several mea-
First, the true value of variables changes sures. Sometimes it is possible to re-
over time. If we repeat a measure several peat a measure several times and then
times to make certain that we have an ac- use the average of these measures to
curate reection of the variable, differ- reect the variable. A subject could be
ences will be caused by variations in the weighed on three different scales and
true value, random error, and systematic the average used. In this way, the re-
error. (Respondents may change their at- searcher hopes to average out random
titudes over time, so the true value itself measurement error.
may shift; theoretically, it is xed only at Tip 2. Use several different indicators. In
the time measured.) measuring a variable such as attitudes
Researchers can do several things to toward abortion, a researcher would
reduce random and systematic error. typically pose several questions and
Random error refers to uctuations that then combine the responses to form
are unsystematicfor example, a rean index. By combining the responses
spondent who cannot decide whether to to several questions, the researcher
rate the quality of lectures in the prenatal hopes to minimize the effect of any one
course as a 6 or a 7 on a 9-point index and question.
nally decides on a 6, may have made a Tip 3. Use random sampling procedures.
random choice. Systematic error is error By giving all people an equal chance of
being included in your study, it is pos- one interpretation. You might say, In
sible to minimize distortions that oc- the following items, circle a number to
cur when people select themselves. If indicate the strength of your opinion.
the goal, for example, was to estimate Tip 6. Error checking data. Both system-
the popularity of a national health-care atic and random errors can occur eas-
program, one would want to reect the ily. Conducting error checks on the data
views of the whole country, not special should eliminate these. If data are being
groups within it. By reporting on what coded (the process of assigning cate-
the average person thinks about the gories to responses) by several people,
health program, you will minimize the interrater reliability checks should be
random uctuations that might occur carried out in training the individuals to
if you talked to very few, unsystemati- do the work. Only when high levels of in-
cally selected respondents. By not terrater reliability are achieved should
oversampling those who spend more the coding proceed (see Chapter 6). The
time at home, you will have avoided use of video recording equipment has
systematically biasing the study by permitted those doing observational
overrepresenting the views of people studies to conrm their data by having
with that characteristic. other experts examine the work.
Tip 4. Use sensitive measures. In asking Tip 7. Reduce subject and experimenter
questions, provide respondents with as expectations. As discussed in Chapter
broad a range of response categories as 8, subject and experimenter expecta-
possible so as not to constrain them. In tions can systematically alter the mea-
questions used to create indexes, pro- sures achieved in a research project.
vide many response categories; in ask- Attempt to control these expectations
ing respondents to estimate the popu- as much as possible.
lation of their community, provide
many categories. By providing many re- Precise measurement is, indeed, a chal-
sponse alternatives (or, indeed, allow- lenge. Any time a measurement is taken,
ing the respondent simply to respond errors will occur. The errors may be slight
without using any categories at all), the or substantial. Measurement should be
researcher can decrease the amount of seen as a matter of probability; most of
random error in measurement. How- the time, a researchers measurements
ever, you can go overboard in suggest- will fall within a given margin of error.
ing increased numbers of response cat- This implies that, some of the time, mea-
egories. If there are too many response surements will not fall within a given (and
categories, a respondents ability to acceptable) margin of error. Researchers
make distinctions may be exceeded. To attempt to estimate the amount of error
ask someone to report their weight to that is likely in their measurements. The
two-decimal point accuracy would be specic procedures for doing this are ex-
silly; few people know their weight with plained in Chapter 11.
such accuracy. Figure 13.3 presents a target-shooting
Tip 5. Provide clear instructions for how analogy to help show the relationship be-
questions are to be answered. Re- tween reliability, true value, and validity.
spondents varied understandings of In measurement, we attempt to consis-
what is being asked of them are one tently (reliability) hit the bulls eye (the
source of random error. Thus, varia- true value) and if we do so, we have mea-
tion in response may simply be the re- sured what we intended to (validity).
sult of these different understandings. An inspection of Figure 13.3 shows that
Attempt to develop instructions and a measure may be reliable but lack valid-
questions that are clear and have but ity; however, a measure that is valid is
True Value
Diagram 1: Not reliable (note the spread in bullet holes)
Not valid (cannot be valid if not reliable)
True Value
Diagram 2: Reliable (note the tight grouping)
Not valid (missed the bull's eye or True Value)
True Value
Diagram 3: Reliable (tight grouping)
Valid (tight grouping and on bull's eye, or true value)
Figure 13.3 True value, reliability, and validity.
also reliable because measures that hit by the measurement level achieved in
the bulls eye are closely grouped and, data collection. As a general rule, one
therefore, must be reliable. The three di- should attempt to achieve the most pre-
agrams illustrate the point. cise measurement possible.
One way to begin to understand levels
Diagram 1 depicts measurements that
of measurement is to ask if the variable
are neither reliable nor valid; the mea-
being measured has an underlying con-
surements have missed the true value
tinuum (does it vary from low to high and
(bulls eye) and produced a lot of scat-
do the categories imply an underlying
ter.
quantitative dimension?). If there is no un-
Diagram 2 displays a reliable measure
derlying continuum, with the variable be-
but one that lacks validity. Note that
ing made up of a number of discrete cate-
the shots are closely grouped together
gories, then the measurement will be at
(hence reliable) but that they are not
the nominal level; if the variable has an
right on the bulls eye, which repre-
underlying continuum, then the level of
sents the true value.
measurement will be either ordinal, inter-
Diagram 3 displays a valid and reliable
val, or ratio. As discussed in Chapter 11,
measure in which the shots are nicely
this text distinguishes three levels of mea-
grouped on the bulls eye.
surement: nominal, ordinal, and ratio.
D. LEVELS OF MEASUREMENT 1. Nominal Measurement
Three levels of measurement were pre- The nominal level represents the lowest
sented in Chapter 11. An understanding of level of measurement. Examples of nomi-
levels of measurement is necessary be- nal measurement (those with an arbitrary
cause the ways in which one should go assignment of numbers to categories) in-
about analyzing a variable are constrained clude measurements of such variables as
religious afliation, gender, program of measurement involves an underlying con-

study, political party afliation, and ethnic tinuum with the numerical values ordered
origin. Although there may be underlying so that small numbers refer to lower levels
continua related to religious afliation on the continuum and larger numbers to
(e.g., degree of religious commitment or higher points; however, the distances be-
frequency of church attendance), by itself, tween the assigned numbers and the un-
the religious organization is a nominal cat- derlying continuum are not in a one-to-one
egory; it may be Baptist, Buddhist, Hindu, relationship with each other. Note the fol-
Lutheran, Roman Catholic, or Jewish. lowing questionnaire item asking about
When a respondent checks off which (if the size of a respondents current home
any) religious afliations he or she was as- community:
sociated with while growing up, the mea-
surement level attained is nominal. One Under 5000 1
category is neither higher nor lower than 5000 to 19,999 2
any otherthe categories are simply dif- 20,000 to 99,999 -3
ferent. For example, Catholics are no 100,000 to 999,999 -4
higher or lower than Protestants. Nominal Over 1 million -5
measurement involves no underlying con-
tinuum and the numerical values assigned Note that the numbers assigned by the
are arbitrary and have no arithmetic researcher (1 through 5) are arranged so
meaning. Such values cannot be added, that higher numbers refer to larger popu-
subtracted, multiplied, or divided. Addi- lation centers; but note that the intervals
tional examples are provided in Table 13.5. between the numbers are not equal
(whereas the second category spans a
2. Ordinal Measurement population range of 15,000, the fourth cat-
egory spans a range of 900,000).
The ordinal level represents the second Another frequently used type of ques-
lowest level of measurement. Ordinal tion asks respondents to assess their de-
Table 13.5 Examples of Nominal Level Measurement*

Type of Research Design Variable Measure
Experiment Expectancy created? Subject run under conditions in which the
outcome is:
Expected - - - - - - - -1
Not expected - - - - 2
Survey questionnaire Place of residence The place where I live now is:
Canada - - - - - - - - - -1
Mexico- - - - - - - - - - 2
United States - - - - 3
Other - - - - - - - - - - -4
Participant observation Behavior related to illegal George:
activities Initiates activities - - - - - - - - - - - - - 1
Supports activities of others - - - - 2
Rejects activities of others - - - - - -3
Content analysis Type of gender role TV Character X plays a gender role that is:
character plays Traditional - - - - - - - - 1
Nontraditional - - - - - 2
*Nominal measurement involves no underlying continuum; numeric values assigned have no meaning; values cannot be
added, subtracted, or multiplied.
gree of agreement with a statement such subtracted but not multiplied or divided
as the following one: because there is no true zero point.
Nurse practitioners enhance the quality of
care in primary health centers. 3. Ratio Measurement
Strongly disagree 1 2 3 4 5 6 7 8 9 Strongly agree
Finally, ratio measurement is the highest
This kind of item provides ordinal level. Ratio measurement involves an un-
measurement. Although we know that derlying continuum. The numerical val-
high numbers indicate a greater degree of ues assigned are ordered with equal in-
agreement with the statement, we do not tervals; the zero point is aligned with true
know whether the distances between the zero; and when different ratio measures
values are equal; the distance between 4 are combined, the values may be added,
and 5 on the scale may not be the same subtracted, multiplied, or divided. Fur-
size as the distance between 8 and 9. Or- thermore, it should be readily apparent
dinal measurement orders values but that a ratio scale has all the qualities of
does not assure equal gaps between the lower levels of measurement (nominal,
measurement points. Additional exam- ordinal, interval), that an interval scale
ples of ordinal level measurement are has all the qualities of the nominal and or-
provided in Table 13.6. dinal scales, and so on.
In conclusion, ordinal measurement in- For example, in the case of income, the
volves an underlying continuum; numeri- nature of the variable is such that it is pos-
cal values assigned are ordered but inter- sible to represent income with a number
vals are not equal. When ordinal measures that reects the income of a person ex-
are combined, the values may be added or actly. In this case, it is also possible to use
Table 13.6 Examples of Ordinal Level Measurement*

Experiment Expectancy level created Subject run under conditions of:
Low level - - - - - - - - - 1
Medium level- - - - - - 2
High level - - - - - - - - 3
Survey questionnaire Size of community The place where I live now has a population
of:
Under 5000 - - - - - - - - - - -1
5,000 to 19,999 - - - - - - - -2
20,000 to 99,999 - - - - - - - 3
100,000 to 999,999 - - - - - 4
More than 1 million- - - - 5
Participant observation Supportive behavior Support of illegal activity:
No support - - - - - - - - - - 1
Some support - - - - - - - 2
Strong support - - - - - - - 3
Content analysis Degree of traditional Character X plays a gender role that is:
gender role Nontraditional - - - - - - - -1
Neutral - - - - - - - - - - - - - - 2
Traditional - - - - - - - - - - - 3
*Ordinal measurement involves an underlying continuum; numeric values assigned are ordered but intervals are not
equal; values may be added or subtracted but not multiplied.
zero to reect no income and other nu- various rates and ratios (e.g., suicide
merical values to reect all other income rate, dependency ratio, gender ratio) are
levels. Here it is correct to say that an in- all ratio level measures (see Table 13.7
come of $50,000 is twice as much as an infor additional examples).
come of $25,000. With ratio-level measure- Some textbooks distinguish equal in-
ment, it is possible to add and subtract terval as a fourth level of measurement.
constants as well as to multiply or divide Box 13.1 explains what happened to the
by them without changing the proportion- fourth level of measurement frequently
ality among the values. distinguished in research texts.
If a researcher were to have respon-
dents indicate with a check mark which of
the magazines on a list that they had E. THE EFFECTS OF REDUCED
scanned or read in the previous month and LEVELS OF MEASUREMENT
a total was taken of the number ticked off,
the result would be a variable that varies This chapter has encouraged researchers
from 0 to the number required if a respon- to achieve the most precise measure-
dent checked off all the items listed. This ments that are practical. To explore the
number would be a ratio measurement. consequences of reducing the level of
Note that a 0 in this example refers to no measurement, the authors examined what
exposure to any of the magazines listed, would happen if a ratio level variable (stu-
and other values simply provide a count of dents high school average grade) was re-
the number of the listed magazines that grouped using random numbers to estab-
the respondent has scanned or read in the lish cutpoints, into 9, 7, 5, and 3 categories.
past month. There is a one-to-one relation- (To regroup the high school average
ship between the value assigned by the re- grade into three categories, two random
searcher and the number of listed maga- numbers were selected that fell between
zines identied by the respondent. the lowest and highest value and these es-
Similarly, when communities are being tablished the cutpoints. If the random
studied, measures such as the propor- numbers were 49 and 72, then all grades
tion of visible minorities, the proportion between 50 and 71 would be coded a 2;
of retired people in the population, and grades between 0 and 49 coded as 1; and
Table 13.7 Examples of Ratio Level Measurement*

Experiment Accuracy of recall Number of correct answers
in trial: ___________
Survey questionnaire Weight? My weight is:
___ pounds
or
___ kilos
Participant observation Community involvement Number of people at health fair on
January 15: ___
Content analysis Conservative gender Ratio of men portrayed in nontraditional
portrayals gender roles to women portrayed in non-
traditional gender roles: ___ : ___
*Ratio measurement involves an underlying continuum; numeric values assigned are ordered with equal intervals, the
zero point is aligned with true zero. When different ratio measures are combined, the values may be added, subtracted,
multiplied, or divided.
BOX 13.1 What Ever Happened to Equal Interval Measures?
The conventional equal interval level of mea- characteristic being measured and not placed in
surement (Stevens, 1951) is not presented in some pre-coded set of categories with uneven
this text for the sake of pedagogical simplicity. category sizes.
For analyses that require equal interval or ratio Similarly, measures reflecting properties of
measurement, in most cases the measurement communities or countries (or other groupings)
will be ratio level. Many texts in the social sci- such as the proportion of nonnative-born mem-
ences use the temperature measures of Celsius bers, the crime rate, and the number of motor
versus Fahrenheit as examples of situations in vehicles per person all represent ratio level
which the intervals between the points are measures.
equal but where the zero point is misplaced (ab- Although most textbook authors present
solute zero is about 273F). The table below the four types of measurement, some do not
shows the traditional measurement levels and (see Levin and Fox, 1991; Rosenthal and Ros-
their associated properties. now 1991; Jackson, 1999). For the sake of sim-
Arguably, most social science measures that plicity, equal interval measures are not identi-
meet the equal interval assumption also meet fied as such in this text. Readers should be
the ratio measurement assumption of a cor- aware that the measurement assumptions of
rectly aligned zero point (see Blalock, 1979). In- various statistics that often call for interval level
dividual variables such as age, weight, height, in- measurement are stated as ratio level measure-
come, years employed, number of magazines ment in this presentation. But because virtually
read last week, and the number of dates one all the social science measures that are at this
had last month all would constitute ratio mea- level are ratio level measures, this should rarely
sures providing the data recorded reflect the prove to be a difficulty.
Table 1 Traditional Levels of Measurement

Measurement Level
Measurement Feature Nominal Ordinal Equal Interval Ratio
Identies categories? Yes Yes Yes Yes
Orders observations? No Yes Yes Yes
Equal intervals between points? No No Yes Yes
Properly assigned zero point? No No No Yes
grades 72 and above coded as 3. New ver- of Box 13.2 to appreciate the importance
sions for each of these variables were cre- of precise measurement. Although this
ated 10 times for the categories 3, 5, 7, and material is somewhat technical in nature,
9.) The new versions of the variables were it illustrates a concrete example of the
then correlated with unchanged ratio vari- three principles identied below.
ables (high school English grade and rst- The analysis suggests that when re-
year university grade). As anticipated, the duced levels of measurement are achieved,
correlations declined further with each the result will underestimate the strength
succeeding grouping into fewer categories of association between variables. Further-
(Box 13.2). It is recommended that readers more, in doing analyses that permit a com-
take time to carefully review the contents parison of the relative effects of indepen-
BOX 13.2 The Effects of Reduced Levels of Measurement
An attempt to examine the consequences of re- be reduced in comparison to those using

duced levels of measurement was carried out. raw scores.
Hypothesis 4: The above effects will be
1. The hypotheses. The hypotheses that guided greater for those variables having fewer cate-
this investigation were as follows: gories.
Hypothesis 1: The greater the number of Hypothesis 5: Beta weights of variables not
categories in an ordinal variable, the less re- categorized will be enhanced in comparison
duction there will be in the correlation be- to those where categorization has been done.
tween the variable and other variables.
A data set containing 3617 cases was used in
2. In doing a regression analysis: this investigation. The variables included uni-
Hypothesis 2: A ratio variable that is mea- versity students high school average, high
sured using a Likert-type question (ordinal school English grade, and average at the end of
level measurement) will reduce the amount their first year of university study. Correlations
of variation explained in the analysis com- were calculated (see Table 1) for the relation
pared with how much variation would be ex- between these variables. The correlation be-
plained had ratio level measurement been tween university average and the English grade
achieved. high school average was 0.464 and 0.573, re-
Hypothesis 3: The beta weights of Likert- spectively; the high school English grade and the
type items (ordinal level measurement) will average high school grade correlated at 0.662.
Table 1 Correlations Between First-Year University Average, Average High School

Grade, and English High School Grade (n 3617)
First-Year Average High English High
Correlations University Average School Grade School Grade
First-year university average 1.000
Average high school grade 0.573 1.000
English high school grade 0.464 0.662 1.000
A regression analysis was done taking first- R2 was 0.341 (34.1% of the variance ex-
year university average as the dependent vari- plained); the beta weight for the English grade
able, using high school English grade and the was 0.152; for high school average it was
high school average grade (excluding the Eng- 0.472. The impact of high school average was
lish mark) as independent variables. The re- about three times as great as the English grade
sults of this analysis are shown in Table 2. The alone.
Table 2 Regression Analysis Predicting First-Year Average Grade (n 3617)

Variable b Coefficient Beta Weight Percent Explained
High school English grade 0.16290 0.15195 8.30
High school average* 0.54188 0.47203 25.78
% Explained 34.08
Constant 11.08264
Multiple R 0.58383
R square 0.34085
*The high school average was calculated excluding the high school English grade.

BOX 13.2 The Effects of Reduced Levels of Measurement
In an attempt to test the hypotheses, a series this way (10 for 9 category variables, 10 for 7
of trials was done on the data testing the effect category variables, 10 for 5 category variables
of regrouping the high school average variable. and 10 for 3 category variables). In all cases, the
We wished to compare what would happen if it new variables were used as an independent vari-
were collapsed into 3, 5, 7, and 9 categories. We able in each trial along with the English grade in
also wished to observe the impact of such col- all cases using the university average as the de-
lapsing on the R2, and the beta weights. A table pendent variable. The results of the correlational
of random numbers was used to determine the analyses are summarized in Table 3; Table 4 sum-
cutpoints; a total of 40 variables were created in marizes the results of the 40 regression analyses.
Table 3 Average Change in Correlations Between High School Average Grade and First-
Year University Grade When Data Are Grouped into 3, 5, 7, and 9 Categories*
Number of Average Decline Standard Number of
Categories in Correlation Deviation Recodings
9 0.03810 0.03151 10
7 0.05790 0.03208 10
5 0.11260 0.09623 10
3 0.18240 0.16038 10
*The raw data, based on 3617 cases (St. Francis Xavier University), was used for this analysis. The high
school average grade was recoded into 9, 7, 5, or 3 categories, 10 recodings for each category. For each,
the new variable was then correlated with the rst-year university average. The above table summarizes the
results. For example, when the data were recoded into three categories, the average drop in correlation
(for the 10 randomly determined cutpoints) was 0.18.
Table 4 Summary of Changes in R2 and Betas When One Variable Is Grouped into 9, 7, 5,
and 3 Categories, Regression Analyses
Number of Categories Mean Change Mean Change Mean Change
in Recoded Variable in R2 in Beta HS in Beta ENG
9 Categories 0.025 0.067 0.059
7 Categories 0.033 0.099 0.094
5 Categories 0.055 0.162 0.135
3 Categories 0.069 0.217 0.179
Probability 0.0059 0.0039 0.0001
HS 5 high school average grade; this is the variable that was recoded.
ENG 5 high school English grade; this variable was left in its raw form for the analysis.
This analysis summarizes the results of 40 separate regression analyses (10 for recoding into 9 categories;
10 for the recoding into 7 categories; 10 for the recoding into 5 categories; and 10 for the recoding into 3
categories). The total n for each analysis is 3617. Table 2 presents the results when the high school average
is used in its raw form.
SOURCE: Winston Jackson. The author wishes to thank Dr. Bernard Liengme, registrar, St. Francis Xavier Univer-
sity, for providing the data for the analysis.
dent variables on a dependent variable able relative to other, more precisely mea-
(see Chapter 18), the effect of an indepen- sured variables.
dent variable that has a reduced level of
measurement will be underestimated. The
research suggests the following general
principles: F. PROCEDURES FOR DATA
COLLECTION
Principle 1. The greater the reduction of
measurement precision, the greater
Data collection is the process of gathering
the drop in the correlation between
information from identied participants to
the variables.
answer a research question. The amount
Principle 2. In analyses comparing rela-
of time, planning, and energy required for
tive effects of variables, the effects of
data collection depends upon the specic
those variables with reduced levels of
research question, design, and measure-
measurement are underestimated by
ment tools available. In preparation for
comparison with those whose mea-
data collection, the researcher must an-
surement is more precise.
swer a number of questions, such as:
Principle 3. Conversely, variables whose
measurement is more precise will have What data are to be collected?
their effects overestimated relative to What methods will be used to collect
those whose measurement is less pre- the data?
cise. From whom will data be collected?
These three principles suggest that we
How will participants be identied and
selected?
should attempt to reect the underlying
concepts as precisely as possible. But
Who will collect the data?
there are constraints. It is foolish to ask
Where will the data be collected?
people to give us exact answers when there
When will the data be collected?
is little likelihood that they will be in the po-
How will consistency of data collection
methods be established and moni-
sition to make precise estimates. For ex-
tored?
ample, asking anyone on campus to esti-
mate the weight of a school bus would Each question necessitates an answer
invite highly variable and inaccurate an- by the researcher before data collection
swers. Furthermore, it is sometimes inap- commences. The answer varies depend-
propriate to ask for extremely precise ining on the nature of the research project.
formation if a respondent will consider The researcher should detail strategies
such information too personal. For exam- used to address each question in the writ-
ple, in some populations, asking someone ing of the research report. This informa-
his or her exact annual income would fall tion is useful in evaluating the study re-
into that category. Sometimes, in order to sults or planning a replication of the study.
soften a question and to make it less threat- Box 13.3 illustrates a description of the
ening, we reduce its precision. In asking data collection procedure as it would ap-
about age, for example, we sometimes ask pear in a journal article. Note the article
people to indicate into which age category describes who the participants were, how
they fall rather than reveal their exact age. they were selected, the sites for data col-
So the rule to be followed is to achieve as lection, and the measures used to collect
precise measurement as is practical. Re- the information from participants.
searchers recognize that weaker measure- A variety of quantitative and qualita-
ment typically result in underestimating tive data collection methods are avail-
the importance of a poorly measured variable to nurse researchers to gather infor-
STRESS, COPING, AND SOCIAL SUPPORT

AMONG HOMELESS YOUTHS
Participants in this study were homeless or po- reviewing the agencys sign-in roster to deter-
tentially homeless youths who were ages 13 mine how many youths were in the agency, they
through 23 years. Homelessness included living determined the sampling fraction from a sam-
in one of the following places during the past 12 pling fraction table according to the number of
months: street; subway; abandoned car or potential participants. Using the sampling frac-
building; park; beach; car, bus, or van; girlfriend, tion and a predetermined random start number,
boyfriend, lover, or friends house or apart- youths were selected from the sign-in roster and
ment; foster home; group home; halfway house; invited to participate in the study.
treatment center; shelter; mission; motel; hotel; Street sites. Street-based sampling occurred
boarding house; or jail. along five major boulevards. Along these boule-
vards, 73 street areas, four parks, and three
Procedure restaurants were identified as natural PSUs. In-
Systematic sampling methods were used to ob- terviewer teams were assigned randomly to
tain a sample of homeless youth in Los Angeles, one of these PSUs, where they determined the
California. Sampling occurred through two number of potentially eligible youths in the area
sampling frames: one fixed or service sites (i.e., and then determined the sampling fraction.
shelter or drop-in centers) and one for natural Youths were then counted and selected ac-
street sites (i.e., street corners) where home- cording to the sampling fraction and the prede-
less youth gather. termined random start number.
Fixed sites. Previous research indicated that Youths selected for recruitment were asked
approximately one half of the Los Angeles home- to answer an eight-item screening instrument
less youth population could be reached through to determine whether they were eligible to
shelters and drop-in centers. Three shelters and participate in a 60- to 75-minute structured in-
six drop-in centers were selected as fixed pri- terview. The interviews were conducted at a
mary sampling units (PSUs), and 50 percent of nearby coffee shop and fast-food restaurants,
the sampling assignments were at these fixed permitting the interview to occur in a safe, quiet
sites. Interview teams consisting of two to four setting. Compensation for participation in-
members were given random assignments to cluded a meal worth $10 and $10 in food
one or more of these fixed sites each day. After vouchers.
SOURCE: Summarized from Unger, J., Kipe, M., Simon, T., et al. (1998). Stress, coping, and social support among
homeless youth. Journal of Adolescent Research, 13(2), 134157.
mation. The choice depends on three iological measures; projective techniques;

concerns (Massey, 1995): and the Delphi technique. Qualitative
methods include unstructured interviews,
1. The identied research problem, ques-
focus groups, observation sessions, and
tion, or hypothesis
records or historical documents.
2. The research design
In qualitative research, data collection
3. The amount of information already
is greatly facilitated by using an organized
known about the variables
process. Indeed, Montgomery (2000) has
The main methods used in quantitative suggested the use of a qualitative data
research include questionnaires; special collection bag that nurses can use to pre-
scales and indices to measure knowledge, pare for successful data collection. The
skill, and attitudes; biochemical and phys- items in the bag include:
Consent forms tions depending on the data produced.

Participant information sheets Not all respondents may be asked the
Pens and paper same questions, nor may all respondents
Notebook for thoughts, feelings, mood be included in all aspects of the investiga-
Materials to record the decision trail tion. The sample may not be determined
Tape recorder with batteries in advance of the data collection proce-
Electricity adapter dure; rather, it must remain uid through-
Audio tapes out the study as the researcher takes the
Return self-addressed, stamped en- liberty of adding new participants de-
velopes ( if needed) pending on the emerging ndings.
Compensation (monetary and non- Nursing deals with many complex phe-
monetary) nomena. Measurement of such phenom-
Interview guide ena presents a challenge to nurse re-
searchers. As the discipline of nursing
Keeping the well-stocked bag close at develops, new constructs are emerging
hand facilitates timely and efcient data and, along with them, so is the need for
collection, minimizing loss of data collec- valid and reliable instruments to measure
tion opportunities. these constructs. Some phenomena of in-
Consistency in data collection proce- terest to nurses remain unexplored be-
dures is important in reducing bias that cause no appropriate strategies exist to
may be introduced if more than one indi- collect data about them. The situation is
vidual is involved in data collection. As changing, however, as more nurse re-
much as possible, data should be collected searchers engage in instrument develop-
in the same way from each participant. For ment studies or borrow instruments de-
example, if more than one instrument is ad- veloped by others and modify them for use
ministered, they should be given in the in nursing investigations. The most com-
same order to all participants, in the same mon data collection strategies in nursing
context and setting, and with the same set research are discussed in this section and
of directions. When assistants are hired to include various indexes and scales; physi-
collect data, they must be trained to do so ologic measures; special measurement
systematically. Usually a written protocol procedures such as magnitude estimation
that outlines the procedures for data col- and visual analogues; and interviews, ob-
lection is provided to research assistants. servations, and questionnaires.
Interrater reliability needs to be estab-
lished and monitored throughout the data
collection period. 1. Indexes
In quantitative designs, data collection
methods are predetermined, standard- Combining several indicators into one
ized, and specied in advance so that little score results in an index or scale. Al-
is required in terms of input from the in- though these terms are often used inter-
vestigator after data collection begins. changeably, when a distinction is made,
The situation is quite different in qualita- an index refers to the combination of two
tive designs in which greater exibility or more indicators. A scale refers to a
and less structure are evident in the data more complex combination of indicators
collection process. In qualitative designs, in which the pattern of the responses is
the researcher is considered the data col- taken into account. The terms are used
lection instrument. The researcher may interchangeably in this text.
vary the questions asked of respondents Indexes are routinely constructed to re-
and may eliminate or introduce new ques- ect psychosocial variables such as social
support, socioeconomic status, satisfac- with a computer program, we would use a

tion with care, group dynamics, or an atti- matrix that has nurses and questions on
tude toward an issue of concern to the the dimensions. Each cell would identify a
eld of nursing. (For details of many clas- correct or an incorrect response to each
sic indexes and scales, see Miller, 1977, or question for a particular nurse. Now we
Frank-Stromberg et al., 1997.) Scales and could order the nurses by the number of
indexes may also reect self-report mea- correct responses and then choose the top
sures on various physical sensations such and bottom quartiles. Table 13.8 shows the
as pain, nausea, appetite, fatigue, sexual percent from each quartile getting each
satisfaction, and so on. Frequently, a re- question correct.
searcher will construct subindexes that The next step is to select those items in
may be treated alone or combined with which the performance of top and bottom
other subindexes to form a composite nurses differs most. We assume here that,
measure. For example, a researcher mea- overall, the questions measure knowl-
suring attitudes toward abortion might edge of the subject matter and that we are
construct a subindex for soft reasons simply choosing those items that discrim-
(e.g., economically inconvenient, prefer- inate best.
ence for having a baby later) and hard The rst two questions discriminate
reasons (e.g., pregnancy as a result of wellthe high scorers do considerably
rape, severely handicapped). These sub- better on those items than the low scor-
indexes might also be combined to form ers. The third question would be re-
an overall index. In each case, however, jected; although 55 percent of the nurses
the researcher will have to ensure that ap- in the top quartile gave the right answer,
propriate items are included in each sub- so did 60 percent of the bottom group.
index. Similarly, question 4 would be dropped
because a similar proportion of top and
a. Item Analysis bottom nurses answered correctly.
Similar procedures may be used to se-
It is important that the components of an lect high discrimination items for in-
index discriminatethat is, various ele- dexes. Suppose, for example, that we had
ments must discriminate between high 15 items for an index measuring nurses
scorers and lower scorers. To illustrate, job satisfaction and we wished to deter-
suppose that you are attempting to de-
velop a set of multiple choice questions
to measure pediatric nurses knowledge Table 13.8 Discrimination Ability of 100 Items:
of growth and development theory and Percent Correct for Each Item by Quartile
that you wish to identify the items that Percent Correct Each Item
best measure mastery of the subject mat-
Question# Bottom 25% Top 25%
ter. Let us suppose that you have 100
questions on the test and that you wish to 1 40.0 80.0
identify the best 50 questions for a future 2 5.0 95.0
test. The issue is to select the items that 3 60.0 55.0
best discriminate between high and low
4 80.0 80.0
performance on the test. Let us suppose
that we have given the preliminary test to 5 10.0 40.0
200 nurses working in pediatric settings. 6 20.0 60.0
We could proceed by grading the test
and computing the total correct responses
100 30.0 20.0
for each nurse. If the marking were done
mine which items to include in the index. titioners, one might identify which, if
We want items that do two things: (1) any, events would lead a respondent to
validly reect the dimension of the con- favor use of nurse practitioners. Items
cept they are supposed to and (2) dis- may initially be selected on the basis of
criminate between high and low scorers. face validitythat is, if the item ap-
We might want to do the following: pears, on the face of it, to represent a
part of the theoretical continuum being
Include the items that have face validity.
measured, then an effort should be
Add them together, coming up with a
made to get a measure of that item.
total score for each individual.
Items usually consist of a stem state-
Split the sample into the top and bot-
ment relating to the phenomenon being
tom quartiles in job satisfaction scores.
measured, a set of scale values, and an-
Test each items ability to discriminate
chors at each end of the scale values.
between high and low job satisfaction.
These aspects are illustrated in the fol-
Select that items that best differentiate
lowing example:
high scorers from low scorers.
Stem statement: I support the use of
nurse practitioners for treatment of
b. Selecting Index Items common health problems.
Anchors: Strongly disagree Strongly
Indexes are constructed by combining sev-
agree
eral individual questions and represent an
Scale values: 1 2 3 4 5 6 7 8 9
attempt to summarize, in one score, a mea-
The following three items on the use
sure of a variable. Indexes are constructed
of nurse practitioners may form part of
in situations in which we have a single-
an index:
dimension variable but where one ques-
I support the use of nurse practitioners in un-
tion might not adequately measure the
derserviced areas of the healthcare sys-
variable. Indexes can be constructed by
tem.
combining a number of similarly formatted
questions or combinations of questions
I support the use of nurse practitioners in all
with different formats. In all cases, the
sectors of the health-care system.
indicators are combined and possibly
weighted to sum to a single index score.
I support the use of nurse practitioners in fam-
The steps involved in developing an index
ily practice centers.
are identied below.
Step 1. Review conceptual denition. As Step 3: Pretest index. Index items should
in developing other measures, the rst always be pretested. A pretest is sim-
step is to review the conceptual de- ple; all you have to do is complete the
nition of the variable. Some sense of index yourself and then sit beside indi-
the range or the dimensions involved viduals who are within the population
in the variable should be developed. being studied and have several com-
The chances are that several questions plete the index items.
can then be designed to measure the Encourage these pretest respon-
variable, reecting each of the dimen- dents to ask questions, telling them in
sions in the conceptual denition. advance that you are trying to nd out
Step 2. Develop measures for each di- if all the questions are clear. Reword
mension. Here the same principles ap- questions to achieve clarity. Pretests
ply as for the development of individual almost always lead to revisions in the
measures. For example, in measuring wording of index items.
attitudes toward the use of nurse prac- Step 4. Pilot test index. If time permits, in-
dexes should be pilot tested. A pilot c. The Rationale for Using Several
study involves having a number of re- Items in an Index
spondents complete a questionnaire
containing the proposed index items. In Suppose we are interested in measuring
a pilot study, you take what you con- nurses attitudes toward abortion. Would
sider to be the nal version of the index it be better to use one question or several
to a sample of respondents similar to questions in our measure? The general
those in the proposed study. The re- consensus among most researchers is
sults are analyzed to see if the index that measures with multiple items have
items discriminate and if they are inter- an advantage over single-item measures
nally consistent. Items may be dropped in attaining precise measurement. Jack-
or altered on the basis of a pilot study. son (1999) has summarized the argu-
If time is not available to do a pilot ments in favor of using multiple items in
test, at least ensure that a full range of measurement:
possible variability is represented by
the index items. When the data are be- An attitude toward abortion would al-
ing analyzed, the various items can be most certainly be complex, and one
evaluated using the SPSS procedure of would want to reect this complexity in
RELIABILITY. This procedure provides any measurement (e.g., would nurses
a method of checking the internal con- have the same attitude toward abortion
sistency of index items. Items may be if the pregnancy threatened the life of
dropped if they do not prove to be in- the mother, was the result of a rape, or
ternally consistent. was economically inconvenient?).
Gillis (1990) developed and pilot A single item may lack precision, may
tested the Knowledge of Growth and lack a sufcient range in values (a ques-
Development Test (KGDT), a 35-item, tion such as, Are you in favor of abor-
multiple choice test designed to mea- tion? with yes or no response cate-
sure a nurses knowledge of child gories would be inadequate. Even with
development on three conceptual di- more response categories, it would still
mensions: (1) knowledge of play, (2) be weak because the question does not
teaching and procedure preparation, identify conditions in which a person
and (3) the stages of psychosocial de- might favor or not favor abortion).
velopment across the ve stages of Single items are also less reliable and
childhood. The KGDT was pilot tested more prone to random measurement
on a sample of senior baccalaureate error. If a question like the one above
nursing students. The effectiveness of were repeated in a questionnaire, the
each item was determined by an item respondent, having indicated opposi-
analysis. Item analysis provided infor- tion to abortion the rst time, might
mation about the percentage of the well change her answer the second
group that answered the item cor- time because she thinks: Well, it de-
rectly and the discrimination index or pends on the situation. If the female
ability of the item to differentiate be- was raped and is 14 years old, I would
tween nurses of high and low ability. be in favor of her right to have an abor-
All items with a D value (discrimination.
tion index value) of 0.65 were main- With multiple items, it is possible to
tained. Three items were reworded be- evaluate the reliability and validity of
cause of their initial negative D values. the index; with single indicators, it is
The KGDT had good internal consis- more difcult to gauge the amount of
tency (Cronbach alpha 0.80). measurement error.
2. Likert-Based Indexes items, fewer values are needed to achieve

high levels of reliability.
Rensis Likert (1931) proposed that in-
dexes or scales could be constructed by
summing respondents answers to a num- a. Tips for Constructing a Likert-
ber of related items. This type of question Based Index
is widely used in nursing research. In the
original format, the respondent is asked The following tips may be helpful in con-
to (1) strongly disagree, (2) disagree, (3) structing such items for use in a Likert-
be undecided or neutral, (4) agree, or (5) based index.
strongly agree with a statement. Such
items were, and continue to be, popular in Tip 1. Avoid the word and in Likert items
measuring attitudes and perceptions. Box if such usage makes the item multidi-
13.4 illustrates some of these questions. mensional. In such indexes, we are at-
Readers should note that the items in tempting to measure a single variable,
Box 13.4 deviate somewhat from the orig- so it is important for us not to add a sec-
inal format. Given a preference for in- ond dimension inadvertently. Suppose
creasing the variability in such items, the we asked a respondent to assess how
number of response categories has been well he or she gets along at home by
increased from ve to nine. This increase making the following statement:
does not take more space on the ques- I get along well with my mother and father.
tionnaire or more space when entered Strongly disagree 1 2 3 4 5 6 7 8 9 Strongly agree
into a computer. Using more response What number is the respondent to
categories (9 point rather than 5 point) circle if he or she gets along well with
should lead to slightly higher correla- mother but ghts continuously with fa-
tions between index items and prove to ther? The question has to be subdi-
be somewhat better at reecting true un- vided into two questions, one asking
derlying values. (See the earlier discus- about relations with mother and one
sion on the effects of reduced levels of asking about relations with father.
measurement.) When there are only a few Watch out for the word and in a Lik-
items in an index, increasing the number ert item. Most of the time, you will
of scale values usually increases the reli- have to change the item.
ability of the scale. When there are many Tip 2. Place the Strongly Agree on the

right-hand side of the scale, with 9 indi- their answers. Likert items result in an
cating strong agreement. Some re- ordinal level of measurement. Typi-
searchers prefer to vary the response cally, such items are combined to form
categories, by, for example, reversing indexes by adding together the values
the side on which the agree and dis- on individual items (after having re-
agree labels are placed. This is done to versed the scores on the negative mea-
prevent response set, a situation in sures). For example, when scoring a
which the respondent tends to answer scale measuring the construct satis-
all items similarly. Although switching faction with a distance education nurs-
the side on which the agree and dis- ing course, the scale values for the neg-
agree categories are placed may reduce atively stated items would be reversed
the tendency to respond in a set man- so that the highest value reects satis-
ner, it may also introduce additional er- faction rather than dissatisfaction. The
rors in response. (Respondents may values for the negative items would
not notice that you have switched the then be summed and added to the val-
agree and disagree categories.) This au- ues for the positive items. Likert in-
thors preference is to maintain a uni- dexes are frequently referred to as sum-
form presentation. Response sets are mated rating indexes. In such indexes,
best avoided by wording some ques- the scores for each item are summed or
tions positively and others negatively. summed and averaged to yield each in-
Tip 3. Avoid multiple negatives that may dividuals score. Typically, the result-
confuse respondent (e.g., statements ing scores are considered to have been
such as, I dont think the university ad- transformed into ratio level data that
ministration is doing a bad job will al- permits the use of more sophisticated
most certainly confuse and slow down data analysis procedures.
respondents). If the respondent thinks
the university administration is doing a b. Evaluation of Likert-Based Indexes
good job, should the respondent circle
the 9 for Strongly Agree, or the 1 for We have assumed that the summation
Strongly Disagree? A direct simple score reects the true underlying vari-
statement is better: The university ad- able. If an item does not correlate with
ministration is incompetent or The this score, we assume that the item is not
university administration is excellent. appropriate. It is, however, possible that
Tip 4. Vary the strength of wording of the other items are in error. If index items
questions to produce variation in re- are not well thought out, the chances are
sponse. Most items should be stated that the correlations between the items
as moderately positive or negative. will generally be low and that this infor-
Statements that reect extremes or are mation may then be used to reject items.
neutral produce less variance in re- To perform this analysis using SPSS, see
sponse and consequently are not ef- the RELIABILITY procedure in Appendix A.
fective in discriminating participants Likert-based indexes are widely used
on the construct being measured. Sim- in nursing research. Their popularity is
ilarly, if there is uncertainty as to because of these factors:
where responses will fall, use more
than one item with different intensities
They are easy to construct.
in the wordings.
There are well-developed techniques
for rigorously assessing the validity of
Tip 5. Before the rst Likert-type item is
potential items.
presented, provide a brief explanation
of how respondents are to indicate
They are relatively easy for respon-
dents to complete.
They require fewer items than other BOX 13.5 Semantic Differential
indexes to achieve the same level of re-
liability when 9-point scales are used. 62. Circle a number to indicate where you
They can be used for group and indi- think you fit on a continuum between
vidual comparisons. the two opposites.
They can be administered in oral or 621 Shy 1 2 3 4 5 6 7 8 9 Outgoing
written formats. 622 Passive 1 2 3 4 5 6 7 8 9 Dominant
623 Cautious 1 2 3 4 5 6 7 8 9 Daring
The disadvantages of Likert-based in- 624 Bookworm 1 2 3 4 5 6 7 8 9 Social
dexes include susceptibility to: (1) social butterfly
desirability response set bias, the ten- 625 Quiet 1 2 3 4 5 6 7 8 9 Loud
dency to give an answer that is consistent 626 Serious 1 2 3 4 5 6 7 8 9 Humorous
with current social values; (2) extreme re- 627 Conformist 1 2 3 4 5 6 7 8 9 Leader
sponse set bias, the tendency to select an 628 Cooperative 1 2 3 4 5 6 7 8 9 Stubborn
attitude at the extreme; and (3) acquies-
cence response set bias, or the tendency
to always agree with statements regard-
less of their content (Massey, 1995).
where a group would be placed on an hon-
3. Semantic Differential est/dishonest continuum or on a hot/cold
Procedures dimension. Respondents are encouraged
to answer the questions quickly, letting
Osgood (1957) is most associated with the their guards down, and thus revealing how
development of the semantic differential they see various categories of individu-
measurement technique. There are nu- als or objects.
merous index applications for this type of Factor analytic studies (see Chapter 18)
measurement technique in nursing. Origi- of semantic differential indexes suggest
nally, these measurement techniques were that such scales assess three factors of
used to study subjective feelings toward meaning. These include (1) evaluation, (2)
objects or persons. For example, stereo- potency, and (3) activity. Table 13.9 pre-
typing behavior, measuring how respon- sents the pairs of adjectives frequently
dents view various out-groups, has been used to dene these factors. There are a
investigated using this approach. More variety of traditional and nontraditional
recently, the semantic differential index uses for such items. Researchers can com-
has been used in nursing to measure atti- pare participants scores for rating various
tudes toward such constructs as caregiv- constructs or compare various groups of
ing roles, menopause, smoking, risk-taking respondents scores. Questions can be
behavior, computer-assisted learning, self- used to measure individual variables or be
concept, and maternalchild bonding. The combined to create indexes. The items are
format of semantic differential questions is scored similarly to items on a Likert index.
shown in Box 13.5. These questions con- Remember to take into account the direc-
sist of a series of bipolar adjectives, indi- tion of the positivenegative bipolar adjec-
cating two extremes, placed at the margins tive pairs.
of the page; the respondent is asked to in-
dicate where, between the two extremes, 4. Magnitude Estimation
he or she would place the group, individ- Procedures
ual, or object being evaluated by selecting
a value on a 9-point scale. For example, re- Magnitude estimation procedures are
spondents might be asked to indicate useful when comparative judgments are re-
Table 13.9 Anchors for Semantic Differential Factors

Factor
Evaluation Potency Activity
Bipolar adjectives Good/bad Strong/weak Active/passive
Fair/unfair Tense/relaxed Quick/slow
Positive/negative Large/small Severe/lenient
Honest/dishonest Hard/soft Sharp/dull
Successful/unsuccessful Rigid/exible Stimulating/boring
Valuable/worthless
Useful/useless
Source: Adapted from Waltz, C., Strickland, O., and Lenz, E. (1991). Measurement in Nursing Research (2nd ed).
Philadelphia: F.A. Davis.
quired. When these procedures are used, through a list of students, indicating in
a respondent compares the magnitudes each case the amount of popularity each
of a series of stimuli. These techniques has relative to that of the average stu-
emerged out of work done by S.S. Stevens dent. If the respondent thinks that Joan
(1966a, 1966b, 1951), who attempted to is two and a quarter times as popular as
examine the relationship between physi- the average, a value of 225 would be as-
cal stimuli (e.g., the physically measured signed. Alternatively, if the respondent
roughness of different sandpapers) and thinks that Joan has three quarters the
respondents perception of their rough- popularity of the average student, then
ness. This powerful technique is useful a value of 75 would be assigned.
when comparative judgments are re- A second approach is to have respon-
quired. The technique results in ratio dents draw different length lines to indi-
level measurement and should be in the cate their perceptions of the differences
repertoire of the nurse researcher. This between stimulus objects. In this case, a
technique is used best in interview or in standard line is given, and the respon-
group-administered questionnaires. The dent is asked to draw lines relative to the
instructions need to be reviewed carefully standard line. (Sample instructions for
with the respondents, so it is necessary to using the line method are presented in
have a researcher present when the in- Box 13.6.)
strument is administered.
Typically, one of two methods is used
to do magnitude estimations. The rst is a. Tips for Using Magnitude
to have respondents provide numerical Estimation Procedures
estimates. If one were estimating the rel-
Some tips for the utilization of magnitude
ative popularity of different students in a
estimations are:
residence, for example, one might begin
by having respondents compare the pop- Tip 1. Only use magnitude estimations
ularity of each student with that of a stu- when a researcher is present to ex-
dent the researcher has identied as be- plain the method. The mailed ques-
ing about average in popularity. The tionnaire is not a suitable vehicle for
researcher would then assign this aver- using this technique.
age student 100 units of popularity. The Tip 2. Use magnitude estimations when
respondent would be asked to proceed comparative judgments are required.
BOX 13.6 Sample Magnitude Estimation Instructions: Perceptions

of Hospitals
Now we would like to have you rate the vari- tal from your home; Dallas General, Mount
ous local hospitals on a number of dimensions. Saint Vincent Hospital, and St. Marys are about
The way you will do it will be to draw different average; and St. Francis Xavier Hospital is about
lengths of line to indicate how each of the hos- one third as far as the average hospital.
pitals compares with one another. If you draw Now we would like you to give us estimates
a line twice as long as the standard line, it on the following 25 aspects, using the technique
means you think that particular hospital is twice that was described above.
as good as average for the particular dimension
How good are the family room facilities for fam-
being rated. For example, suppose you were
ily members of hospitalized clients?
asked to indicate how far each hospital is from
Standard ____________
your home. You would then draw lines indicat-
Breckenridge
ing the relative distances from your home. You
Dallas General
might draw the following lines:
Mount Royal Hospital
IF THE AVERAGE HOSPITAL IS ____________ Mount St. Vincent Hospital
THIS FAR, HOW FAR IS: (suppose you then St. Francis Xavier Hospital
drew the following lines) St. Marys Hospital
Breckenridge __________________
How good a professional reputation do you
Dallas General ____________
think the nursing staff at each hospital has?
Mount Royal Hospital _______________
Standard ____________
Mount St. Vincent ____________
Breckenridge
St. Francis Xavier ____
Dallas General
St. Marys ____________
Mount Royal Hospital
We would interpret your lines to mean that Mount St. Vincent Hospital
you think that Breckenridge Hospital is about St. Francis Xavier Hospital
1.5 times further than the average local hospi- Saint Marys Hospital
Tip 3. Use a stimulus category some- used, the value 999 is used to indi-
where near the middle of the range you cate an item that has not been
intend to use as a standard. Avoid answered; 001 generally is used to in-
choosing a standard that is near the dicate a zero response. Lines are gen-
extreme high or low. erally measured in millimeters, and
Tip 4. After the standard has been as- numerical estimates are entered di-
signed, leave the respondent free to as- rectly without change.
sign all other values.
Tip 5. Randomize the order of presenta- b. Evaluation of Magnitude
tion and avoid starting with the ex- Estimations
tremes of the continuum.
Tip 6. Before the session begins, tell the The exciting part of this technique is its
respondent how to indicate a zero ability to provide ratio level measurement
response or a nonapplicable one. for perceptions about social and psycho-
Tip 7. Data derived from magnitude esti- logical phenomena of interest to nurses. It
mations generally require three col- is possible, for example, to measure the
umns of space for each variable en- status of occupations and people with dif-
tered. When lines or numbers are ferent incomes and education levels
(Hamblin, 1971); the perceptions of those mediate feedback of numerical measures

who have migrated (Jackson, 1999); the on the back side (Frank-Stromberg and
relative attractiveness of universities to Olsen, 1997).
high school students (Jackson, 1973); the
difculty of adopting selected activities in
meeting a goal (Sennott-Miller and Miller, a. Steps for Constructing and Using a
1986); dyspnea in clients with obstructive Visual Analogue Index
lung disease (Janson-Bjerklie et al., 1986);
and the difculty of caregiving tasks Step 1: A VAS measures only one dimen-
(Waltz et al., 1991). sion of a phenomenon at a time; there-
In collecting data, respondents some- fore, in constructing a scale, the re-
times use numbers incorrectly, rank or- searcher must rst decide which
dering the stimuli rather than maintaining dimension is to be measured with the
the proportionality between the stimulus scale. For example, you may be inter-
items. One has to be careful to make cer- ested in measuring the intensity of pain
tain that respondents understand the pro- experienced by participants. If other
cedure. A further limitation is that a re- dimensions of pain are to be measured
searcher must be present to train the by the researcher, additional visual
respondents in the use of the technique analogue scales will be required for
and, hence, it is not applicable to mailed each dimension.
questionnaires or phone interviews. Step 2: Identify the bipolar anchors that
will appear at each end of the scale.
They are placed beyond each end of
5. Visual Analogue Scales the scale, not above or below the line.
The anchors should reect the entire
Visual analogue scales (VAS) were devel- continuum of responses possibly re-
oped more than 60 years ago to measure lated to the phenomenon of interest.
various subjective phenomena. Today For example, if you are measuring in-
they are gaining increasing popularity in tensity of pain, one anchor may be no
nursing research as an easily administered pain and the other anchor may be
tool appropriate for most clinical popula- the worst possible pain.
tions. VAS measure the intensity of partic- Step 3: Draw the scale. The scale is a line
ipants sensations and feelings about the of 100 mm in length with right angle
strength of their attitudes, beliefs, and stops at each end.
opinions about specic stimuli such as For example:
pain, fatigue, nausea, quality of life, health No pain II Worst possible pain
status, appetite, and self-care ability. VAS Step 4: Some researchers prefer a vertical
requires participants to place an x or format for the scale because it elimi-
mark on a line or linear scale that reects nates problems with leftright discrim-
the intensity of their feelings, beliefs, or ination and is more analogous to the
opinions. Usually a 100-mm line is used more (high, top) and less (low, bot-
with two anchor words or phrases at each tom) ends of the continuum measure
end of the line to describe the endpoints of by a visual analogue scale. The partici-
the scale. The line can be drawn vertically pant in the sitting position should mark
or horizontally. A mechanical format vi- the scale. It is not known what the im-
sual analogue scale is now available. It is pact of reclining in the supine position
similar to a slide rule plastic device that may be on the perception of scale
shows a 100-mm horizontal line on the length for bedridden clients.
front and a movable tab that provides im- Step 5: For children, pictures may be
used at each end of the scale as an- the two extremes of the dimension be-
chors. The pictures should represent ing measured. For example:
Step 6. Printing is preferred to photo- 6. Delphi Technique

copying visual analogue scales be-
cause photocopying the instrument The Rand Corporation (a think tank in
may cause small alterations in the California) originally developed the Del-
length of the line, which may produce phi technique in the 1950s as a means to
invalid data if not corrected. structure group opinion and discussion
Step 7: Instructions should be clear and (Waltz et al., 1991). It has been used in
include a written example of how to nursing research to assess priorities
mark the scale. Place a mark on the such as national research directions, de-
line that indicates the degree of pain termine goals and plans for curriculum
you are having at this moment, is the revision in nursing education programs,
type of instruction you would provide and obtain the opinions of experts on a
to participants. variety of topics without the necessity of
Step 8: To score the visual analogue scale, meeting face to face. It is especially use-
measure the distance of the mark from ful in exploring planning options; identi-
the end of the scale in millimeters. Usu- fying the pros and cons of expert nurses
ally the low end of the scale is used as related to health reform policies; and de-
the zero point. If more than one scorer veloping national agendas around issues
is measuring the points marked, it is such as nursing education, research, or
important to have identical rulers so professional practice roles.
that systematic bias is avoided.
a. Steps in Using the Delphi
b. Evaluation of Visual Analogue Technique
Indexes
Step 1: The rst step is to identify and se-
A major advantage of the VAS is its ease lect a panel of experts on the topic of in-
of administration and its sensitivity to terest to the investigator. The panel
changes in levels of the stimulus. They should be as representative of various
are readily understood by participants demographics and expertise as possi-
and can be completed quickly and easily. ble so that bias is avoided in the mem-
They do not limit participants to a de- bership. For example, if you applied the
ned number of responses but rather en- Delphi technique to determine national
able them to make ne distinctions in re- nursing research priorities, you should
sponses. In this manner, they are more have a panel composed of nurse re-
sensitive than categorical scales. Relia- searchers from across the country, in-
bility and validity are similar to more cluding large urban centers as well as
time-consuming measures. Validity is de- rural areas, nurses from various sec-
termined by using the contrasted groups tors of health care and academia, and
approach or by correlating the VAS nurses with expertise in various re-
scores with other measures of the phe- search methods and interests.
nomenon. Reliability is assessed using Step 2: Next a questionnaire that focuses
testretest procedures. on the phenomenon of interest is de-
veloped and circulated to the expert sensus. It is not advisable to go beyond

participants. Usually the question- ve cycles because of the risk of pan-
naire is administered by mail, inter- elists attrition because of disinterest.
view, or an interactive computer con-
ference line. The respondents do not The Delphi technique is exible, can ac-
discuss issues face to face at any time. commodate a variety of research topics,
The questionnaire is developed ac- and obtains the opinion of experts with-
cording to the guidelines presented in out the nancial cost or the time inconve-
Chapter 14 and contains a set of in- nience of bringing experts together. The
structions for participants. Most Del- opinions of a variety of experts can be
phi questionnaires contain close- concisely condensed into a precise state-
ended questions, but the opportunity ment. Honest opinions are likely because
can easily be provided for open-ended anonymity is protected. It is unlikely that
statements as well as scales. For ex- verbally aggressive experts will unduly in-
ample, in generating research priori- uence the group. The disadvantage is
ties, you may begin by asking respon- that the procedure is costly in terms of
dents to list research questions they multiple data analysis and processing and
consider to be of major importance. In mailing and printing costs. Also, the time
subsequent rounds of the question- requirement for collection of data de-
naire, you may present lists and have pends on busy experts returning ques-
respondents prioritize the items. tionnaires so that subsequent rounds of
Step 3: Returned questionnaires are ana- data collection and analysis can proceed.
lyzed and results tabulated and re- Panel members may become bored and
turned to participants. Anonymity of withdraw before completion of the study
participants responses is protected, al- if data collection does not proceed at a
though the researcher may list names reasonable pace. This creates problem
as part of the study. Descriptive statis- with sample size.
tics, frequencies and measures of cen- Box 13.7 illustrates the use of the Delphi
tral tendency, and dispersion for each technique to predict changes in nursing
item are usually circulated to partici- education. The phases in the question-
pants. Occasionally participants with naire analysis cycle are described. For a
extreme responses may be asked to ex- complete discussion of the procedure and
plain their choices. ndings, consult the original source.
Step 4: After the primary round of ques-
tionnaires, the experts again respond
to a second questionnaire developed 7. Physiological Measures
by the researcher. Participants re-
spond to the new information in the Nurses use physiologic measures on a reg-
questionnaire that has resulted from ular basis to assess the health status and
the statistical analysis of the primary physiological functioning of their clients.
questionnaire and return it to the re- Data such as blood pressure values, tem-
searcher for subsequent analysis. The perature, pulse rate, respiratory rate,
experts can adjust their opinions based height, weight, percent of body fat, muscle
on feedback received after each ques- strength, electrocardiogram and elec-
tionnaire. This procedure is repeated troencephalogram readings, pulmonary
until the data reect a consensus of artery pressure, cardiac output, salivary
opinions or beliefs among the panel. enzyme levels, urinary or serum glucose
The questionnaire-analysis cycle may and ketone levels, and oxygen saturation
take three to ve times to reach con- levels are a few of the many physiologic
THE DELPHI TECHNIQUE: A POSSIBLE TOOL FOR PREDICTING

FUTURE EVENTS IN NURSING EDUCATION
The purpose of this study was to explore the po- Round III: Panelists received feedback from
tential of the Delphi technique in predicting their Round II predictions plus the corre-
events of the next 50 years in nursing education. sponding response from the total group for
Experts were selected on the basis of educa- each statement. If a panelists prediction dif-
tional level, rank in educational institution, or po- fered from the group response, he or she
sition in agency. The selected panel consisted of was requested to revise his or her prediction
16 members representing education, practice, or to support his or her position. The results
the professional association, and government. were again tabulated and reported for each
The Delphi technique included four rounds, each statement. Statements achieving consensus
involving a questionnaire and questionnaire were announced. Reasons for dissenting
analysis, conducted in the following manner. opinions were incorporated into the next
questionnaire.
Round I: Panelists were requested to make a
Round IV: Panelists were asked to reconsider
maximum of 10 predictions regarding the fu-
their predictions in view of the dissenting
ture of nursing education in the next 50
opinions and to revise them if they so desired.
years. A grouping and collation was done to
The additional predictions that achieved con-
reduce the number of predictions to a man-
sensus were identified. A description of the
ageable size.
events that would occur in the future, as pre-
Round II: The predictions were presented and
dicted by the consensus of the panel of ex-
the panelists were asked to predict in which
perts, was composed and sent to the pan-
time interval they would occur. Results were
elists.
tabulated and reported for each statement in
terms of number and predictions in each time For a discussion of the results, interested
interval. readers are referred to the original source.
SOURCE: Summarized from Bramwell, L., and Hykawy, E. (1999). The Delphi technique: A possible tool for predict-
ing future events in nursing education. The Canadian Journal of Nursing Research, 30(4), 4759.
measures that clinical nurses quantify on and scientic equipment to yield biologi-
a regular basis to assess the level of func- cal data.
tioning of their clients. For example, Kang and associates
These measures are of value to nurse (1998) studied the relationship of social
researchers as well. Physiologic measure- support to stress responses and immune
ment is particularly appropriate in studies function in healthy and asthmatic adoles-
designed to assess the impact of nursing cents. They drew blood samples and mea-
interventions on bodily functions. Tradi- sured the natural killer (NK) cell function
tionally, nursing research has tended to in the immune response. Monitoring the
use psychosocial measures because of numbers and types of NK cells in the
problems inherent in physiological mea- circulation is important for evaluating
surement strategies. Often nurses do not changes in immune responses. Results in-
have control over or access to the equip- dicated that high social support appeared
ment or resources available to monitor to attenuate the magnitude of examina-
changes in physiological functioning. The tion-induced reduction in NK cell function
situation is changing, however, and many activity, suggesting that social support
nurse researchers are now engaged in plays a protective role against immune
studies that use physiological measures decrements during times of stress.
Well-designed and appropriately im- 1. The observations undertaken must be

plemented physiological measurement consistent with the studys objectives
devices are among the most precise and aims.
methods of measurement one can use in 2. There is a systematic and standard-
nursing research. They provide objective ized plan for the observation and the
and sensitive measurements that are dif- recording of data.
cult for the participant to distort. They 3. All of the observations are checked
usually yield ratio level data that can ac- and controlled.
commodate a range of statistical proce- 4. The observations are related to scien-
dures. The disadvantages include the ex- tic concepts or theories.
pense, storage requirements, additional
training requirements, and the intrusive- Observations may be structured, semi-
ness created by some devices. Error is structured, or unstructured. They may
also a potential concern because of envi- occur in natural settings or in controlled
ronmental factors such as temperature laboratory settings. In structured obser-
or barometric pressure changes that may vations, the researcher carefully denes
affect readings; different users and pro- what observations are to be made and
cedures that lead to inconsistency in how they are to be recorded and coded
measurements; changes in calibrations of (see Fig. 7.1). Usually a category system is
equipment that affect precision of mea- devised for organizing and sorting the
surement; or carelessness on the part of data. Tally sheets, checklists, and rating
those recording the measurement values. scales are also used. Consistency of ob-
servations and recording is very impor-
8. Observational Measurement tant. Usually interrater reliability is es-
tablished and monitored periodically
Observational measurement is frequently throughout the study if observation is the
conducted in nursing research to deter- primary means of data collection. This is
mine how participants respond in specic important in terms of determining the de-
situations. It is a data collection method gree of condence to place in the data.
that is particularly well suited to phenom- In qualitative investigations, unstruc-
ena that are best viewed from a holistic tured observations are conducted. These
rather than a reductionistic perspective. involve spontaneously observing what
For example, if one is interested in study- occurs in natural settings with little pre-
ing the impact of caring behaviors by planning. Participant observation is a
nurses on separation anxiety in hospital- special type of unstructured observation
ized children, a researcher could elect to used in qualitative investigations. It en-
interview the child or administer a ques- ables the researcher to collect informa-
tionnaire, but direct observation of the tion that otherwise may go unrecorded.
childs behavior would most likely yield Detailed eld notes, logs, and narratives
the best data to comprehensively answer are used to record such data.
the research question. A distinction must Observation is difcult at the best of
be made between the general types of ob- times because the very presence of the ob-
servations that nurses engage in as part of server may change the situation so that
the nurseclient interaction and the scien- one is not certain of what they are observ-
tic observation that is associated with re- ing. Participants may behave in atypical
search measurement. To be scientic, ob- ways that may distort the study ndings. If
servations must meet four conditions as one proceeds to observe participants
outlined by LoBionda-Wood and Haber without informing them about the study,
(1998, p. 313). These include: ethical dilemmas may result. Although ob-
servations tend to be more subjective than recorded; (2) semistructured with specic
other measurement techniques, studies topics focusing the discussion; or (3) non-
using these techniques remain as some of structured, in which questions are not pre-
the best available means to capture the determined but are left to the discretion of
complexity of both the subjective and the the researcher, who frequently takes his or
objective complexity of human behavior. her lead from the respondents cues. Usu-
The advantages include the exibility they ally a broad research question such as,
provide to the researcher in reconceptual- Tell me what your experience was like
izing the problem based on emerging data. with . . . is used to initiate the discussion.
They provide a direct, rst-hand measure- Interviews have the advantage of solic-
ment of events. For a complete discussion iting more in-depth data than would be
of observational measurement, see Chap- available through questionnaires; they
ter 7, which considers both overt and nat- provide the opportunity for probing of re-
uralistic observational eld studies. sponses; and they can be used with par-
ticipants who are not able to read or write.
9. Interviews They are limited by time and cost factors
that make it difcult to use large sample
The interview is the data collection sizes unless adequate nancial and human
method used most frequently by nurses resources are available. When more than
to obtain information from clients and one interviewer is used to collect data, it is
their families. An interview is dened as a important that sufcient time be spent
face-to-face verbal interaction in which training the interviewer in data collection
the researcher attempts to elicit informa- procedures. Inconsistencies in data col-
tion from the respondent, usually through lection procedures and subject selection
direct questioning. Because nurses use bias are always potential threats to the va-
the interview technique so frequently in lidity of ndings.
their practice, the dynamics of interview- If you are planning on using an inter-
ing are well known to them. Similar to sci- view to collect data for your research
entic observation, however, care and project, you are advised to refer to Chap-
precision are required in conducting inter 7 for a discussion of in-depth and fo-
terviews for purposes of research. The in- cus group interviews.
terpersonal skills of the researcher are an
important asset in collecting interview 10. Questionnaires
data. The response rate from interview
studies is usually higher than from ques- The questionnaire is a self-administered
tionnaires. A skilled interviewer can elicit form designed to elicit data from the
rich and accurate descriptions from re- respondent through written or verbal
spondents that would not be possible by responses. Questionnaires have many im-
other means. Because interviews are self- portant uses in nursing research. Chapter
reports, the researcher must assume that 14 provides a comprehensive discussion
the data reported are an accurate account of questionnaire development and admin-
of the phenomenon under investigation. istration.
The interview is classied according to
the degree of standardization implicit in
the interview questions. The interview G. DEVELOPMENT OF AN
may be (1) structured with the researcher INSTRUMENT
exercising a maximum amount of control
by predetermining questions and the Because nursing research is still evolving,
range of response categories that can be nurses frequently encounter the problem
of not being able to locate a measurement form of the ALQ comprised 66 items in
tool for a construct of interest. When this seven categories: physical participa-
occurs, the nurse researcher is faced with tion, nutrition, safety, social support,
the challenge of developing an instrument health awareness, stress management,
or modifying the study. The process of in- and identity awareness. The seven cate-
strument construction is a complex one. gories were considered dimensions of a
It involves the following steps: healthy lifestyle in adolescents. The in-
strument used a ve-point response for-
1. Dene the construct to be measured. mat of 1 never; 2 rarely, 3 some-
2. Create the instrument items. times, 4 often, and 5 almost always
3. Assess the content validity of the items. to obtain an ordinal level of measure-
4. Pretest and pilot test the items. ment.
5. Assess the validity and reliability of Step 3: Assess the items for content va-
the instrument. lidity. Content validity was assessed
6. Publish results so that others can avail by eight nurses with expertise in ado-
themselves of the measurement tool. lescent health promotion. Four were
asked to rate each item using four cri-
Let us consider the development of teria: readability, cultural relevance,
the Adolescent Lifestyle Questionnaire age appropriateness of items, and con-
(ALQ) as an example of each of these ceptual congruence with the construct
steps. Gillis (1997) sought to develop a of a healthy lifestyle. The remaining
valid and reliable instrument to measure four nurses were asked to place the
healthy lifestyles in adolescents. Before items in the seven categories accord-
this, no instrument was available to mea- ing to denitions provided by the re-
sure healthy lifestyles in adolescents. searcher. Items were added, modied,
and deleted based on input from the
Step 1: Dene the construct. Gillis de- panel of nurse experts. The resulting
ned healthy lifestyle as a multidimen- instrument contained 56 items.
sional pattern of discretionary activi- Step 4: Pretest and pilot test the instru-
ties and perceptions that are part of an ment. The pilot version of the instru-
adolescents daily approach to living ment was tested on a sample of 73
and that signicantly affect health sta- school-based adolescents over a 3-
tus in a positive manner. week period for reliability, item clarity,
Step 2: Create the items. Items for the ALQ and response variance. Results of relia-
were developed from qualitative re- bility testing using the testretest pro-
search interviews with 30 adolescents. cedure yielded a coefcient of 0.76 for
Two broad interview questions were the total instrument, indicating stabil-
posed: (1) What does it mean to you as ity. Cronbachs alpha was calculated as
a teen to live a healthy lifestyle? and (2) a measure of internal consistency. The
What kinds of things do teens your age alpha coefcient for the total instru-
do on a regular basis to keep healthy? ment was 0.93 and ranged from 0.60 to
Item development was guided by a re- 0.87 for the subscales.
view of the adolescent and adult health Examination of frequency distribu-
promotion literature and by analysis of tions indicated that the full range of
the qualitative interviews. Factors were responses was used for the majority of
selected that a priori were congruent items. Some items were deleted or re-
with the denition of lifestyle and were worded because of confusion over
sufciently general enough to apply to meaning and terminology. As a whole,
large numbers of adolescents. The pilot the instrument appeared to have suf-
cient reliability to warrant further de- and validity testing of the nal version
velopment. of the ALQ.
Step 5: Estimate reliability and validity. Step 6: Publish the results. Nursings
Empirical validation of the ALQ fol- ability to study phenomena of concern
lowed the process suggested by Nun- to the discipline and thereby advance
nally (1978). This included item analy- the science of nursing is limited by our
sis of the pool of 56 items to determine ability to measure such phenomena.
which contributed most to the internal Hence, it is important that researchers
consistency of the measure; factor who invest considerable time and re-
analysis to dene the factorial compo- sources in instrument development
sition of the item pool and establish va- studies share their results with others.
lidity; and reliability measures to esti- (This study is published in the follow-
mate the internal consistency of the ing source: Gillis, A. (1997). The ado-
nal version of the ALQ. Box 13.8 pro- lescent lifestyle questionnaire: Devel-
vides detailed results of the reliability opment and psychometric testing.

Canadian Journal of Nursing Research, on the care and attention that the re-
29(1), 2946.) searcher pays to appropriate data collec-
Data collection and measurement tion and measurement issues. Determin-
strategies constitute critical aspects of ing what measurement tool to use and
the research process. They mark the tran- conrming that the tool validly and reli-
sition from the theoretical to the empiri- ably reects the construct of interest is
cal phase of the research study. Much of central to doing good science and to ad-
the success of a research project depends vancing nursing.
E X E R C I S E S
1. Would it be possible to have a valid 3. In our culture, teenage males will over-
measure that is not reliable? Could you estimate their weights; teenage females,
have a reliable measure that is not however, slightly underestimate their
valid? weights but show less variability in their
estimates of their true weight.
Discuss this quotation, making clear
the differences between true value,
2. Discuss sources of random and nonrandom error, and systematic error in
random measurement error in reactive measurement.
research designs. To what extent are
such errors caused by different per-
ceptions of the question being posed
or differences in image management? Is 4. Develop and pilot test one of the fol-
image management more likely to lead lowing types of indexes or scales: Lik-
to random or nonrandom errors? ert, semantic differential, visual ana-
logue, or magnitude estimation. Report
fully on the methods you used to de-
velop the scale or index.
Carmines, E.G., and Zeller, R.A. (1979). Relia- searchers can use to prepare for qualita-
bility and Validity Assessment. Beverly Hills: tive data collection.
Sage. This is an excellent introduction to Lodge, M. (1981). Magnitude Scaling: Quantita-
the assessment of reliability and validity. tive Measurement of Opinions. Beverly Hills:
Frank-Stromberg, M., and Olson, S. (1997). Sage. Another in the Sage series; this intro-
Measurement in Nursing Research (2nd ed.). duction to magnitude estimation proce-
Boston: Jones and Bartlett Publishers. An dures neatly summarizes the alternative
excellent resource for both the develop- procedures and reviews the ndings that
ment and use of instruments in nursing re- have resulted.
search and clinical practice. Sandelowski, M. (2000). Combing qualitative
Montgomery, K. (2000). Getting organized: and quantitative sampling, data collection,
Qualitative data collection. Applied Nursing and analysis techniques in mixed method
Research, 13(2), 103104. This article de- studies. Research in Nursing and Health, 23,
scribes an organizational schema re- 246255. This article provides examples of
data collection combinations, including with recent illustrations of the work done
the use of instruments for fuller qualitative using this tradition.
descriptions, for validation, and as elicita- Waltz, C., Strickland, O., and Lenz, E. (1991).
tion device interviews. Measurement in Nursing Research (2nd ed).
Strauss, A., and Corbin, J. (1997). Grounded Philadelphia, PA: F.A. Davis. This compre-
Theory in Practice. Thousand Oaks: CA: hensive text describes the process involved
Sage. This book provides an updating on in selecting, testing, and designing instru-
the development of grounded theory along ments for use in nursing investigations.
Chapter 14
Questionnaire
Development
CHAPTER OUTLINE
A. General Guidelines for Questionnaire D. Ordering Questions, Formatting, and

Development Presentation
B. Types of Questions Illustrated 1. Ordering Questions
1. Precoded, Single-Choice Questions 2. Formatting
2. Open-ended Questions 3. Presentation
3. PresenceAbsence Questions E. Evaluation of Questionnaires
4. Rank-ordering Questions
1. Advantages
5. Likert-type Questions
2. Limitations
6. Index Development
3. Challenges
C. Steps in Developing a Questionnaire
KEY TERMS
Closed-ended questions Pilot study Questionnaires

Likert-type questions Precoded single-choice Rank-ordering questions
Open-ended question questions Response set
Presenceabsence questions
471
A
s discussed in Chapter 5, a ques- Guideline 1. Consult the participant. All
tionnaire consists of an organized questionnaires are an imposition on
set of written questions that are those who are asked to complete them.
presented to respondents for their reply. It is important, if you wish to have a
Questionnaires have many applications; high completion rate, not to impose on
they are used to do interviews, conduct re- your respondents. Above all, respon-
search, and compare outcomes of differ- dents must be made to feel that they are
ent programs among other applications. A being consulted and can freely express
review of the literature suggests that nurs- opinions. The cover letter, question-
ing research has relied heavily on ques- naire, and interview schedule should
tionnaires as a primary means of collect- be designed to be inclusivethat is, de-
ing data for the past 40 years. This trend is signed to make all participants feel that
likely to continue in the new millennium. their opinions are both valued and ac-
A well-designed questionnaire does not ceptable.
impose on the patience of the respondent. Guideline 2. Keep it short. Frequently pro-
It should be possible to move through the jects involve a group of researchers and
questionnaire rapidly, without becoming there may be considerable difculty in
bored and without having to reread ques- keeping the questionnaire from becom-
tions because of ambiguity. An easy-to- ing too long. Asking too many questions
complete questionnaire is more likely to not only is an infringement on the re-
be lled out successfully. For those doing spondents time but also creates addi-
a survey, Chapter 5 contains a discussion tional work in data entry and error de-
of how to administer different types of sur- tection. To overcome this problem, it
veys. Although Chapter 13 discusses is- takes a careful negotiator to persuade a
sues related to validity and reliability of colleague that some of the proposed
questionnaires, this chapter focuses on questions be left for a later study. One
the development of questionnaires. All test when such difculties arise is to re-
questionnaires contain the following com- quest individuals to indicate precisely
ponents: (1) a cover letter, (2) a set of di- how the variable will be used to meet
rections to respondents, (3) a set of ques- the objectives (hypothesis) of the
tions, (4) a set of response categories, and study. Such discussions force some
(5) a demographic section. Each of these careful thinking about the survey, the
elements is addressed in this chapter. We objectives, and the analysis. All too of-
begin our consideration with some gen- ten a number of variables remain unan-
eral issues, then pay attention to the alyzed, either because they are consid-
phrasing of specic questions, and con- ered to be poorly measured or because
clude with issues of layout and format. their relationship to the study is un-
clear.
There cannot be any strict rules for the
A. GENERAL GUIDELINES FOR
length of a questionnaire. It may be use-
QUESTIONNAIRE
ful to develop a guide that identies:
DEVELOPMENT
The variables that need to be mea-
The following section presents a set of sured
general guidelines for developing a ques- The content that is interesting but not
tionnaire. In some situations, the guide- essential to the study
lines should not be followed in detail; The approximate number of questions
simply use them with intelligence and a needed to cover the study variables
good dose of common sense. The maximum number of questions
QUESTIONNAIRE DEVELOPMENT 473
that are practical to ask in the study, that interviews, whether using structured
given who your respondents are and or semistructured questionnaires, take
taking into account their needs and at- much more time because of the tendency
tention spans of respondents to stray from the topic.
An estimate of the time it will take to Questionnaires administered to an assem-
complete the instrument bled collection of individuals can involve
70 or more questions. Here, as long as the
From this guide, researchers can decide
questions are well designed, respondents
to increase or reduce the number of ques-
can move through the instrument rapidly.
tions to truly reect the study variables.
But no matter how the data are to be gath-
If the questions are easy to answer and
ered, the researcher should strive for
the respondents have a particular inter-
brevity and simplicity. There is no point in
est in the survey, it is possible to extend
gathering data that you do not have the
the length of the questionnaire. However,
theoretical or analytical skills to process.
it is always wise to use as few questions
as possible and never ask questions Guideline 3. Achieve precise measure-
merely for interests sake. Table 14.1 sug- ment. Generally, researchers should
gests guidelines for maximum lengths for try to obtain the most precise measure-
typical questionnaires. ment possible. Collect data in the
The table can only be a guide because rawest form, such as income to the dol-
enormous variations may exist in the lar, precise occupation rather than a
complexity of the questions, in the so- general category, and age to the nearest
phistication of the respondents, and in year rather than an age category that
the respondents interest in the survey. spans 10 years. This recommendation
Phone surveys have to be particularly has to be moderated where other fac-
easy to respond to. Each question must tors argue against precise measure-
have simple response options because ment. You might avoid precision when
the respondent will have difculty trying it would entail asking for information
to remember all the response categories that is too personal, require respon-
if many are presented. Keep phone ques- dents to make distinctions beyond
tionnaires simple and short. those they normally use, or when the
Other questionnaires can involve more methods needed to ensure such preci-
questions and greater complexity. Mailed sion would be too cumbersome. Argu-
questionnaires need to be kept somewhat ments in favor of precise measurement
shorter than the others are because the are presented in Chapter 13.
researcher is not present to provide en- When asking about income, for ex-
couragement. Interviews using a question- ample, the question should be clearly
naire can safely be extended to about 60 stated to determine if the income is in-
questions and take as much as an hour dividual income or household income.
to complete. It should be remembered Also, the researcher may consider ask-
Table 14.1 Maximum Number of Questions on a Questionnaire

Type of Questionnaire Maximum Length Maximum Time
Phone questionnaire 20 questions 10 to 20 minutes
Mailed questionnaire 50 questions 15 to 20 minutes
Group administered 70 questions 20 to 30 minutes
Interview 60 questions 60 to 90 minutes
ing the source (e.g., base pay from the 1. Precoded, Single-Choice
primary employer or from extra jobs or Questions
overtime). Some people consider in-
come and age as personal information. In precoded, single-choice questions,
Researchers need to determine the im- the respondent is asked to indicate with
portance of the measurement scale (or- a checkmark which category applies to
dinal versus ratio data) because the him or her. Box 14.1 provides illustra-
measurement scale determines the sta- tions of such questions. Note that only
tistical analysis. Researchers take a the question is numbered; it would look
chance of not getting any response or a too cluttered if the categories are also
nonresponse on those items if the in- numbered. After the category label, a
formation asked appears too personal. dashed line is used so that the respon-
dents eye moves laterally to the check-
off category; the number beside the
square box is the value that will be used
B. TYPES OF QUESTIONS when the data are entered into the com-
ILLUSTRATED puter. To avoid a cluttered look, there
should be no space between the number
This section illustrates a variety of for- and the square box.
mats for typical questionnaire items. Al- The next items in Box 14.1 illustrate
though it is not possible to anticipate all slightly more complex forms of the check-
types of questions, the same principles off question. Question 5 illustrates how to
can be applied to many questions. (A accommodate two columns of check-off
compendium of questionnaires designed categories; in this case, the questionnaire
by students in our research methods provides one for the mothers and one for
course is available for review on the In- the fathers educational level. Once again,
ternet at the following address: http// the computer codes are placed next to the
www.stfx.ca/people/wjackson). categories because it facilitates data en-
BOX 14.1 The Simple Precoded Question
4. What year of school are you in? 11. What was the approximate population of
Freshman ------------------1 your home area before you began college?
Sophomore ---------------2
Junior ----------------------3 Rural area ----------------01
Senior ----------------------4 Small town under 999 ----------------02
5.1/5.2 What is the highest education com- Between 10004999 ----------------03
pleted by your mother and father? Between 50009,999 ----------------04
Between 10,00019,999 ----------------05
Mother Father Between 20,00029,999 ----------------06
Grades 06 -------------1 -------------1 Between 30,00049,999 ----------------07
Grades 79 -------------2 -------------2 Between 50,00074,999 ----------------08
Grades 1012 -------------3 -------------3 Between 75,00099,999 ----------------09
Some post- -------------4 -------------4 Between 100,000249,999 ----------------10
secondary Between 250,000999,999 ----------------11
University -------------5 -------------5 More than 1,000,000 ----------------12
graduate
try. Similar space saving can be achieved BOX 14.2 Two Versions of
by splitting long category lists into two Population of My
and placing them side by side. Home Town
The question on population size sim-
ply illustrates that when there are more Flawed Version: Categories Are Not
than nine categories, it is important to Mutually Exclusive
place the leading zero in the computer Example 1. The population of the place I
codes. If the leading zero is omitted, data considered my hometown when growing
entry errors will almost certainly occur. up was:
The particular population categories Rural area ---------------------1
would, of course, have to be altered, de- Town up to 5000 ---------------------2
pending on the target population. 5000 to 20,000 ---------------------3
Two points also need to be mentioned 20,000 to 100,000 ---------------------4
that are not illustrated in Box 14.1. Fre- 100,000 to 1,000,000 ---------------------5
quently, it is not possible to name all the 1,000,000 or over ---------------------6
possible responses that would be appro- Better Version: Categories Are Mutually
priate (e.g., religious afliation) and the Exclusive
researcher will have a nal category that Example 2. The population of the place I
says: considered my hometown when growing
up was:
Other - 12
Please specify ________________. Rural area ----------------------1
Town under 5000 ----------------------2
The most obvious reason for using a 5000 to 19,999 ----------------------3
please specify category is that there 20,000 to 99,999 ----------------------4
would not be sufcient room to list all 100,000 to 999,999 ----------------------5
possible religions. However, even more 1,000,000 or over ----------------------6
importantly, you do not wish to insult
your respondent by not having included
his or her religion. In actual fact, if the ma-
jor religious groupings have been in- tions is they ensure comparability of
cluded, the person who species a cate- responses and facilitate analysis of re-
gory not included in the questionnaire sponses. They are particularly easy to
will simply be coded as an Other. administer, but this advantage must be
Finally, please note that the response balanced with the time-consuming chal-
categories provided should cover the full lenge of questionnaire construction.
spectrum of possible responses and it
should not be possible for a respondent
to check off two of the categories. Note in 2. Open-Ended Questions
Box 14.1 that the population categories
An open-ended question asks the re-
are mutually exclusiveno category over-
spondent to answer some question or to
laps another one. A common error is il-
offer a suggestion or opinion, but to do so
lustrated in Example 1, Box 14.2.
without any preset categories being pro-
Example 1 illustrates a situation in
vided for the answer. There are at least six
which the researcher has failed to pro-
reasons for including some open-ended
vide mutually exclusive categories; Ex-
questions in a questionnaire: open-ended
ample 2 corrects the error.
questions are preferred if:
Precoded, single-choice questions are
also referred to as closed-ended ques- The responses are too numerous (e.g.,
tions. The major advantage of such ques- year of birth).
The researcher does not wish to impose BOX 14.3 Sample Open-Ended
response categories on the respondent. Questions
The researcher wishes to create the
sense that the respondent is really be- 20. Approximately, what was your average
ing consulted, by being asked to offer in your final year of high school? _ _ %
his or her opinions. 21. What is (or was) your fathers occupa-
tion? (e.g., foreman, railway machine
The researcher wishes to provide a
shop . . . supervises work of about 25
qualitative dimension to the study
people)?
open-ended responses are often used
Job ____________________________
as a source of quotations for the nal Brief job description ______________
report. _______________________________
The study is being done as a pilot and _______________________________
the appropriate response categories _______________________________
have not been determined. 22. In what year were you born? 19 _ _ .
You wish to change the pace or format 23. What is the one thing that you would
for the reader. like to see changed at the university
health and counseling center?
Although open-ended data are not al- _______________________________
ways analyzed when writing the nal re- _______________________________
port on a project, they can provide in- _______________________________
sights to the researcher that might be 24. In your opinion, what is the single best
thing about the health and counseling
missed if such questions were not asked.
center?
Box 14.3 illustrates some variations in
_______________________________
typical formats. Questions 20 and 22 are _______________________________
questions where the space for the re- _______________________________
sponse indicates that two numbers are ex-
pected. By providing two blanks, the re-
spondent is being prompted to enter two
numbers. The percent symbol in question
20 also helps to indicate exactly what is ex- quired, the researcher must be careful to
pected. This helps to prevent frivolous request it.
replies, such as writing in average in the Questions 23 and 24 simply seek the
case of question 20, which asks about the opinion of the respondent on two issues.
respondents average in the nal year of Typically, the responses that are given
high school, or writing in a long time ago would either be listed (simply typed) or
in the case of question 22, which asks coded. If they are coded, the categories
about year of birth. would be determined after the data have
Question 21 illustrates a method of ask- been collected and the responses exam-
ing about occupation. The additional line ined.
(Brief Job Description) is included to en- A good rule to follow is to minimize the
sure that the respondent provides suf- number of open-ended questions. Many
cient detail to enable the researcher to at- researchers minimize the number of
tach an occupational rating code to the opinion-seeking, open-ended questions
response. If such specication is not re- because they are time consuming to
quested, some respondents will simply in- code; tend to generate responses that
dicate the employer (e.g., writing in Mi- are inconsistent; and are more likely to
crosoft) and the researcher will not be be left blank. Indeed, respondents fre-
able to attach an occupational rating to quently fail to complete a questionnaire
that kind of response. When detail is re- that has too many such questions. Many
respondents appear to believe that ask- list do or do not apply to them. Box 14.4
ing them to write a sentence or two is too provides examples of such questions. Of
much of an imposition. the two versions presented, the second
When a pilot study is conducted and one is preferable. In it, either a yes or a
the research team is uncertain about the no is expected for each item. Although
appropriate response categories, it is a this is a little more work for the respon-
good idea to pose the question in an dent, the researcher can then be more
open-ended form. After you analyze the condent that each item has been con-
results, you can then base the categories sidered. In question 24, if an item is left
to be used in the nal study on those sug- blank, does that mean that the respon-
gested in the pilot study. dent is opposed to eating that item? Or
Placing an open-ended question at did the respondent simply fail to con-
about the two-thirds mark of a long ques- sider that item?
tionnaire may well provide the relief In contrast to the questions presented
needed to sustain the respondents inter- in Box 14.1, in which the respondent was
est and ensure completion of the ques- asked to check one of a set number of an-
tionnaire. If both open-ended and xed- swers, square boxes are not used for the
choice questions are asked, it is advisable
to place the open-ended version rst so
that the respondent is not inuenced by
the xed-choice options. BOX 14.4 Presence/Absence
Michael D. Smith (n.d) has explored the Check-Off Questions
effectiveness of open versus precoded
Flawed Version: Possible Ambiguity of
questions in revealing the incidence of Items Left Blank
physical abuse. Both formats were
24. On a day you consider routine, which of
equally effective in detecting abuse; how-
the following food choices would you
ever, it was noted that if questions about include in your diet for a 24-hour pe-
abuse were asked a second time, some 21 riod? (Check as many as appropriate.)
percent of the victims revealed their vic-
Dairy products _________1
timization on the second question. The
Meat products _________1
ndings suggest that, with questions of a
Whole grain products _________1
personal nature, it is a good idea to repeat Fruit products _________1
a question a second time, using a variant. Vegetable products _________1
Some may object that such persistent
Better Format: Less Ambiguity About Items
questioning may create an expectation Left Blank
(demand characteristic) that the respon-
23. Have you ever had contact with handi-
dent should make a positive response.
capped people in any of these groups?
However, if the question is properly (Circle to indicate yes or no for each
worded this should not be a problem. group.)
Use opinion-seeking, open-ended ques- Yes No
tions sparingly, but keep in mind that they Community 1 0
are an excellent vehicle for providing a Family 1 0
change in pace for the respondent or for Relatives 1 0
exploring new issues in detail. Elementary school class 1 0
Junior high school class 1 0
Senior high school class 1 0
3. PresenceAbsence Questions University class 1 0
As coworker 1 0
Presenceabsence questions request re-
spondents to check off which items in a
answer in Box 14.4. Instead, respondents spondents, and they should be kept as
are asked to circle a 1 (for yes) or a 0 (for simple as possible.
no). Readers should also note that the Second, it should be noted that the in-
computer codes used involve 1 for pres- structions are embarrassingly explicit.
ence and 0 for absence. During analysis, Although rank ordering may be an ob-
a total experience variable may be cre- vious and simple idea for many people,
ated by simply adding the score on each the detailed instructions will minimize
item together. A total score of 5, for ex- the number of respondents who will sim-
ample, would mean that the individual ply place a checkmark beside three items
has had experience with disabled people or check just one of the items. It is sug-
in ve of the settings identied in the gested that you use bold type for rank
question. order with a larger size font.
Note that the respondents are given a
short line on which they are to write their
4. Rank-ordering Questions answers. Again, it is important to provide
a different look for the item in order to
Rank-ordering questions are those in
cue the respondents that this is not a
which a respondent is asked to indicate
check-one question. Be assured that if
an ordering of response items, usually
you use a check-one format, a good
from most preferred to least preferred.
number of questionnaires will be re-
Asking respondents to rank order a list
turned with one of the items checked.
has to be done with great care. In this
One suggestion for coding these re-
case, detailed instructions should be pro-
sponses into the computer is to use the
vided for the respondent. Box 14.5 in-
values provided by the respondents ( i.e.,
cludes an example of such a question.
1 1, 2 2, and 3 3) and code all items
First, note that respondents are only
left blank as a 4. In this way, means can be
asked to pick out the three most impor-
computed during analysis. (This is a situ-
tant items; in most cases, respondents
ation in which it seems to make sense to
will not be able to accurately pick more
compute a mean using ordinal data.) In
that the top three (Jackson, 1999). These
the few cases in which respondents sim-
are difcult types of questions for re-
ply tick off three items, each one ticked
can be given a 2. Of course, if the whole
item is left blank, the researcher is forced
BOX 14.5 Rank-ordering to assign the missing value code to each
Questions of the items.
Avoid overusing rank-ordering items
31. RANK ORDER the three most im- in questionnaires. They slow down re-
portant things you want in your nursing spondents and increase the risk of losing
job. (Place a 1 beside the most impor- the respondents cooperation.
tant one; a 2 beside the next important
one; and a 3 beside the next most im-
portant one.)
5. Likert-type Questions
High salary _____
Satisfaction _____ Chapter 13 presented Likert-type ques-
Continued interest _____ tions under the discussion on index con-
Power _____ struction. Readers are referred to that
Prestige _____
chapter for a more extensive discussion
Excitement _____
of these items. But because they are used
as stand-alone items, a brief comment
will be made on them here. Likert-type nal format. Given a preference to increase
questions are those that ask respondents the variability in such items, the number
to indicate the strength of their agree- of response categories has been increased
ment or disagreement with a statement. from ve to nine (see Chapter 13 and Jack-
This type of question is widely used in so- son, 1999). This increase does not take
cial science research. In the original for- more space on the questionnaire or more
mat (Likert, 1931), respondents are asked space when coded into the computer.
to react to a statement by indicating The following tips may be helpful in
whether they (1) strongly disagree, (2) constructing such items:
disagree, (3) are undecided or neutral,
(4) agree, or (5) strongly agree. Such Tip 1. Avoid the word and in such items
items were, and continue to be, popular if its usage makes the item multidi-
in measuring both matters of fact and at- mensional.
titudinal issues. Box 14.6 provides some Tip 2. Place the Strongly agree answer
examples of these questions. on the right-hand side of the scale, with
Readers should note that the items in 9 indicating strong agreement. Some re-
Box 14.6 deviate somewhat from the origi- searchers prefer to vary the response
categories, such as reversing the side
on which the agree or disagree labels
are placed. This is done to prevent re-
BOX 14.6 Likert-type Items
sponse setthat is, a situation in which
In the following items, circle a number to in- the respondent tends to answer simi-
dicate the extent to which you agree or dis- larly to all items. Although switching
agree with each statement. the side on which the agree and dis-
agree categories are placed may reduce
52. I believe self-confidence represents the
the tendency to respond in a set man-
most effective determinant of healthy
lifestyle. ner, it may also introduce additional er-
Strongly disagree 1 2 3 4 5 6 7 8 9 Strongly agree rors in response. (Respondents may
not notice that you have switched the
53. I believe alcoholics can be rehabilitated agree and disagree categories.) Our
to lead healthy lifestyles.
preference is to maintain a uniform pre-
sentation. Response set is best avoided
54. I believe daily exercise is one of the by wording some questions positively
most effective ways of lowering blood and wording others negatively.
pressure. Tip 3. Avoid negatives that may confuse
respondents. For example, statements
55. I believe a positive attitude is important such as: I dont think the hospital ad-
in shaping a healthy lifestyle. ministration is doing a bad job will al-
Strongly disagree 1 2 3 4 5 6 7 8 9 Strongly agree most certainly confuse and slow down
56. I believe a mothers lifestyle is a pow- respondents. Use direct statements
erful determinant of lifestyle choices in such as: The administration at my hos-
children. pital is terrible: or The administration
Strongly disagree 1 2 3 4 5 6 7 8 9 Strongly agree at my hospital is outstanding.
Tip 4. Select a strength of wording to pro-
57. I believe a fathers lifestyle is a power-
duce variation in response. For exam-
ful determinant of lifestyle choices in
children. ple, if you were asking patients about
Strongly disagree 1 2 3 4 5 6 7 8 9 Strongly agree the quality of care given by nurses in a
hospital, you might ask several ques-
tions, varying the strength of the word-
ing in each item. You might well suspect sent an attempt to summarize, in one
that nursing care will be rated high, so score, a measure of a variable. Indexes
you might wish to strengthen the word- are constructed when we are dealing with
ing from: a single-dimension variable, but in con-
Version 1: The nursing care I received at texts in which one question may not
St. Marthas was good. measure the variable adequately. Indexes
can be constructed by combining a num-
ber of similarly formatted questions or
combinations of questions with different
To
formats. In all cases, the indicators are
Version 2: The nursing care I received at combined and possibly weighted, so that
St. Marthas was perfect in every in- they add up to one index score. (See
stance. Chapter 13 for details on different meth-
ods of constructing indexes. Appendix A
contains illustrations for index construc-
The strengthening of the wording in tion using the Statistical Package for the
version 2 will almost certainly produce Social Sciences [SPSS].)
more variation in response. Version 1
would be expected to produce 8s and
9s almost exclusively. (A further illus-
C. STEPS IN DEVELOPING A
tration is provided in Rule 7 below.)
QUESTIONNAIRE
Tip 5. If there is uncertainty about where
responses will fall, use more than one
This section reviews the steps that need to
item with different intensities in the
be taken in developing your questionnaire.
wordings.
Tip 6. Before the rst Likert-type item, Step 1. Make a list of variables. A com-
provide a brief explanation of how re- prehensive questionnaire designed to
spondents are to indicate their an- reect the variables of a study can be
swers. thought of as consisting of four major
Tip 7. When respondents fail to answer a groupings:
question, use the 0 value as a missing The background characteristics or
value code. (The 9, which is usually demographics
used for missing values, reects a real The dependent variable or variables
value in the case of a 9-point Likert The independent variables
item.) The other types of variables; that is,
intervening, antecedent, and sources
Likert items result in an ordinal level of
of spuriousness variables
measurement. Such items are frequently
Questions to measure variables in
combined to form indexes by adding to-
each of these groups have to be devel-
gether the values on individual items (af-
oped. Some researchers prefer to use
ter having reversed the scores on the
four or more separate questionnaires
negative measures) and researchers con-
for each separate grouping rather than
ventionally treat such scores as ratio
combine items into one major instru-
level data (see Chapter 13).
ment.
The rst task is to create a list of vari-
6. Index Development ables; during this stage it is not neces-
sary to worry about how the variables
Indexes are constructed by combining will be measured; a list is all that is re-
several individual questions. They repre- quired. The list may be derived by:
Applying relevant theoretical mod- amine relationships are constrained

els and identifying variables that by the level of measurement attained
should apply. in operationalizing each of the vari-
Reviewing the literature, paying ables. This discussion may also in-
particular attention to which vari- clude how the analysis will proceed,
ables were measured by other re- given different outcomes of prelimi-
searchers. nary runs on the data. Important new
Examining other questionnaires for variables frequently emerge as a result
ideas as to which variables should of this process. Table 14.2 indicates
be included. the types of procedures that would be
Reviewing the causal models that appropriate given different levels of
have been developed for the cur- measurement of the variables involved
rent project. in the study. The procedures outlined
Thinking about which variables in the table are explored in greater de-
make sense, given the topic of the tail in Chapter 16.
research. Suppose that you wished to do re-
Step 2. Anticipate how data will be ana- gression analyses on the data. In such
lyzed. After the list of variables is de- cases, you would prefer to have ratio
veloped, it is important to discuss level measurement of all the variables.
what statistical procedures will be You might want to gure out how to use
used to analyze the data when they are magnitude estimation procedures or
collected. Indeed, it is possible to set how to construct indexes that would
up proposed analysis tables showing meetor be close to meetingthe
the relationships that are to be ana- standard for ratio measurement. On the
lyzed and indicating the procedures other hand, if the nature of the depen-
that will be used to examine each of dent variable (nominal measurement)
them. It is critical to have some under- dictates a reliance on cross-tabular
standing of how the analysis is to pro- analysis, then you would be content to
ceed because the methods used to ex- get ordinal and nominal level measure-
Table 14.2 Appropriate Analysis Procedures by Levels of Measurement

INDEPENDENT VARIABLE
DEPENDENT Nominal Ordinal Ratio
Nominal CROSSTABS CROSSTABS CROSSTABS MEANS*
DISCRIMINANT
SPEARMAN R SPEARMAN R
DISCRIMINANT
Ratio/Interval MEANS MEANS CORRELATION
ANOVA ANOVA GRAPH
T TEST T TEST PARTIAL CORRELATION
MANOVA
REGRESSION
*In SPSS for this case, run the independent variable as though it were the dependent variable; the interpretation of the
test of signicance would be standard.
ments of your variables. The majority of positive and negative characteristics of

questionnaire items tend to be mea- questionnaire items.
sured at the nominal or ordinal level. Step 4. Double-check to make certain
Questionnaire items often address a you have all the variables. It is essen-
wide variety of topics. Because of this, tial to ensure that all the variables that
researchers frequently analyze indi- play a part in the hypotheses and mod-
vidual items, rather than summing els related to the project have been in-
items and getting a total score. Likert cluded. The list of variables must be
scales and summated rating scales are checked and double checked. It is not
exceptions to this. unusual for a researcher to discover
Step 3. Write the proposed questions on that a key variable has become lost in
index cards. During the development of the shufe. With checking completed,
a questionnaire, it is a good idea to it is now time to begin drafting the
write proposed questions on index questions themselves. Once again,
cards (usually 3 5 inch cards are large there are some rules to keep in mind as
enough). Using cards facilitates both this process begins.
the quick editing of items and rearrang- Step 5. Review conceptual denitions.
ing their location in the questionnaire. When developing the wording for a par-
The researcher should develop a check- ticular question, it is important to review
list of positive and negative attributes of the conceptual denition of the con-
questionnaire items and check each struct that is being measured. Knowing
question against the list. The placement the conceptual denition provides an
of items on the index cards facilitates important guide as to what the question
this checking. Box 14.7 illustrates the is to measure. To illustrate, suppose one
BOX 14.7 Positive and Negative Characteristics of Questionnaire Items
Positive Characteristics
Sensitive questions: Do not phrase the
Clarity: Questionnaire item has only one question in a manner that angers or emo-
meaning. tionally upsets the respondent or family
Brevity: Question is as brief as possible yet members.
it retains its clear meaning. Jargon-filled items: Does the question use
Simplicity: Vocabulary in questionnaire jargon, technical terms, professional litera-
items is no higher than a grade seven level. ture, or slang with which respondents are
Applicability: Respondents have the ability not familiar.
to answer the questions appropriately. Complex questions: Does the question
Relevance: Items are congruent with and use long phrases or complex clauses that are
reflective of the variable to be measured. not understood by the general public?
Negative Characteristics
Biased questions: Does the question lead
the respondent in one direction to answer
Double-barreled questions: Does the the item? Does the item contain emotionally
question ask two questions in one and have charged words such as abortion?
a hidden premise? Inappropriate questions: The respon-
Double negatives: Avoid use of words dent does not have sufficient information to
such as dont and not. respond to the item.
SOURCE: Adapted from Seaman, C.H. (1987). Research Methods: Principles, Practice, and Theory for Nursing. Los Altos,
CA: Appleton Lange.
is attempting to measure socioeco- provide invaluable and crucial guides

nomic status. And suppose that, in the to valid measurement.
conceptual denition of socioeconomic Step 6. Develop wording for questions.
status, reference is made to a hierarchi- There are a few tips to keep in mind as
cal continuum of respect and prestige. you commence writing drafts of the
Given this conceptual denition, the re- questions to be included in your ques-
searcher might be led to adopt a mea- tionnaire.
sure particularly designed to reect the Tip 1. Words must be understood. All
relative amount of prestige an individual respondents must understand the
has. In this case, one might opt to use an words used in questionnaires. Err on
occupational prestige scale such as the the side of simplicity. If a survey is be-
one developed by Featherman and ing done on high school students, play
Stevens in the United States (1982) or Pi- it safe and use words that a student in
neo and Porter in Canada (1967). Al- grade 7 can handle. Most word pro-
though this procedure might yield a cessing programs can assess the read-
valid measure of the concept in general, ing level of your questionnaire. Fur-
experienced researchers would realize thermore, the words selected should
that the measure may be weak on the be those that have but one unam-
respect aspect of the denition. It biguous meaning. Showing off an im-
would also run into some serious dif- pressive vocabulary has no place in a
culties if there were a number of house- questionnaire that is intended to pro-
wives or single stay-at-home parents in duce valid and reliable data. At the
the sample (such individuals are not as- same time, using slang or jargon that
signed scores in occupational prestige is unfamiliar to respondents may be
scales). Therefore, one might choose to offending and lead to a refusal to an-
measure the respect and prestige an swer the question. For example, if you
individual has by getting a number of in- are interested in determining the
dividuals to rate each individual com- amount of drug use by respondents, it
pared with others. Perhaps the latter ap- is better to ask, Approximately how
proach would yield the most valid often do you take sedatives or tran-
measure of the concept even though it quilizers per week? rather than, Ap-
would be rare for a research design proximately how often do you get high
(probably a study of a particular group) on drugs per week?
to allow for the practical use of this Tip 2. Pay attention to the and alert.
method. Individual questions should be unidi-
Suppose, alternatively, that the pro- mensional. Avoid the trap involved in
posed conceptual denition of socio- a question such as:
economic status stressed variations in
My lifestyle choices are similar to
access to scarce resources. In this
those of my mother and father.
case, one might attempt to reect the
variable by obtaining an indication of
the individuals total income. Here the The problem with the question, as
assumption is that an individuals total worded, is that some respondents
income would be a good reection of will not make choices that are similar
that individuals ability to buy access to both their mothers and fathers
to scarce items in the society. choices. As a result, some respon-
The more general point is that re- dents may indicate a 1 or a 2 if they
searchers must pay careful attention do not make choices similar to both;
to conceptual denitions because they others might indicate a 6 or 7 as a
kind of average of the relationship statement such as, Women play a

between their choices and those more important role in shaping fam-
lifestyle choices of their mother and ily health than men, or Women play
father. The point is that two dimen- the single most important role in
sions have been introducedthe shaping the health of family mem-
lifestyle choices of the mother and bers. In both examples, we have
the lifestyle choices of the father. made it less likely that all respon-
Such questions should be divided dents will remain at the Strongly
into two separate questions. It is al- agree end of the continuum.
ways useful to scan ones questions In dealing with a variable such as
to eliminate such unwanted confu- job satisfaction in nursing, a review of
sion; double check any question that the literature will reveal that most re-
has the word and in it. spondents will report themselves to
Similar to the and alert, avoid be relatively satised with their jobs.
double-barreled questions that ask Knowing this, researchers strive to
two questions in one statement. They identify items that will induce some
are recognized by words such as ei- participants to report that they are
ther/or, and/both, and therefore. less than fully satisedperhaps the
Tip 3. Vary wording to produce vari- researcher will ask if the respondent
ability. It is important that the items would like her or his child to have a
in a questionnaire produce variabil- job similar to their own or ask if they
ity in response. If most respondents think they are paid the right amount
provide similar responses to a ques- for the responsibilities that they have.
tion, then the question will have lit- Similarly, a question about the sat-
tle use during analysis. Try to ensure isfaction of clients with services in an
that the respondents will scatter acute care facility will need to ensure
themselves across the response con- that the questions ferret out the slight-
tinuum. A simple example will illus- est dissatisfaction. Otherwise it is
trate the point: likely that virtually everyone will rate
the service as good or excellent.
Women play an important role
When one is phrasing questions so
in shaping the health of family
as to ensure that respondents vary
members.
in their scores, one is not attempting
to distort realityto show dissatis-
Using common sense alone, one faction when none is present; rather,
could anticipate that most respon- one is attempting to develop mea-
dents will be in strong agreement sures that are highly sensitive. And if
with this statement and will circle 8 one is attempting to understand
or 9. If, indeed, respondents do not what leads clients to be relatively
vary their responses much, the ques- more, or relatively less, satised
tion will not prove useful in discrimi- with their treatment in a health care
nating between respondents views facility, one would need highly sensi-
about the role of women. In this case, tive measures, given the fact that
what one does to produce more vari- participants generally tend to re-
ability in response, is to strengthen spond positively. If all participants
the wording of the item, so that more simply reported that they were sat-
respondents will move toward the ised, we would not be able to iden-
disagree end of the continuum. For tify what factors inuence levels of
example, the researcher could use a client satisfaction.
Tip 4. Avoid complexity. Try to keep education or occupational prestige,

questions simple; avoid asking par- might be considered preferable to in-
ticipants to do difcult tasks. When it come itselfsimply to avoid prying
is necessary, for example, to have unnecessarily into what may be per-
participants rank order a list, it is usu- ceived to be a highly personal matter.
ally best to ask them to rank order the Show acceptance of socially unde-
three most important items rather sirable behavior and attitudes by us-
than asking them to go through the ing strategic leads such as Many
whole list. In most cases, the slight re- adolescents have suicidal tenden-
duction in discriminatory power will cies. Have you ever thought about
be offset by a higher response rate to suicide? This is more likely to get an
the question. honest response than if you bluntly
Tip 5. Use existing wordings for com- ask how many times the respondent
parative analysis. When a researcher has thought about suicide.
wishes to compare data that he or Tip 7. Avoid asking participants to
she is generating to data reported by speculate on why they act the way
other researchers, it is important that they do. Generally we are not inter-
the wording of questions be identical. ested in polling participants opinions
Here one has to weigh the advantages on whether certain relationships ex-
of improving the wording of a ques- ist. For example, if you were studying
tion against the advantage of main- the relationship between participa-
taining identical wordings to facilitate tion in regular physical activity and
comparative analyses. levels of self-esteem, you would nor-
Tip 6. Take the edge off sensitive ques- mally not ask participants if they think
tions. Using a combination of experi- there is a relationship between these
ence and common sense, the survey two variables. Instead, the researcher
researcher soon learns that there are would get separate measures of the
some issues that respondents may be two variables (among others) and
reluctant to report. Illegal activities, would then analyze the data to deter-
evaluating friends or neighbors, and mine if there is any relationship.
indicating age or income can all be Tip 8. Be precise and highly specic
sensitive issues. By asking for the re- when choosing wording. If you wish
spondents year of birth rather than to measure, for example, how much
age, the researcher makes the ques- people smoke, ask questions that
tion sound scientic and more likely pinpoint the number of cigarettes
to be answered; if age is asked, the re- smoked in a given period and the oc-
spondent may feel some invasion of casions when they light up. Here
privacy is occurring; the same with in- the time period will have to be rea-
come. Here many researchers will ask sonable. It is silly to ask how many
respondents to indicate into which cigarettes someone has had in the
broad category of income they fall. Al- past 2 years; few people could make
ternatively, if income is a key question a reasonable estimate without con-
and if the question is located among siderable thought. Box 14.8 illus-
those dealing with conditions of work, trates some possibilities in asking
respondents will generally provide about smoking patterns.
the information. However, if the in- Question 1 suffers because of vari-
come question is being used as a mea- ations in what respondents consider
sure of socioeconomic status, then light, moderate, or heavy smoking
other indicators, such as years of behavior. Question 2 is better be-
BOX 14.8 Sample Smoking questionnaire, they may inquire,

Behavior Questions When you ask about the size of my
family, do you mean the family I was
Flawed Version: Fuzzy born into or the one I have with my
1. How would you describe your smoking husband? Also in that question, do
behavior? you want me to include the parents
in the count? If so, clearly you will
Abstainer ----------------0
need to change your wording of the
Light ----------------1
question. With experience, you will
Moderate ----------------2
Heavy ----------------3 learn to be highly specic in your
wording of questions. The goal is to
Flawed Version: Too Specic; Recall
minimize variations in participants
Problem
understanding of each question. It is
2. On Christmas Eve 2000, how many cig-
always a good idea to review your
arettes did you smoke?
questionnaire and search for possi-
_________ cigarettes
ble ambiguities. The goal, although
Better Version: Specic
actually impossible to achieve, is to
3. In the past 7 days, how many cigarettes have all participants understand
did you smoke? each and every question in an identi-
None ----------------0 cal manner.
1to 15 cigarettes ----------------1 Tip 10. If resources allow, do a pilot
16 to 32 cigarettes ----------------2 study. It is recommendable that re-
33 to 66 cigarettes ----------------3 searchers use pilot studies in a num-
67 or more ----------------4 ber of situations. A pilot study in-
volves having a small sample of
respondents complete the question-
naire or undergo the interview. Pilot
cause it pins down a specic time pe- studies are used to determine items
riod, but perhaps it goes overboard; to be included in indexes and to de-
it is possible to become too specic. termine, from open-ended ques-
Question 3 offers a reasonable com- tions, which categories should be
promise because it is specic and used in a xed-choice format. Fur-
refers to a reasonable and recent ther renements in questionnaires
time period. may be achieved through the use of
Tip 9. Pretest the questionnaire. It is a pilot study.
important to have a few individuals
complete the questionnaire or the in-
terview before settling on the nal D. ORDERING QUESTIONS,
wordings of the questions. Start by FORMATTING, AND
lling it out yourself. Make any nec- PRESENTATION
essary corrections and then adminis-
ter it to a few other people. Generally 1. Ordering Questions
it is best to sit with the individuals
completing the questionnaire. Before Rule 1. Introduce the questionnaire to re-
they begin, tell them exactly what spondents. Normally, questionnaires
you are doing (try to remove any am- contain a brief statement introducing
biguity in the questionnaire) and en- the study to the respondent. It should
courage them to ask for any clarica- be short (usually three or four lines is
tions. Perhaps, as they complete the sufcient) and should inform the re-
spondent about who is doing the study, ductory statement on a questionnaire

who is sponsoring it, and what the and the letter of introduction do not re-
study is about. These few lines should place an informed consent form. The re-
attempt to establish the legitimacy of searcher must provide evidence that re-
the project. By identifying who is doing spondents have given free and informed
the study and who is sponsoring it, the consent to participate in the study. This
idea is conveyed that the survey is im- evidence is ordinarily obtained in writ-
portant. Such identication also pro- ing. Figure 14.1 includes an informed
vides the respondent with some addi- consent form that should be signed by
tional information that he or she can respondents as evidence of their con-
use to decide whether to ll out the sent. (For a review of protecting sub-
questionnaire. The researcher should jects rights, see Chapter 10.)
not identify the specic hypotheses of Rule 2. Ease them into it. In deciding the
the study; to do so might well bias the order of questions, consideration
responses. Finally, if the survey is should be given to starting with ques-
anonymous, respondents should be as- tions that are easy to answer. It is im-
sured that their anonymity will be pro- portant not to start by asking ques-
tected. A good way to achieve the latter tions that may be regarded as too
goal is simply to ask respondents not to personal. Background information
write their names on the forms (see Box questions are usually placed at the end
14.9 for a sample wording). of questionnaires (Dillman, 1999; Er-
In a mailed questionnaire, a letter de- dos, 1983; Jackson, 1999). However,
scribing the project and requesting the when students are being surveyed or
cooperation of the recipient is gener- in situations in which there is little
ally included. Ideally, this letter should problem with refusals, it is possible to
be written on letterhead stationery and begin with items that reect the re-
be signed by the head of the organiza- spondents place of birth, gender, and
tion. One attempts to communicate the the size of the community in which the
importance of the project by identifying respondent lives. Perhaps then it will
its sponsors and by the professional ap- be possible to move to issues such as
pearance of the mailing. Normally, a the respondents year of birth.
stamped, return envelope is included Whether one begins with background
for the respondents convenience. information or other questions, it is im-
Figure 14.1 illustrates a sample letter portant that the respondent be able to
that might accompany a mailed ques- move quickly through these rst items,
tionnaire. It is important that the intro- creating the impression that it will take
BOX 14.9 Sample Introductory Questionnaire Statement
We hope that you will be willing to complete you may stop completing the questionnaire at
this important survey of health behavior being any time you wish without penalty. We hope,
conducted by a group of nursing students as however, that you will complete it in its en-
part of their course requirements. We hope tirety. There are no right or wrong answers.
you will find the questions interesting. Please Please be as truthful as possible. Your answers
complete the questions in the order in which are completely confidential. You do not need to
they are presented. You have the right to skip put your name on the questionnaire, and no
any question you do not wish to answer, and one will know how you answered the items.
Figure 14.1. Sample introductory letter and consent form to accompany a questionnaire.
only a few minutes to complete the dering is to begin with questions that
whole questionnaire. Researchers are the respondents will consider impor-
always concerned to do everything that tant. Respondents are more likely to
they can to increase the proportion of complete a questionnaire if they view
people who will complete the form suc- it as salient (Goyder and Leiper, 1985;
cessfully. If obtaining the background Jackson, 1999).
information involves questions that Some researchers suggest the funnel
may be considered highly personal, technique for ordering items. This pro-
then it is best to place such items at the cedure begins with very broad ques-
end of the questionnaire. tions and narrows down to quite speci-
Another suggestion for question or- c items. Whatever the method chosen,
it is important to follow a logical order. check-one-of-the-following questions

For example, if time is used to order in the same manner, the respondent
questions, one might begin with the soon gets the idea that questions pre-
past, move to the present, and then the sented in this manner are to be re-
future. Do not move back and forth be- sponded to by choosing one of the op-
tween times because it would confuse tions provided by the researcher. Later,
respondents. The order in which ques- when a ll-in-the-blank question is
tions are posed can inuence subse- asked, the researcher can draw atten-
quent responses. tion to the fact that something else is re-
Rule 3. Key and repeated variables at quired by varying the format. For exam-
one third point. In most projects, there ple, if the respondent is to be asked to
are a few variables that are particularly rank order a list or to choose the three
important to the study. In most cases, most important items from a list, then a
there will be a major dependent vari- different look should be given to the
able. Ordinarily, it is a good idea to question by varying the way in which
place key variables about one third of the response categories are set up
the way through the questionnaire. One perhaps a short line opposite each item
wants the respondent to be fresh and rather than the little tick-off boxes. (See
be paying maximum attention when Boxes 14.1 through 14.5 for illustra-
these key variables are presented. tions; also see Appendix B and check
If there are to be any reliability the Web page at http//www.stfx.ca/
checks (repeated questions), then one people/ wjackson). The key point is that
needs as much separation as possible different formats should be used for dif-
between the rst and second presenta- ferent types of questions; this will help
tion of such questions. The rst pre- respondents answer the questionnaires
sentation of any repeated questions rapidly.
should also be placed near the begin- Note that the format in Box 14.1 re-
ning of the questionnaireafter the requires the respondent to place a check-
spondent is warmed upbut before fa- mark in the appropriate box. Check-
tigue or boredom sets in. marks are slightly easier than circling
Rule 4. Group questions by type. Some responses, so they are preferable. Any
grouping by type of item is advisable. device to make the task easy for the re-
Minimize the amount of shifting be- spondent should be incorporated.
tween open-ended questions and pre- Rule 6. Anticipate computer data entry.
coded questions. If there are a number Virtually all questionnaires of any size
of Likert-type items, it is a good idea to will be analyzed using a computer. To
group some of them together (Jack- simplify the entry of the data into the
son, 1999). The idea is to get the re- computer, it helps enormously if the
spondents used to a particular format values that are going to be entered into
in order to permit quick movement the computer are right on the question-
through the items. Continual shifting naire. Indeed, it is now quite common
between types of questions will only for researchers to avoid what was once
slow down respondents and increase referred to as coding the data. By cod-
the risk of error. ing, reference is made to researchers
going through the questionnaire and
2. Formatting assigning values to each response cate-
gory, recording this information on
Rule 5. Begin conditioning respondents. code sheets, and then having the data
The rst questions begin a condition- entered into the computer off the code
ing process. By always presenting the sheets. This process was expensive,
time consuming, and likely to introduce the use of code values, uses square
additional errors into the nal data set. brackets for the check-off responses,
When possible, avoid the whole coding and places the response brackets be-
process by placing the code values di- fore the category (Babbie, 1992, p. 155).
rectly on the questionnaire. These Dillman favors the use of capital let-
numbers can then be used when the ters to identify response categories,
data are entered into the computer. and, in mailed questionnaires, places
They can be placed unobtrusively on the code values to the left of the re-
the questionnaire and will not distract sponse categories. Dillman recom-
the respondent when they are complet- mends placing the code values to the
ing the form. right of the categories in personal in-
Box 14.1 illustrates a set of questions terview forms and telephone surveys
in which such numbers have been (1978, 1999). Both Babbies and Dill-
used. Readers should note that there mans styles are acceptable.
are no blanks between the response Another formatting style is illus-
category and the check-off space (usu- trated by questionnaires developed by
ally a small box). Note, as well, that the Population Research Laboratory at
only question numbers are used and the University of Alberta. Harvey Krahn
that response categories are num- and his associates tend to include the
bered beside the place where the re- actual computer code values on the
spondent makes his or her checkmark. questions, placing them to the right of
As readers will appreciate, when the the categories. Having respondents cir-
data are entered into the computer, it cle numbers placed to the right of the
is easy to enter the number immedi- category ensures that right-handed re-
ately adjacent to the checkmark made spondents do not cover the category
by the respondent. label with their hands. This formatting
Box 14.10 shows the formatting style style is simple, uncluttered, and rec-
favored by Earl Babbie, Don A. Dillman, ommended.
Harvey Krahn, and the authors of this Our preference is to include code val-
text. Babbie does not appear to favor ues and place them to the right of the

response categories, using a series of ing one of the techniques shown in Box
dashes to carry the respondents eye 14.11. Note that those who responded
out to the check-off brackets. Note that positively to question 9 are directed to
no space is left between the code value question 9.1.
and the brackets. (The intention is to
make the numbers disappear as much
as possible; when spaces are left be- 3. Presentation
tween the number and the bracket, the
Rule 9. Give questionnaires a distinctive
number seems to gain prominence.) Be-
look. Besides paying attention to issues
cause the numbers that will eventually
of layout, it is also important to give
be entered into the computer are di-
questionnaires a distinctive appear-
rectly beside the checkmark, data entry
ance. The use of colored paper is one
is simplied and less subject to error.
way to achieve this. Colors should be
Some researchers place the com-
pretested for effectiveness but, in gen-
puter screen column numbers on the
eral, pastels have been found to elicit
questionnaire that each variable will oc-
the highest response rate (Seaman,
cupy. These are placed on the extreme
1987). Yellows and pinks are especially
right-hand side of the questionnaire.
noted for eliciting a high return rate. An-
The advantage of this is that errors can
other useful tip is to have the question-
be avoided when entering the informa-
naire prepared on 81/2 14 paper and
tion into the computer. Certainly, in the
then have it reduced to the standard 81/2
case of interviews or phone question-
11 paper. When this is done, use
naires, it is advisable to include such
small margins because a wide margin
codes. However, in cases in which re-
will show up after the reduction is
spondents are lling in the question-
made. Although slightly smaller print
naires themselves, such numbers add
will result from reducing the page size,
to the clutter on the questionnaire and
this will permit more space between
should probably be avoided. (In Chap-
questions and greater exibility when it
ter 16, additional suggestions are made
to help reduce errors in data entry.)
Rule 7. Vary placement of response cat-
egories. Box 14.1 also illustrates how BOX 14.11 A Branching Question
the appearance of a questionnaire can Illustrated
be improved by varying the placement 9. Have you smoked cigarettes in the past
of the response categories for each 7 days?
question. Although this might be an Yes ---------- 1
aesthetic rather than a scientic con- No ---------- 2
sideration, the appearance of a ques-
tionnaire in which the response cate-
9.1 If Yes: How much did you smoke
gories are all lined up on the right in the past week?
margin is not pleasing to the eye, nor is Less than one package----1
it easy to see where one question ends 1 to 3 packages ------------2
and another begins. The rule, then, is 4 or more packages-------3
to vary the location of the response
category sets.
Rule 8. Clearly indicate any branching. 10. Have you ever been married?
Some questions are only to be an- Yes ---1 [Answer questions 1114]
No ---2 [Go to question 15]
swered by some respondents. In such
cases, guide your reader clearly by us-
comes to formatting each page of the scriptive information from respondents

questionnaire for optimal appearance. about phenomenon such as attitudes,
When choosing font size, consider the knowledge, behaviors, opinions, or facts
age of the expected respondent; older about events or people. Questionnaires
respondents, in particular, will appreci- are economical in that they can be distrib-
ate a somewhat larger print. uted to large samples through the mail
Rule 10. Do not squeeze too much onto or directly; they can be administered with
one page. Avoid trying to squeeze too ease; and they can be completed and
much onto a single page; question- analyzed in a limited time period. They
naires should permit the respondent present items in a consistent manner; min-
to move through each page rapidly. imize bias that is often present in inter-
Squeezing material looks bad and dis- views and other forms of data collection,
courages your respondent. Moreover, and offer the possibility of anonymity.
if it is squeezed too tightly, it will be
harder to code.
2. Limitations
Although questionnaires provide many

E. EVALUATION OF advantages, they are not without limita-
QUESTIONNAIRES tions. Questionnaires require consider-
able thought, planning, and testing to cre-
1. Advantages ate a valid and reliable data collection
instrument. Because individual items of-
The questionnaire is one of the most pop- ten address a variety of different topics
ular tools for data collection in nursing re- within one questionnaire, it is difcult to
search. It enables the researcher to collect determine the reliability of all items (see
self-report data from respondents either Chapter 13). Table 14.3 outlines some
through written or verbal responses. questions that may be asked in helping
Questionnaires are especially useful when beginning researchers to assess various
one wishes to collect a broad range of de- questionnaire items.
Table 14.3 Queries to Help Judge Questionnaire Items*

Item Query
Format In xed choice questions, are the question response categories mutually exclusive?
Does the formatting make data entry easy?
Is the length of the questionnaire likely to discourage completion of the instrument?
Language Will the wording be easily and consistently understood by the respondents?
Are the questions culturally appropriate, not offending some respondents?
Does the language invite the respondents to express their opinions?
Does the strength of the wording provide for sufcient variation in response?
Indexes If the index was developed by other researchers, has its reliability been tested?
Has the index been used on respondents drawn from similar populations?
Does the content of the items cover the essential variables of the study?
Validity Are the proposed measures congruent with the denitions implied by the constructs?
Have other researchers established the validity of the measures?
*If you answer yes to all these queries, then you can be condent that the questionnaire is appropriate to use in your
study. If you answer no to one or more of these queries, then you should revise the questionable items. If you are
considering the use of an existing instrument designed by another researcher and you answer no to some of the
queries, you should consider developing a new instrument.
Clear written instructions must ac- is very good (Babbie, 1992). Hand delivery
company the instrument so that the re- and pick up of completed questionnaires
spondent responds appropriately to the may be a feasible way to increase re-
items provided. Data collected by ques- sponse rates. An alternate method may be
tionnaires tend to be less in depth and to mail the questionnaires and then visit
more supercial than those collected by with respondents on prearranged dates to
interviews. Even open-ended questions pick up the completed questionnaires.
that allow respondents to express their To increase response rate, the re-
perspectives do not probe for deeper searcher may use inducements. A variety
meanings. of approaches have been used with vary-
ing success. Items such as a small pay-
3. Challenges ment to compensate for time consumed
by the questionnaire completion, the
Questionnaires require researchers to placement of a pencil in the envelope to
provide, in addition to clear, coherent ease completion, the promise of a sum-
questions, a sense of motivation that in- mary of the study results, and an appeal to
spires respondents to complete the in- the altruistic nature of the respondent to
strument. It is often a challenge to get re- contribute to knowledge development or
spondents to complete and return increase understanding of some phenom-
questionnaires. This is particularly true enon important to society may yield re-
with mailed questionnaires. The response spectable return rates. In addition, a non-
rate for mailed questionnaires is often low response questionnaire procedure can be
(25 to 30 percent). To be respectable, a used; in 2 weeks, the researcher may send
mailed questionnaire should hope to elicit a postcard to remind the participants to
a 50 percent response rate (Seaman, 1987; return the questionnaires; 4 weeks after
Jackson, 1999). Others indicate that a 60 the initial mailing, the researcher may
percent rate is good and a 70 percent rate send another questionnaire.
E X E R C I S E S
1. Develop a series of background infor- b. Sit down with a fellow student and
mation questions that would be ap- have him or her complete the re-
propriate for a study you would like to vised questionnaire. What changes
do. Be careful to obey as many of the resulted from having another per-
rules for questionnaire construction son complete the questions? Revise
as you can. the questionnaire to remove any
ambiguities.
2. Using the questions developed in

question 1: 3. Develop a series of items for an index
a. Complete the questionnaire your- that is of interest to you. Dene the
self. Decide which changes you conceptual aspects of the index and
would make, based on problems suggest indicators.
that emerged when you completed
the questionnaire. Revise the ques-
tionnaire.
4. Construct a 10-item questionnaire to to catch up with other health-care

collect data about womens views of professionals?
health and the factors that inuence
health-seeking behaviors. Include a set
of instructions to respondents, 10
questions and response categories, Do you agree that to be in good
and a section on demographics (back- shape is to feel physically t?
ground information)
Are nurses underpaid and underval-

ued in society?
5. Examine the following questions and
identify what, if anything, is wrong
with the items. How often have you engaged in pre-
marital sexual intercourse?
Do you disagree that dieting is not a

good means to enhance body shape Do you support mandatory use of
in adolescents? seat belts and older ages for is-
suance of a drivers license?
Nurses are underpaid. Therefore,

should the government give them a How often do you tend to get de-
substantial increase to enable them pressed?
Bradburn, N.M., and Sudman, S. (1980). Im- Frank-Stromberg, M., and Olson S. (1997).
proving Interview Method and Question- Measurement in Nursing Research (2nd ed).
naire Design. San Francisco: Jossey-Bass Boston: Jones & Bartlett. An excellent re-
Publishers. source for both the development and use
Dillman, D.A. (1999). Mail and Electronic Sur- of instruments in nursing research and
veys: The Tailored Design Method. New clinical practice.
York: John Wiley & Sons. This book is an Kerlinger, F.N. (1986). Foundations of Behav-
excellent presentation of many of the is- ioral Research (3rd ed). New York: Holt,
sues involved in doing surveys. Rinehart, and Wilson.
Dillon, J.T. (1990). The Practice of Questioning. Sudman, S., and Bradburn, N. (1983). Asking
New York: Routledge. This is a frequently Questions. San Francisco: Jossey-Bass Pub-
cited resource for questionnaire develop- lishers.
ment. Waltz, C., Strickland, O., and Lenz, E. (1991).
Erdos, P.L. (1983). Professional Mail Surveys. Measurement in Nursing Research (2nd ed).
Malabar, FL: Robert E. Krieger Publishing. Philadelphia: F.A. Davis. This comprehen-
This book should be regarded as required sive text describes the process involved in
reading for anyone considering doing a selecting, testing, and designing instru-
mail survey. ments for use in nursing investigations.
Chapter 15
Sampling and Sample Size
CHAPTER OUTLINE
A. The Terminology of Sampling 2. Steps in Determining Sample Size and

B. Fundamental Sampling Techniques Estimating a Nominal Variable
1. Probability Sampling Techniques 3. Power Analysis and Sample Size
2. Nonprobability Sampling Techniques D. Other Sampling Issues
3. Qualitative Sampling Techniques 1. Sample Size and Accuracy
C. Sample Size Determination 2. Sample Size and Condence Limits
1. Steps in Determining Sample Size and 3. The Impact of Refusals
Estimating a Ratio Variable 4. Conrming Representativeness
KEY TERMS
Accessible population Nonprobability sampling Sample

Alpha techniques Sampling error
Beta Population Sample interval
Biased sample Power analysis Sampling fraction
Cluster sample Primary sampling units Sampling frame
Convenience sample Probability sampling Simple random sample
techniques
Effect size Skip interval
Purposive sampling
Expert sampling Snowball sampling
Quota sample
Matched pairs Stratified sample
Response rate
Multistage area sample Systematic sample
495
N
urse researchers are interested in The Literary Digest poll described in
studying elements of the nursing Box 15.1 was decient in that:
world. One way to do this is to take
measures on the relevant variables for all
It lacked a representative sample of
the general population.
people in the population of interest. Al-
though this solution is possible if one is
Circumstances altered substantially
between the time of data collection
studying a small population (e.g., clients
and the time of the election.
who received nursing care in the emer-
gency department of the local hospital), The lesson to be learned from the Lit-
when the concern is to understand how erary Digest debacle is that, for reasons of
people from larger aggregations such as a both economy and accuracy, a well-
community, region, or country feel about selected small sample is to be preferred
certain health issues, then it is necessary to a large, poorly selected sample.
to gure out ways to get a sense of their Sampling may involve the selection of
beliefs and attitudes without having every- people, events, organizations, periods of
one complete a questionnaire or agree to time, records, and many other phenom-
be interviewed. In such cases, a sample ena of concern to the eld of nursing. It is
that can reect or represent the views of used in a great variety of nursing research
the larger population may be drawn. A projects, such as choosing which client
sample represents a microcosm of the records will be subjected to a content
population you are interested in studying. analysis, choosing and then assigning
It is a way of representing the variability nurse volunteers to different conditions in
within a large population through the experimental designs, or deciding which
study of a smaller number selected from members of an HMO are to be included in
that population. If the sample is represen- a telephone survey.
tative of the population from which it is
drawn, the researcher can have con-
dence in concluding that the results are
A. THE TERMINOLOGY OF
generalizable to the entire population and
SAMPLING
setting studied. This chapter discusses
sampling techniques, procedures for se-
To begin our discussion, some terms
lecting samples, and methods for estimat-
need to be distinguished:
ing sample size.
Sampling is done to save time and Population. The population is the en-
money. If you can estimate who is going to tire group that you wish to describe. In
benet from a health promotion program Box 15.1, the population would be the
with a sample of 250, why poll 1000 peo- American electors of 1936. If you were
ple? And although common sense sug- studying clients with asthma, the popu-
gests that larger samples would be more lation might be dened as all people
accurate than smaller ones, this is not nec- with asthma residing in the United
essarily the case. The key issue is that a States. Other examples of research
sample must be representativeit must populations of interest to nurses in-
reect the population accurately. Box 15.1 clude successful code blue procedures
reports a classic case of an inaccurate at hospital X in 1999; preschool chil-
sample. In this famous case, a huge sample dren with hearing disorders in Ontario,
(more than two million people) failed to Canada; nurses with clinical specialist
predict the winner of the 1936 American registration in the state of California; 5-
presidential election (Simon and Burstein, year survivors of breast cancer in North
1985, p. 108). America; state-funded baccalaureate
SAMPLING AND SAMPLE SIZE 497
BOX 15.1 Predicting the 1936 American Election
In 1920, 1924, 1928, and 1932, the New York was with considerable confidence that the Lit-
based Literary Digest conducted polls in an effort erary Digest conducted its 1936 poll. On Octo-
to predict the winner and the winning margin in ber 31, 1936, the magazine published its results
the American presidential elections. The polls based on mailings to some 10 million Ameri-
were accurate (within 1 percent in 1932), so it cans. The results indicated:
Literary Digest Prediction of 1936 American Presidential Elections*

Candidate Number of Votes Percentage of Votes
Landon 1,293,669 57.1
Roosevelt 972,897 42.9
TOTAL 2,266,566 100.0
*Responses from more than 10,000,000 ballots mailed out (under 23 percent response rate).
The poll suggested that Landon would easily the new voters came largely from poorer
win the presidential election, but it turned out groups (few of whom owned cars or tele-
that Roosevelt won by a substantial margin. So phones). Gallop also noted that those most
why, with a sample of more than 2 million, did likely to return their ballots would overrepre-
the Literary Digest miss the mark so badly? sent the better educated and the higher so-
The founder of the American Institute, cioeconomic categories.
George Gallop, predicted in July of 1936, some Gallop thus challenged the sampling frame
months before the Literary Digest began mailing used by the Literary Digest. The sampling frame is
out its ballots, that it would not be accurate. the target population from which the researcher
Gallop pointed to the flaw of using mailing lists draws a sample. The sampling frame did not ac-
based on telephone listings and listings of auto- curately reflect American electors; rather, it fa-
mobile owners. Gallop pointed out that such vored the more prosperous elements in society,
lists largely favored the more economically and Roosevelts strength was among the poorer
prosperous; during the years of the Depression, elements of American society.
Source: Summarized from Simon J.L., and Burstein P. (1985). Basic Research Methods in Social Science (3rd ed). New
York: Random House, pp. 107110.
nursing programs; or health records of may be interested in studying percep-

all women diagnosed with depression tions of caring in clients who receive
at hospital X in the year 2000. As the ex- care in emergency departments. The re-
amples suggest, a population may be search population is all clients who re-
dened broadly and potentially include ceive care in emergency departments. It
millions of people, phenomena, or is impossible to get a complete list and
events, or it may be dened narrowly contact all such clients. Therefore, you
and include only a few hundred people should choose your sample from the ac-
or events. The terms population and cessible population; that is, those clients
universe are used interchangeably. to whom the researcher has feasible ac-
Accessible population. This is the pop- cess. You may have access to clients in
ulation that is feasible for the re- emergency departments of the four gen-
searcher to access. For example, you eral hospitals in your county, the re-
gional trauma center, and the local hos- interest to the research study? For
pital emergency department. Any of the example, if you are interested in
clients at these six centers would con- studying smoking behaviors in dis-
stitute the accessible population for the advantaged women and you select
study. your sampling frame from a list of
Sampling frame. This is the list (or women attending a private clinic,
lists) from which you draw a sample. you most likely will miss many dis-
In Box 15.1, the sample frame included advantaged women who cannot af-
all those people listed as automobile ford to attend a private smoking ces-
owners plus all those who had a sation clinic.
phone. Ideally, the sampling frame and If the above four factors are consid-
the population are identical; in prac- ered, it is likely you will increase the
tice, however, it is often not possible probability of getting a representative
to get a complete list of the population, sample of the population you wish to
so the sampling frame may not per- study.
fectly reect the population. Almost
any list will not be exactly up to date;
Sample. The sample refers to those in-
dividuals (or units) selected for a study.
some people will be left off and other
people may be listed twice. The goal is
Response rate. The response rate
refers to the percentage of delivered
to get as accurate a sampling frame list
questionnaires that are completed and
as possible. In evaluating the accuracy
returned.
and completeness of the sampling
frame, you should consider the follow-
Biased sample. The sample is not rep-
resentative of the population it is in-
ing:
tended to reect.
1. What systematic omissions from the
list have resulted from the enumera-
Sampling error. This represents the
difference between the values emerg-
tion method? For example, are seri-
ing from the sample and those emerg-
ously ill clients, socially disadvan-
ing from the population on some phe-
taged, or illiterate clients omitted?
nomenon.
2. Will omissions bias the study nd-
ings and generalizability of results?
For example, if seriously ill clients
are omitted from a list of clients re- B. FUNDAMENTAL SAMPLING
ceiving care in emergency depart- TECHNIQUES
ments, will this bias results of a study
of perceptions of caring by emer- There are two categories of sampling pro-
gency clients? cedures used in quantitative studies:
3. Is there any systematic bias inher- probability sampling techniques and non-
ent in the ordering of the sampling probability sampling techniques. Table
frame? For example, if the sampling 15.1 summarizes the major quantitative
frame systematically alternates all sampling procedures. You may wish to re-
names by gender, a sample may be fer to this table as you read about each
selected that is all male or all female. sampling method. Qualitative studies for
This is certain to happen in system- the most part use nonprobability sam-
atic sampling if every second name pling techniques. These are discussed in
is selected from the list. Chapters 6 and 7. Table 15.2 presents a ty-
4. Does the list of potential partici- pology of common sampling strategies
pants represent the population of used in qualitative research.
Table 15.1 Summary of Quantitative Sampling Procedures

Type Description of Procedure Strengths and Limitations
Probability
Simple random All members of a population are enumer- Risk of bias is low
ated and then the sample is selected Representativeness is enhanced
from the sampling frame using random Time consuming to draw sample
procedures
Stratied random Selection of participants from two or Tedious to draw sample
more strata Requires knowledge of underlying
structure of population
Representativeness is high
Bias is low
Systematic Selection of every kth person in the Requires a list of the sampling frame
sampling frame Observe for systematic bias in the
ordering of the sampling frame
Representativeness limited if bias is
present in the sampling frame
Cluster Multistage sampling in which large Efcient way to draw a large sample
clusters are sampled rst, followed by if elements in a population are not
successive sampling of smaller and available
smaller units Higher sampling error than simple
random
Convenience Participants selected because they are Easily accessible
in the right place at the right time High risk of bias
Questionable representativeness
Quota Inclusion of participants based on pre- Uses knowledge of population to
determined characteristics; a type of enhance representativeness of
convenience sampling sample
Bias is present
Snowball Participants are referred by earlier Facilitates access to difcult-to-nd
participants subjects
Representativeness is suspect
Bias is high
Purposive Selects subjects on the basis of personal Bias may be high
judgment about representativeness Limited ability to generalize results
Easy to draw sample
Expert Subjects selected on the basis of their Representativeness is questionable
expertise Achieves consensus efciently and
effectively
1. Probability Sampling probability sampling (see Chapter 12). If

Techniques your study involves the use of such tests,
be certain to meet the sampling assump-
Probability sampling techniques are used tion. The sampling units usually are indi-
in selecting sampling units so that each viduals, but may also be other levels of
unit has a known chance of being included. analysis, such as communities or coun-
It is important to appreciate that tests of tries. Probability sampling usually yields
statistical signicance assume that sam- samples that are representative of the pop-
pling has been done using some form of ulation from which they were drawn and
Table 15.2 Typology of Sampling Strategies In Qualitative Inquiry

Type of Sampling Goal
Maximum variation Deliberate selection of participants who are outliers from the norm
to see if the commonalities exist with the larger group
Homogenous Deliberate selection of participants with a range of similar experi-
ences; this facilitates cohesion during the interview process
Critical case Important case that permits generalization of ndings to other cases
Theory based Selection of the sample is determined by the goal of nding examples
of a theoretical construct; participants are selected on the basis of
whether or not they present an instance of the theoretical construct
Conrming and disconrming Sample is selected to maximize both deviant and typical cases
cases
Snowball or chain Sample selected by participants referring the researcher to other
people who share common experiences
Deviant case Selection of outlier cases that show great variation in the phenome-
non from the majority of the sample
Typical case Represents what is normal or average in the sample
Politically important case Participants selected for their value-added potential to anticipated
issues in the analysis
Random purposeful When a potential purposive sample is too large, a random purposeful
sample is used to select a sample of appropriate size
Stratied purposeful Uses subgroups to select participants that represent dimensions on
which variability is sought
Criterion selection All cases meet some criterion
Reputational case Selection of participants on the basis of reputable key informants
Convenience Selection of participants on the basis of ease and convenience but at
the expense of information and credibility
Source: Summarized from Miles, M.B., and Huberman, A.M. (1994). Qualitative Data Analysis (2nd ed). London: Sage.
avoid bias. Representativeness of the sam- ticipation in a study. This procedure is

ple is necessary if the researcher is to the best-known probability approach. It
make generalizations about study results requires that a list of the potential re-
to the larger unmeasured population. spondents is available to the researcher.
Probability sampling involves the tech- Such lists might include student lists,
niques of simple random sampling, sys- lists of clients from a health-care facility,
tematic sampling, stratied sampling, and lists of employees, lists of nursing pro-
multistage or cluster sampling. The programs, and so forth. The following steps
cedures for selecting each of the major are necessary when selecting a simple
types are listed below. random sample:
Step 1. Number the units on the list.
a. Simple Random Sample Step 2. Computers are typically used to
generate the random numbers used in
A simple random sample provides each sample selection. Alternatively one
unit (usually a person) in the population may use a table of random numbers
an equal chance of being selected for par- such as the one provided in Table 4.1
(Chapter 4) to select the required num- For instance, one could choose students
ber of units. (Sample size determination living in every fourth residence room
is discussed later in this chapter.) To and, as long as one numbers the rooms
use a table of random numbers, shut systematically, one could proceed with-
your eyes and place a pencil mark on out a list of the students.
the table. The number closest to where The critical issue is that every person
the mark is made is used as the starting must have a known (usually equal) chance
point. From this point, one reads the of being selected. The steps in selecting a
numbers systematically (perhaps read- systematic sample are as follows:
ing down the column, continuing at the
top of the next column, and so forth), Step 1. Get a list, map, or diagram as ap-
placing a checkmark beside the cases propriate.
whose number shows up on the ran- Step 2. Having determined the sample
dom number table. This process con- size required plus the additional num-
tinues until a sufcient number of cases ber for replacements (for the refusals
for the study has been selected. or for those with whom contact cannot
Step 3. Additional replacement units be made), these two gures should be
should be selected and kept on a sepa- added together and be regarded as the
rate list, so that when a sampling unit total sample requirement.
cannot be contacted or an individual or Step 3. Divide the total sample require-
program does not participate, then ment into the total number of units in
that unit will be replaced by the rst re- the population being surveyed. This
placement unit. Replacements should number should then be rounded to the
be identied and numbered because nearest, but lower, round number. (e.g.,
they will be used in the order in which if the number you get is 8.73, round it to
they have been selected. Numbering the nearest, lower, whole number; in
them as R1, R2, and so on is a conve- this case, 8). This number represents
nient way of noting them. what is known as the skip interval or
the sample interval.
One caution that must be mentioned
b. Systematic Sample in the use of systematic samples is that
if the list is ordered in some fashion,
A systematic sample provides each unit there may be a problem. Suppose you
(usually a person) in the population an have a listing of couples with the name
equal chance of being selected for partic- of the man always listed rst. In such a
ipation in a study by choosing every nth case, if the skip interval were an even
unit, starting randomly. The systematic number, then all those selected would
sample provides a somewhat easier way be women. If the list is patterned, as in
of selecting cases from a list of potential the example discussed, a different sam-
respondents. In this case, names listed in pling procedure will have to be used.
birth announcements, phone books, Step 4. Using a table of random numbers,
health maintenance organization directo- select a number between 1 and the
ries, street maps, dormitory diagrams, value of the sample interval. The num-
nurse registration lists, student lists, or ber selected becomes the starting
surgical lists all might be sources from case, the rst one selected to partici-
which one might draw a systematic sam- pate in the survey. Suppose we were
ple. In the case of systematic samples, it doing a survey of nursing students liv-
is even possible to proceed with sam- ing in campus dormitories and we
pling when no list exists before sampling. have determined the skip interval to
be 8 and the starting case to be 3. In lescents in the public school system, such
this case, we would develop a system- a procedure might be somewhat wasteful
atic procedure for numbering the dor- because some schools will be made up
mitory rooms and for moving from predominantly of young adolescents, oth-
oor to oor, dormitory to dormitory. ers will be made up of middle adolescents,
We would begin with the third door, and others will be made up of older ado-
then go to the eleventh, nineteenth, lescents. As a result, a very large sample
and twenty-seventh. In rooms with two would be required to provide a sufcient
nursing student residents, we would number of respondents to allow general-
ask both to participate; otherwise, stu- izations about lifestyle patterns across the
dents in double rooms would have less adolescent age span. In situations such as
chance of being selected for the study. this one, it is useful to draw a stratied
However, caution has to be exercised sample. A stratied sample gives respon-
because, in some universities, nursing dents within each of the three adolescent
students may select their roommates age groups an equal chance of selection,
and, if this is the case, there would be but at the same time, ensures that an
a lack of independence in some of the equal number will be selected from each
units sampled. If this is a problem, the of the three age groups or strata.
researcher would be well advised to
Step 1. Determine the sample size re-
obtain a list of all nursing students in
quired from each of the categories.
residence and then use a random sam-
Step 2. Develop a list for each of the age
pling procedure.
strata from which you wish to draw
When systematic samples are being your sample.
selected from lists, it is a rather straight- Step 3. Using either a systematic or ran-
forward matter to go through the list, dom sampling procedure, choose the
placing checkmarks beside cases that cases for the sample along with the re-
have been selected, perhaps marking quired number of replacements.
every fifth one. Additional units should
Samples may be stratied by more than
be selected for use as replacements and
one variable. In order to achieve more pre-
marked with R1, R2, and so on to indi-
cise estimates, we might also have strati-
cate the order in which they should be
ed the previous sample by gender and by
used.
grade level. The procedures are identical;
simply identify the stratication dimen-
c. Stratified Sample sions an then select respondents using an
equal probability procedure. The number
There are times when a simple random or of adolescents in each stratum should be
systematic sample would not provide an the same as the proportion of the group in
appropriate solution to sample selection. the total population. For example, if the
This is true if you are interested in study- population is the total adolescent age
ing a stratum or subpopulation such as span and the strata are early adolescents
juveniles with diabetes, people with con- (33.3 percent of the population), mid ado-
genital heart defects, or individuals in a lescents (40.7 percent of the population),
particular age group. Suppose, for exam- and late adolescents (26.0 percent of the
ple, that you are doing a survey of lifestyle population), the sample selected should
patterns among the adolescent age group reect similar proportions. Figure 15.1 il-
(early, mid, and late adolescent strata) at- lustrates the use of proportional stratied
tending public schools. Although it would random sampling to select a sample of
be possible to do a random sample of ado- adolescents.
Total
1000 adolescents in School District X
population
Stratum 1 Stratum 2 Stratum 3

Early Mid adolescence Late adolescence
adolescence 40.7% = 407 26.0% = 260
Proportional 33.3% = 333 adolescents adolescents
stratification adolescents
RANDOM SELECTION
Proportional
sampling
from each 110 Early 166 Mid 68 Late
strata adolescents adolescents adolescents
Figure 15.1 Proportional stratified random sample of adolescents.
d. Multistage Area Sample eventually get down to the level of the in-
dividual who is randomly selected from
When the task involves developing a sam- the smallest cluster. In a simplied form,
ple to reect a large unit such as a state, the procedures for large-scale surveys
province, or country and no list of the such as a national surveys may be sum-
population is available, then one develops marized as follows:
a multistage area sample or cluster sam-
ple. The key point is that at each succes- Step 1. Identify primary sampling units
sive stage of the sampling process, every (these may be census tract areas or
individual (or unit) must have a known other similar units, normally several
chance of being selected. The difference hundred of them); these units are num-
between a simple random sample and a bered, and a selection of units is made
cluster sample is that a simple random from them, using an equal probability
sample is a one-stage process: a popula- technique.
tion is listed and items are selected from Step 2. Within the selected areas, identify
it at random. In a cluster sample, the pop- the city blocks (in urban areas) or
ulation is divided into groups or clusters; square miles (rural areas). From these,
the rst units or clusters to be sampled choose an appropriate number of units,
are the largest, followed by smaller and making sure to use an equal probability
smaller units selected at random. You technique
Step 3. Within the selected areas, number then randomly select nursing students
the housing units and select the units from the schools.
that will be used randomly from among It should be noted, however, that no
these. sampling choices should be made by data
Step 4. For each household, list the people collection personnel. All choices should be
who fall within the desired sampling pa- made by probability procedures. When a
rameters (e.g., adults, older than age 18 selected participant is not available (e.g.,
years who have lived in the community he or she has moved away, not home after
for 1 month or more). Table 5.2 (Chapter three callbacks, or refuses to cooperate),
5) gives an example of a sample form. then replacements are used, in the order se-
Step 5. An equal probability procedure lected. It is critical that interviewers not
such as simple random or stratied simply replace the unavailable respondent
methods is then used to select respon- with the nearest, most convenient replace-
dents from a list of those who are eligi- ment. This would bias the sample toward
ble. those who are at home or are more coop-
erative. Whatever rules are established,
Cluster or multistage sampling has these should be communicated clearly to
many applications in nursing research. those doing the data collection and should
For example, it may be used by a school also be made clear in any technical reports
health nurse who is interested in studying on the research project (see Table 5.2 for a
lifestyle patterns in school-aged children. sample form that may be used).
The school nurse may work with a nurse Box 15.2 presents the telephone intro-
researcher to select a random sample of duction instructions for the 1993 Alberta
schools from the school district; within Survey (Kinzel, 1992). Although the tele-
the schools, she may randomly select phone numbers selected for dialing were
classrooms and within classrooms, she randomly determined, the procedures fa-
may randomly select children. This may vored interviewing men, particularly those
yield a representative sample more ef- who happen to have answered the phone
ciently than randomly selecting children call from the interviewer. In their proce-
from all schools in the district. Cluster dures, any time a man answered the phone,
sampling may also be used by community he was invited to participate in the inter-
health nurses to study family health prac- view. This procedure was an attempt to
tices. In this case, the researcher would compensate for the overrepresentation of
randomly select census tracts within his teenagers and women, who typically were
or her district, then blocks of households found to be more likely to answer the
within the census tracks, and then fami- phone. These procedures were followed to
lies within the household blocks. create a greater equality in the participa-
When it is impossible to obtain a list tion rates of male and female respondents.
of all the elements in a population, clus-
ter sampling is a practical approach to
ensure the representativeness of the 2. Nonprobability Sampling
sample. For example, if you are inter- Techniques
ested in studying student nurses atti-
tudes toward mandatory registration in Nonprobability sampling techniques are
your state, you may not have the ability commonly used by nurse researchers.
to get a list of the names of all nursing These procedures do not provide poten-
students. You could, however, get a list tial respondents with a known chance of
of nursing programs in your state and being asked to participate in a study. They
randomly select schools from that list, are, nonetheless, important to know about
BOX 15.2 The Alberta Survey
Telephone Introduction Sheet 1993 study to have the highest number of people
1. Hello, Im calling (long distance) on behalf of who have been selected take part.
the Population Research Lab at the Univer- 10. I cant replace your household with an-
sity of Alberta. My name is . . . (full name or other one without destroying the random-
Mrs. XXXXX) ness of the survey.
2. I have dialed XXX-XXXX. Is this correct? 11. Do you think anyone else in the household
3. Your telephone number was selected at ran- could take part (either now or later)? (do not
dom by computer. pause) Could someone assist us with the
4. Just a moment of your time to explain why study?
Im calling. Requesting to Speak to Someone Else in the
5. The lab at the university is currently con- Home
ducting an important study on current issues 12. We dont always interview the person who
(e.g., the family, health and well-being, AIDS, answers the telephone. In your household,
credit cards, crime, work, and social issues). it is one of the (male/female) members we
6. In order to determine who is eligible for the need to talk to.
study, please tell me how many women and 13. May I please speak to (him/her) (one of
men aged 18 years or older live on a regular them who is free at the moment)?
basis at this number? 14. Hello, Im calling from the Population Re-
NUMBER OF WOMEN _____ and NUMBER search Lab at the University of Alberta. My
OF MEN _____ This total includes yourself as name is . . . (full name or Mrs. XXXXX)
a member of this household over the age of 15. The lab at the university is currently con-
18. ducting an important study on current is-
Requesting an Interview with Person who sues (e.g., the family, health and well-being,
Answers the Phone AIDS, credit cards, crime, work and social
7. I would like to interview you. Im hoping that issues).
now is a good time for you. Your opinions 16. Your telephone number was selected at
are very important for the research that is random by computer. Im hoping that now
being done at the University of Alberta. is a good time for you. Your opinions are
8. Before we start, Id like to assure you that very important for the work that is being
your participation is voluntary and that any in- done at the University of Alberta.
formation you provide will be kept confiden- 17. Before we start, Id like to assure you that
tial and anonymous. As I mentioned, there is your participation is voluntary and that any
a wide variety of questions. If you do not wish information you provide will be kept confi-
to answer any of them, please feel free to dential and anonymous. As I mentioned,
point these out to me and well go on to the there is a wide variety of questions. If you
next question. Of course, you have the right do not wish to answer any of them, please
to terminate the interview at any time. feel free to point these out to me and well
go on to the next question. Of course, you
(Optional Read)
have the right to terminate the interview at
We do not need your name, so no one will any time.
know your answers to these questions. If you
(Optional Read)
have any questions about the survey, you can
call the study supervisor in Edmonton at XXX- We do not need your name, so no one will
XXXX for further information. know your answers to these questions. If you
have any questions about the survey, you can
Refusal call the study supervisor in Edmonton at XXX-
9. It is extremely important for the university XXXX for further information.
because much nursing research involves suring the impact of a health promotion
nonprobability sampling techniques. program on participants, the researcher
would try to match each program parti-
a. Quota Sample cipant with another person with similar
predetermined characteristics at the out-
In a quota sample, respondents are se- set of the program. Subjects may be
lected on the basis of meeting certain cri- matched for current health status, similar
teria. Quota sampling requires some un- patterns of health-related lifestyle, age, oc-
derlying knowledge of the population cupation, and so on. Matching pairs is
structure so that the researcher propor- a time-consuming and demanding chal-
tionately represents the strata in the lenge, but it enables control for bias in the
sample. However, no list of potential re- interpretation of ndings and reduces
spondents is required; the rst respon- sampling error (see Chapters 4 and 9).
dent to meet the requirements is asked to
participate and sampling continues until
all the categories have been lleduntil b. Convenience Sample
the quota for each has been reached.
Convenience samples involve selection
Suppose one were asked to compare food
on the basis of ease or convenience. If you
preferences of young adults with those of
were to poll people entering a shopping
elderly individuals. One might do a survey
mall on their attitudes toward current
of nutrition clinic clients, selecting the
health reform initiatives, you would be se-
rst 75 who meet the young adult crite-
lecting a convenience sample. Such sam-
ria and the rst 75 who meet the criteria
ples may, however, involve particular cat-
for inclusion in the comparison group.
egories of individuals. For example, by
Note that this sampling procedure is the
asking a couple of classes of tenth grade
nonprobability twin of stratied sam-
students to complete questionnaires on
pling. It is useful in providing adequate
health behaviors, you would obtain a con-
representation of different groups in a
venience sample of the students present.
population such as various age groups,
Convenience samples may be atypical of
ethnic groups, gender, or socioeconomic
the population with regard to the phenom-
levels. The major distinction between
enon of interest; therefore, caution must
quota sampling and stratied sampling is
be used in interpreting the ndings. More-
that the elements are not randomly se-
over, generalizability of results from stud-
lected in each strata of a quota sample.
ies using this sampling technique is not
The steps in selecting a quota sample
possible. The researcher has no control
are as follows:
over sampling biases or the representa-
Step 1. Dene precisely the criteria for tiveness of the sample. For this reason,
inclusion into each of the categories. convenience sampling is not encouraged
Step 2. Select participants on a rst-come, unless no other alternatives exist or the
rst-included basis until the quota for phenomenon of interest is homogenous
each category has been met. with respect to the characteristics studied.
Quota sampling may also employ
matched pairs. To do so, the researcher c. Snowball Sampling
predetermines signicant characteristics
in the sample that might inuence the de- Snowball sampling is a name for a refer-
pendent variable and matches the sample ral sampling procedure. As you complete
according to those characteristics. For ex- one interview, you ask if there is anyone
ample, if a researcher is interested in mea- else known to the respondent who might
be appropriate for the study. This tech- such as sickle cell anemia. It is also useful
nique is frequently used in situations in when one wishes to examine different ends
which one cannot get a list of the individ- of a continuum on some variable. For ex-
uals who share some characteristic. Some ample, to explore the concept of indepen-
examples in which you may choose to use dence, one may sample a group of early
this sampling procedure are: adolescents and a group of senior citizens
and compare the ndings. Purposive sam-
A study of the marital adjustment of
pling is frequently used in instrument de-
couples whose rst child was born be-
velopment studies in which the researcher
fore the couple married. In this case,
is interested in testing the validity of an in-
you would not easily nd a list of this
strument using the known-group tech-
population. But through referrals, you
nique (see Chapter 13) or pretesting the in-
could identify individuals who would
strument with a purposive sample of
be appropriate for your study.
divergent types of people. A limitation of
A study of prostitutes who are under
purposive sampling is the inability to as-
age 16 years
sess the representativeness of the partici-
A study of people convicted of white-
pants in relation to the population. The re-
collar crimes
searcher often assumes that errors of
A study of closet homosexuals
judgment in overrepresenting or under-
representing elements of the population in
d. Purposive Sampling the sample tend to balance out, but there
is no way to test this assumption.
Purposive sampling, also referred to as
judgmental sampling, uses the researchers e. Expert Sampling
expertise and knowledge of the population
to hand pick the cases to be included in the Expert sampling is actually a type of pur-
sample. The researcher usually selects posive sampling used with the Delphi tech-
participants who are considered typical of nique (see Chapter 13). The researcher
the population of interest. This subjective handpicks a sample of participants be-
sampling technique is common in qualita- cause of their expertise in the study phe-
tive studies. For example, in a qualitative nomenon. A series of questionnaires are
study, Baker (1996) used purposive sam- distributed to the expert sample and re-
pling to select 15 participants with schizo- sponses analyzed. After the analysis of the
phrenia to collect narratives about their questionnaires, questions are reformula-
inner experiences with schizophrenia. In ted and redistributed to the expert sample
addition, participants were purposively se- for completion and subsequent analyses.
lected so that the sample would include in- This is a means to achieve experts con-
dividuals of both genders, various age sensus on an issue.
groups, varying abilities to recognize signs
of relapse, and different lengths of time f. Advantages and Disadvantages of
since diagnosis. Nonprobability Techniques
This approach is useful when a re-
searcher is interested in understanding Note that none of these nonprobability
the experiences of certain segments of a techniques leads to independent, ran-
population. Some examples include the domly selected units. But for some re-
use of extreme cases. These are sought search problems, this type of sampling is
when one wishes to better understand a the only feasible one.
rare or unusual phenomenon such as a Convenience, quota, and snowball sam-
group of participants with a rare diagnosis ples are reasonable approaches to
sampling when one is investigating rela- of the experience with the researcher.
tionships among variables (e.g., the rela- Because the goal is understanding the
tionship between exercise participation meaning of the participants experience,
and self-esteem) and one is interested in qualitative researchers are typically not
trying to understand the conditions under interested in attempting to generalize
which there is (or is not) a relationship be- their results to larger populations. There-
tween the major variables rather than gen- fore, they typically do not use probability
eralizing to the entire population. Purpo- sampling procedures.
sive sampling is especially useful when
one is interested in understanding the ex-
periences of special segments of the pop-
ulation rather than the typical experience C. SAMPLE SIZE
of the entire population. Explanatory stud- DETERMINATION
ies, in fact, frequently use nonprobabil-
ity sampling procedures. However, when A variety of factors inuence the size of
tests of signicance are an important tool the sample that is appropriate for any
in ones research, it should be noted that study; thus, no simple rule can be applied
such tests assume that probability sam- in determining an appropriate sample
pling procedures have been used. There- size. If someone tells you that samples
fore, tests of signicance are not appro- should be 10 percent of the population,
priately used when quota, convenience, the advice is not sound. Sample size de-
snowball, purposive, or expert samples termination involves a series of trade-offs
have been used. between precision, cost, and the num-
bers necessary to do the appropriate
3. Qualitative Sampling analyses. At the same time, it is necessary
Techniques to take into account the amount of vari-
ability in the factors studied and, some-
Although there is considerable overlap times, even the size of the population.
between the sampling techniques used by A general guideline is the larger the
quantitative and qualitative researchers, sample, the more accurate the ndings.
most of those used by qualitative re- This is because a larger sample is more
searchers are variants of nonprobability likely to be representative of the popula-
procedures. Table 15.2 presents a typol- tion. A second guideline is the more het-
ogy of the more commonly used sampling erogeneous the population, the larger the
procedures used in qualitative studies. sample required. This is because there
An examination of the table indicates that will be much variability among variables
although some focus on showing differ- in a heterogeneous population so many
ences, others focus on showing the typi- participants are required to ensure you
cal, the extreme, or ones that identify po- get subjects that are typical of the popu-
litically important cases. Given that these lation from which they are drawn.
are nonprobability sampling procedures, Research design also inuences sample
it is inappropriate to conduct tests of sig- size. Designs such as experimental or eval-
nicance on the data or to attempt ex- uation studies that involve a number of dif-
trapolations to larger populations. ferent groups or cells require the investi-
In most situations, qualitative re- gator to estimate the number of subjects
searchers are interested in samples of par- required for each cell as well as the total
ticipants who have experienced the phe- study. To compare an experimental group
nomenon being studied and who are able with a control group on the dependent
and willing to share their interpretation measure, it is necessary to gure out how
many participants are required for each best to begin with the statistical ones, es-
cell. If the number in each cell is too small timating the required sample size, and
( 10), atypical values will change the then modify this number to take practical
treatment mean dramatically. Guidelines considerations into account.
for the number of participants per cell
range from 10 to 30 (Woods and Catanzaro, Step 1. Decide on the condence level to
1988; Wilson, 1993). Using cell sizes of 20 or be used. If you would like to be con-
more enables more options during statisti- dent that your result will fall within a
cal analysis. Other designs such as surveys given range of precision 95 percent of
require more participants than experimen- the time, use the value 1.96. (Recall
tal designs and challenge researchers to that in a normal distribution, 95 per-
compensate for the problem of low recent of the cases fall 1.96 standard
sponse rate; in contrast, qualitative de- deviation units from the mean.) If you
signs using observation and in-depth inter- wish to be 99 percent condent, a
views require fewer participants than larger sample will be required. (Here
experimental designs. Therefore, one sees
that the answer to the straightforward
question, How many should the sample
include? is never straightforward. BOX 15.3 Estimating the
This section outlines steps that are Standard Deviation
useful to beginning researchers in deter-
Suppose you wished to estimate the stan-
mining sample size (1) when the variable dard deviation of a variable such as the
you are making your estimates about is weight of male undergraduates. Try doing
measured at the ratio level; (2) when the the following steps:
variable is measured at the nominal level;
Step 1. Estimate the average weight of
and (3) when doing a power analysis as a
male undergraduates. Estimate: 160 lb.
means of estimating sample size. In addi- Step 2. Within what weight range would
tion to the methods presented here, in- you expect to find about two-thirds of
terested readers may wish to consult sta- the male students? Preliminary esti-
tistical texts for guidance on sample size mate: 145 to 175 lb.
(Fleiss, 1973; Kraemer and Theimann, Step 3. Because two-thirds of the cases
1987; Polit, 1996; Ritchey, 2000). Cohen represents 1 standard deviation on ei-
(1977) has generated a collection of tables ther side of the mean weight, this indi-
for estimating sample size for a variety of cates that our preliminary estimate of
statistical tests. The tables are based on the standard deviation is 15 lb.
Step 4. Double the estimate in Step 3 (30
anticipated effect size of the study vari-
lb). Does it make sense that about 95% of
ables, signicance level, and the power
all cases will fall within the range defined
associated with the test. by the mean twice the estimate at Step
3 (i.e., 160 30; range, 130 to 190 lb)?
1. Steps in Determining Sample If the band is not wide enough, how much
Size and Estimating a Ratio wider should it be? If it is too wide, how
Variable much narrower should it be? Suppose you
decide that it should be slightly wider, ar-
Sample size determination involves eight guing that about 95% of the cases will fall
basic steps, some of which are statistical between 125 and 195 lb. Divide this range
and some of which are pragmatic. For by 4 to get the estimated standard devia-
tion (70 4 17.5). Standard devia-
this illustration, we will focus on a situa-
tion estimate: 17.5 lb.
tion in which we are making estimates
about a ratio level variable. It is usually
the appropriate value will be 2.58.) For example, if the mean is measured
Most nurse researchers use the 95 per- in pounds, the standard deviation also
cent level in determining sample sizes. has to be measured in pounds.
With this level, you can be condent Step 4. Compute the sample size. Com-
that your sample mean will be within a pute the required sample using the fol-
given precision 19 out of 20 times. lowing formula:
Step 2. Select a major variable to deter-
mine sample size. Normally you focus (Confidence limit) (sd pop) 2
on the main dependent variable of the Required sample size

Accuracy
study in computing the required sam-
ple size. And if the dependent variable
is measured at the ratio level, it will be Suppose, for example, that you wished to de-
easier to determine sample size re- termine the sample size required to estimate
quirements. Such a variable may in- the average weight of male university gradu-
clude the percentage of surgical clients ates. You wish to be 95 percent confident (Z
readmitted with complications after 1.96) and be within 2 lb of estimating the
early discharge or an attitude score. true weight. The estimated standard deviation
You will need to estimate the popula- of the population is 14.0 lb. The values would
tion standard deviation (sd pop) for be plugged into the equation, as in:
this variable. This estimate can be
made by examining results from other
1.96 (14.0) 2
surveys or, failing that, simply by using Required sample size 188
2.0
common sense, noting the mean and
the range within which one expects to
nd two thirds of the cases (Box 15.3). The indicated sample size is 188 and should re-
Another suggestion for estimating an sult in an estimate of the populations average
unknown population standard devia- weight within 2 lb, being confident that the es-
tion is to approximate it by taking the timate will be within this margin 95 percent of
range of values (excluding extreme the time.
cases) and dividing this value by two.
This method generally provides a rea-
sonable estimate. Finally, it must be Step 5. Compute the sampling fraction.
acknowledged that in studies with a The sampling fraction is the sample
large number of variables involved, it is size in relation to the population (see
impossible to make claims, with any Monette et al., 1990, p. 149). Thus, if
certainty, about the precision of all the you have a population of 800,000 and a
variables. Most of them have unknown sample of 188, the sampling fraction
sampling distributions. would be:
Step 3. Determine the minimum preci-
sion that would be acceptable. Do you Sampling fraction Sample size/Population
want to be within 3 lb of the true mean 188/800,000
when estimating weight, within two 0.0002 (0.02%; go to
percentage points in predicting num- Step 7)
ber of clients with complications, or
within three points on a scale with a If the sampling fraction is less than 5
maximum score of 75? The measure of percent, go to Step 7. Alternatively, if
precision should be expressed in the the population was 1200 residence stu-
same units as the standard deviation. dents, the sampling fraction would be:
Sampling fraction 188/1200 Step 8. Adjust the sample size for cost
0.157 (15.7%; go to Step 6) and time factors. A nal step is to ad-
If the computed sampling fraction is just the sample to take time and cost
more than 5 percent of the population, into account. If one requires a sample of
an adjustment may be made in your 3000 but has only the resources to deal
sample size. To compute the adjust- with 1000, then one will have to do
ment, go to Step 6. some rethinking on the precision that
Step 6. Make the small population ad- will be possible. If, on the other hand,
justment. When the sampling fraction one has sufcient resources to increase
is greater than 5 percent, the required the sample size, then one usually does
sample size may be reduced according so because some additional precision is
to the following formula: likely to result.
Adjusted sample size n [1 (n N)] 2. Steps in Determining Sample

188 [1 (188 1200)] Size and Estimating a Nominal
Variable
188 1.157
162 A common situation for a researcher is to

be trying to estimate a population pro-
where n sample size estimated in portion for a nominal level variable. Poll-
Step 4; sters attempting to predict the vote for a
N estimated size of universe.
particular candidate in an upcoming elec-
See Monette, Sullivan, and Dejong, 1990.
tion may want to know the required sam-
Step 7. Are there sufcient cases for the ple size to provide an estimate of the
analysis? The required sample size likely vote and to be within perhaps 5 per-
must now be scrutinized to ensure that cent of the percent likely to support the
it will provide a sufcient number of candidate. Once again, the researcher
cases for the most complex analysis may want to be 95 percent certain that
that is to be done. For example, if one in- the estimate will be within this margin of
tends to do a series of cross-tabulation error. Suppose you believe the support
tables with three categories in each of for the given candidate is probably some-
the independent and dependent vari- where between 35 and 55 percent. How
ables, with a maximum of one control might you determine how many must be
variable with two categories, then one called in a phone survey to more pre-
is proposing a number of 18 cell tables cisely pin down the level of support?
(3 3 2). With 188 cases, the maxi- For nominal or ordinal variables, the
mum expected frequencies would be estimate may be calculated using the fol-
just over 10 cases per cell (188 18 lowing formula:
10.44). This would hardly prove ade-
n (Success) (failure)
quate and, therefore, one would want to
(condence limit Z value)2/(Accuracy)2
increase the sample size or rethink the
proposed types of analyses that are to where n required sample size;
be performed. Correlational techniques success is the estimated proportion voting
place less stringent demands on ones for the candidate;
sample so that, to the extent model test- failure is the estimated proportion not
ing can be done using those techniques, voting for the candidate;
fewer cases are required. 95% condence limit (Z score: use the value
1.96 for the test; 2.58 for the 99% condence then one must question the ethics of con-
limit); ducting the study as presently designed.
and accuracy is the range within which you Power analysis is used to estimate the
want the estimate to be. sample size needed to obtain a signicant
result and allow the researcher to con-
Thus, if we thought a reasonable esti- clude that the research hypothesis is
mate of our candidates vote would be 40 supported. Let us use a research example
percent and we wish to be 95 percent to demonstrate how power analysis is
condent ( i.e., our population estimate used to determine sample size.
will be within the accuracy parameter 19 Suppose you are interested in measur-
out of 20 times), we wish to be within 5 ing the impact of a health promotion pro-
percent of the vote for the candidate. gram on health attitudes. You recruit par-
These values are plugged into the equa- ticipants into the intervention group and
tion, and we note that we will need to the comparison group through the same
complete 369 phone surveys to provide process and randomly assign persons to
our answer. the two groups from the same pool of eli-
gible participants. You wish to test your
n (Success) (failure)
null hypothesis that states that there is no
(condence limit Z value)2/(Accuracy)2
difference in health attitudes between those
(0.40 0.60) (1.96)2
n exposed to the program and those not ex-
0.052
posed to the program. To test the null hy-
0.24 3.84 pothesis, you must determine how many
n
0.0025 participants you need in the intervention
n 369 and comparison group to detect differ-
ences in health attitudes. You conduct a
Thus, to be within 5 percentage points, power analysis to determine how many
it will be necessary to interview 369 per- participants you need for each group.
sons. If you wish to change the parameters, Power consists of four elements: signif-
simply adjust the values in the formula ac- icance level or alpha, sample size, effect
cordingly. You will note that gaining accu- size (ES), and power. If any three are
racy requires substantial increases in sam- known, the fourth can be calculated using
ple size. the power analysis formulas. Let us briey
describe each element. Alpha refers to the
probability of making a type I error (Chap-
3. Power Analysis and Sample ter 12). This is a situation in which you de-
Size tect a difference in health attitudes be-
tween the two groups even though no
Power is the ability of a study to identify difference exists. In this case, the null hy-
relationships or detect real differences pothesis is rejected when it should have
among variables. Power analysis is a pro- been accepted. The conventionally ac-
cedure used to determine the power of a cepted level for a type I error rate is 0.05.
statistical test; in other words, to deter- This means you are willing to accept the
mine the probability that an inferential risk of committing an error ve times out
statistical test will detect a signicant dif- of 100. A stricter alpha level such as 0.01
ference that is real or correctly reject a or 0.001 would decrease the power.
null hypothesis. The minimum accept- A type II error could result if you con-
able power for a study is 0.80. If one does cluded that there is no difference in
not have sufcient power to detect a real health attitudes between the two groups
difference that is present in a population, and a difference actually exists. As a re-
sult, you erroneously accept the null hy- may be conducted to estimate the ef-
pothesis. Beta refers to the probability of fect size of the relationships of interest.
making a type II error. A rule of thumb for A pilot study may be conducted to de-
selecting beta is that it be no more than termine the effect size. The results of
four times the value of alpha. Hence, if the pilot test can then be used to esti-
you are willing to take a chance of nding mate the effect size for the main study.
a difference in health attitudes 5 percent A dummy table analysis can be cre-
of the time when there really is no differ- ated to estimate effect size when no
ence, then you would be willing to risk prior information is available. In the
missing a true difference about 20 per- dummy table, the researcher calcu-
cent of the time (beta 0.20). The power lates the smallest effect size that
of a statistical test is determined by the would still be considered sufciently
following formula: 1 beta power (1 large to have practical or theoretical
0.20 0.80). The conventional standard value. For example, if an exercise pro-
accepted for power is 0.80. gram resulted in a 10 percent weight
Effect size is concerned with the loss for participants, it would be con-
strength of the relationship among study sidered worthwhile and cost effective.
variables. It is a measure of how false the An effect size estimate could be deter-
null hypothesis is, or in other words how mined based on this information.
strong the effect of the independent vari- Estimate on the basis of clinical expe-
able is on the dependent variable. Most rience or prior research whether you
nursing interventions have small effects; consider the effects small, medium, or
therefore, small samples are not likely to large. The qualitative labels can then
detect small differences or relationships be attached to quantitative effect size
between variables. Four strategies exist values to conduct a power analysis. Ef-
for determining effect size (Polit and fect sizes are usually small in new ar-
Sherman, 1990): eas of research. A medium effect size
should only be estimated when you
A review of the literature that has ex- can see the effect with the naked eye.
amined similar phenomena enables the In estimating effect size, it is prudent
researcher to estimate the strength of to be conservative because most nurs-
the effect size based on prior studies. If ing interventions have modest effects.
a substantial number of studies have
been conducted on the phenomena of The effect size is estimated differently
interest, a meta-analysis (a process for depending on the type of statistical test
statistically combining the results from that is performed. Table 15.3 presents Co-
multiple studies on a research topic) hens (1977) estimation of small, medium,
Table 15.3 Definitions of Levels of Effect Size for Common Statistical Tests
Effect Size Values
Test Small Medium Large
T test for two-group means 0.20 0.50 0.80
F test for k independent means 0.10 0.25 0.40
r (r 0) 0.10 0.30 0.50
F (R2 0) 0.02 0.15 0.35
Source: Adapted from Polit and Sherman (1990) and Cohen (1977).
and large effect sizes for four frequently Step 3. Consult a table of sample size re-
used statistical tests. quirements available in statistical texts.
After decisions have been made about Table 15.4 represents such a table and
alpha or signicance level ( ), population provides approximate sample size re-
effect size (
), and power (1 ), it is quirements for various effect sizes and
possible to determine the sample size. powers, and two values of (for two-
Let us return to our example and deter- tailed tests), in a two-group mean dif-
mine the sample size needed to test the ference situation. The table demon-
null hypothesis that there is no difference strates that for of 0.05, the estimated
in health attitudes for those exposed to a sample size needed for an effect size of
health promotion program and those in 0.40 and a power of 0.80 is 99 partici-
the comparison group. The following pants per groupthat is, 99 partici-
steps are required: pants assigned to the health promotion
group and 99 participants assigned to
Step 1. Estimate the population effect the comparison group for a total sam-
size (
). In a two-group situation in ple size of 198 participants. A sample
which the difference of means is of in- smaller than 198 has an unacceptable
terest, the formula for the effect size is probability of resulting in a type II error.
Step 4. If no previous research is available
12 to calculate the effect size, the next

step would be to estimate the expected
effect size based on experience, intu-
where
is the difference between the itive knowledge, and literature. By con-
population means divided by the popu- vention, the value of
in a two-group
lation standard deviation. But we do test of mean differences is estimated at
not know this information beforehand; 0.20 for small effects, 0.50 for medium
therefore, we must estimate the popu- effects, and 0.80 for large effects (Co-
lation means and standard deviation hen, 1977). Again, consult the table of
based on whatever information is avail-
able to us.
Table 15.4 Approximate Sample Sizes* to
Step 2. To estimate the population effect Achieve Selected Levels of Power as a Function
size, review the literature to see if ear- of Estimated Effect Size for Test of Difference of
lier studies examined the effect of a Two Means
health promotion program on health Estimated Effect
attitudes. One study is found that re-
Power 0.20 0.40 0.60 0.80
ported mean scores on a health atti-
tude survey for seniors engaged in a 6- 0.05
month health promotion program and 0.60 246 62 28 217
a comparison group of seniors. The re- 0.70 310 78 35 20
sults of this study were:
0.80 393 99 44 25
0.90 526 132 59 34
X1 (health promotion program) 5.70
X2 (no health promotion program) 5.30 0.95 650 163 73 41
(pooled standard deviation) 0.50 0.99 919 230 103 58
Calculation 5.70 5.30 0.40
5.70 5.30 *Sample size requirements for each group; total sample

0.80 size would be twice the number shown.
0.50 Estimated effect (
) is the estimated population mean
group difference divided by the estimated population
Thus, the value of would be 0.040. standard deviation.
sample sizes to determine that with an within 1 lb of estimating the true value of
of 0.05 and power of 0.80; the total the population, we would need a sample
sample size (number of subjects in four times as large as the one proposed
both groups) for studies with expected for a sample that would get us within 2 lb.
small, medium, and large effect sizes For example:
would be 784, 126, and 50, respectively.
Most nursing studies cannot expect ef- 1.96 (14.0) 2
Required sample size 753
fect sizes in excess of 0.50. The 0.20 to 1.0
0.40 effect size is a realistic expectation
for nursing studies. Polit and Sherman The original sample size required was
(1990) conducted a power analysis on 188 or one quarter of 753. Note that ac-
62 articles published in nursing re- curacy is referred to in statistics texts as
search journals and reported that the the condence interval. In the previous
average power for small effects was example with a sample size of 753, we can
under 0.30. The average effect size for be 95 percent condent that the true pop-
t-test situations was 0.35. ulation mean weight is within 1 lb of the
sample mean.
Power analysis is an advanced statisti-
cal technique that is often required when
applying for external funding. Several 2. Sample Size and Confidence
computer software packages are now Limits
available for performing power analysis
on personal computers. These include The relationship between precision and
programs such as POWER (Borenstein condence limits means that to move
and Cohen, 1988), EX-SAMPLE (Brent et from the 95 percent condence limit to
al., 1988), and SPSS for Windows. Begin- the 99 percent level, one may simply mul-
ning researchers should consult a statis- tiply the sample size by 1.73. Thus, in the
tician for advice in determining appropri- illustration on graduating male college
ate sample sizes. students, to be within 2 lb of the average
To perform this analysis using SPSS, see weight but to be 99 percent condent,
the WEIGHT command in an SPSS manual. one would need to multiply the original
sample size of 188 by a factor of 1.73 (188
1.73 325). A sample of 325 will be re-
D. OTHER SAMPLING ISSUES quired to be 99 percent condent that the
estimate will be within 2 lb of the true
Besides simply estimating the number of population mean. Conversely, to move
cases that will be needed in order to meet from the 99 percent condence limit to
the needs of the research project, there the 95 percent limit, the sample size de-
are a few other issues related to sam- termined for the 99 percent level may be
pling. These are explored here. multiplied by 0.58.
1. Sample Size and Accuracy 3. The Impact of Refusals

The relationship between sample size Although the evidence from studies exam-
and precision of estimations is a simple ining different response rates under con-
one. To double accuracy, sample size trolled conditions shows little variation in
must be quadrupled. In the illustration the descriptive accuracy independent of
concerning the weights of graduating response rate, tests of signicance as-
male college students, if we wished to be sume probability sampling techniques
and assume that there is no systematic sample is not wholly representative, there
bias in who chooses to complete the form. are techniques for weighting results to get
Such tests also assume that any measure- a better reection of the population whose
ment error is random. Because we cannot characteristics one is trying to estimate.
know what impact refusals and measure- However, when there has been a high non-
ment error have on our data, researchers participation rate, there is no guarantee
make every effort to get completed ques- that the sample will be representative for
tionnaires from all respondents. It is not other variables, even if it is representative
clear what conclusions one could legiti- for the age or gender distributions that
mately arrive at if one had a response rate have been checked.
of 20 percent. An alternative procedure that can be
used to conrm representativeness of a
4. Confirming Representativeness sample when the population characteris-
tics are not known is to take a random sam-
Steps, however, may be taken to conrm ple of the nonrespondents, interview them
that ones sample is indeed representative by telephone on key descriptive items, and
of the population about which one is at- then compare the respondents with non-
tempting to make generalizations. For ex- respondents on these items. If there is lit-
ample, one might compare the age, gen- tle difference between the two, you can be
der, and marital status distributions of condent that your sample is representa-
ones sample with known distributions for tive of the population from which it was
the population from census data. If ones drawn (Martof, 2000).
E X E R C I S E S
1. Suppose that you have census infor- 3. If the most complex cross-tabulation
mation indicating that the mean in- table you wish to run in an analysis is a
come of a population is $37,000 with a 2 2 2 table, what sample size would
standard deviation of $21,000. You produce 20 expected cases in each cell
wish to draw a sample that will be (assuming breaks at the midpoint)?
within $2000 of the true population
mean 95 percent of the time. How large
a sample would be required?
4. Suppose you wish to sample staff
nurses opinions of a hospital that has
400 staff nurses. You wish to be 95 per-
2. Assuming the same population values cent condent that you will be within 5
as in the above question, determine percent of estimating the true attitude
the sample size required to produce 99 of the nurses. How large a sample size
percent condence that you will be would you recommend? Show your
within $2000 of the true mean. computations for determining the sam-
ple size.
5. You wish to estimate the popular sup- mate that will be within 4 percentage
port of a local candidate for political points 19 out of 20 times. How large
ofce who is running ahead of the com- should your sample be?
petition. You are asked to do an esti-
Kish, L. (1965). Survey Sampling. New York: Evaluation and the Health Professions, 23(2),
Wiley. 212234.
Konrad, T.R., and De Friese, G.H. (1990). On Polit, D. (1996). Data Analysis and Statistics.
the subject of sampling. American Journal Stamford, CT: Appleton & Lange.
of Health Promotion, 5(2), 147153. Polit, D., and Sherman, R.E. (1990). Statistical
OConnell, A. (2000). Sampling for evaluation. power in nursing research. Nursing Re-
search, 39(6), 365369.

Part 6
QUANTITATIVE
DATA ANALYSIS
Q
ualitative data analysis is discussed within the description
of the various qualitative designs in Part 2 of this text
(Chapters 6 and 7). A special section on analyzing qualita-
tive data by computer is included in Chapter 6.
The three chapters in Part 6 deal with aspects of quantitative data

analysis. Chapter 16 presents various procedures for hand analyzing
data and introduces various computer concepts. Chapter 17 deals
with the issue of examining three- and four-variable causal models,
and Chapter 18 introduces four advanced statistical techniques for
data analysis. The principles of statistical data analysis are discussed
in Chapter 11 and Chapter 12.
519

Chapter 16
Starting the Data

Analysis
CHAPTER OUTLINE
A. Analyzing Data Without Using a 2. Data Entry

Computer 3. Creating and Saving an SPSS .sav File
1. Manual Analysis of Field Study Data 4. Data Cleaning Using SPSS
2. Manual Analysis of Experimental and 5. Steps in Analyzing Data for a Project
Survey Data 6. The 3M Approach to Selecting Analysis
Procedure
B. Analyzing Data Using a Computer
7. When You Get Errors and Warnings
1. Using SPSS for Windows (Version 10.0)
KEY TERMS
Cleaning the data Master table 3M Approach

Data entry Missing-value code Tally sheet
Error checking SPSS
521
522 QUANTITATIVE DATA ANALYSIS
H
aving collected the data, it is now formed by people who feel out of control
time to analyze them. We will ap- of the situation at hand. Signaling in-
proach this task in two ways. We cluded such common behaviors as fre-
begin with a method of hand tabulating quent staring or dgeting with their wrist
the results and then move on to a discus- watches, demanding that the charge nurse
sion of how computers may be used to have their family member seen immedi-
analyze data. ately by the attending physician, pacing
the corridor, or commenting on nervous
feelings and concerns to other waiting
room clients. Three hypotheses were pro-
A. ANALYZING DATA
posed for the study:
WITHOUT USING A
COMPUTER Females would be more likely to signal
than males.
Some analyses that we do with a com- The more people present in the wait-
puter could also be done easily by hand. ing room, the higher the amount of sig-
And sometimes it is even faster to analyze naling.
results manually if one does not have The closer the family members rela-
easy access to a computer. Chapters 11 tionship to the client, the higher the
and 12 present a number of statistics that amount of signaling.
can readily be hand computed if the data
The data were collected on tally sheets
set is small. This section provides a
such as those described in Chapter 7. The
method of collecting and processing in-
student observers indicated on the tally
formation from eld studies.
sheet whether or not family members in
the waiting room signaled. Also entered
1. Manual Analysis of Field Study on the tally sheet were the gender of each
Data family member, the persons relationship
to the family member ( i.e., spouse, par-
We will work through a manual analysis of ent, child, other) and the number of peo-
behavior in a hospital emergency waiting ple in the waiting room.
room. The student researchers who con- Figure 16.1 illustrates how the tally
ducted this naturalistic eld study were in- sheet was set up. Note that each line on the
terested in whether family members in an tally sheets records the information for
emergency situation would communicate one subject. Column 1 records whether
(signal) the nervousness they felt while the family member signaled in some way;
waiting for their family member to receive column 2 records the gender of the family
care. Signaling was dened as any ner- member, column 3 records the number of
vous and/or unnecessary actions per- people present during the observation;
Figure 16.1 Tally sheet for emergency waiting room study.

STARTING THE DATA ANALYSIS 523
and column 4 records the relationship of variables: gender, relationship to client,

the family member to the client. and number in the emergency waiting
Having recorded the information on the room. Each observation ends up in
sheets for 150 family members, the task oneand only onecell of the master
now is to analyze the results. One way to table. Thus a mother, who is one of ve
do this is to set up a master table for the people in the waiting room, and ner-
results, enter the observations into the vously paces up and down the corri-
master table, and then derive the results dors, would end up in one cell only.
tables from the master table. By using a Step 2. Transfer data to the master table.
master table, you will only need to make After observations have been com-
one pass through the raw data. From the pleted, the tally sheet data are trans-
master table you can create any other referred to the master table. Normally
quired tables showing the relationship be- each observation is entered into the
tween the independent variables and the applicable cell with a short stroke;
dependent variable. Here are some steps these are grouped together so that the
for hand analyzing this type of data. fth line entered into a cell is a line
across the other four. Each group then
Step 1. Set up a master table. A master reects ve observations falling into
table is generally set up on 81/2 by 14 the cell (Fig. 16.2 shows a master table;
paper. Using the paper so that the long see Fig. 7.2 for another one). Later
side is across the top, arrange the cate- these can be counted and a total for
gories of the dependent variable on the each cell entered and circled at the
left side of the sheet. In this case, the bottom of the cell.
categories Signal Given and No Sig- Step 3. Create the individual tables. To
nal Given are labels. Categories of the test the three hypotheses, three tables
independent variables are arranged would be created, each showing the re-
across the top of the master table. In lationship between signaling or non-
this case, we have three independent signaling and gender (Table 16.1), rela-
Figure 16.2 Master table for emergency waiting room study.

Table 16.1 Signaling Behavior by Gender

Gender
Male Female Total
Signaling Behavior N % N % N %
Signaling 44 75.9 51 55.4 95 63.3
Nonsignaling 14 24.1 41 44.6 55 36.7
Totals 58 100.0 92 100.0 150 100.0
Chi-square 6.449; degrees of freedom 1; probability 0.05 (reject null hypothesis).
tionship status, and number of people and when the observations were
in the waiting room (Table 16.2). To de- made (see Table 16.2).
termine the number of cases for each Lines to separate sections of the
cell of the new table, you simply add table: After the table title, draw two
up the count in the appropriate cells of closely spaced lines to separate the
the master table. For example, to nd table title from the body of the table.
out how many males versus females Single lines will be used to separate
were seen signaling, one simply adds other sections of the table.
together the count in all the cells for Independent variable across the top
males (the rst eight cells) and then of the page: Arrange categories of the
the cells for females (next eight cells). independent variable (group size)
These two numbers would then be in- across the top of the table, using eas-
corporated into the number of obser- ily understood category labels.
vations of males and females who sig- Dependent variable arranged on the
naled. Similar procedures are then vertical axis: Notice that the depen-
followed to get the count on the num- dent variable (signaling behavior) is
ber of males and females not signaling. arranged along the side of the page.
Step 4. Format the tables. Tables 16.1 Once again, use labels that are easy to
and 16.2 show the format for two of the understand and that will communi-
tables. Various features of these tables cate clearly to your reader what would
are to be noted: be included in each category.
Table title: The title reports the name Calculate column percentages: Note
of the dependent variable (signaling also that the percentages are calcu-
behavior) followed by the indepen- lated on each of the columns. The rule
dent variable (group size). Any con- is to calculate percentages toward
trol variables would follow. Some- the independent variable (adding up
times it is a good idea to report where to 100 percent on each category of
Table 16.2 Signaling Behavior by Number of People in Hospital Waiting Room

Number of People
Eight or Fewer Nine or More Total
Signaling Behavior N % N % N %
Signaling 72 72.0 23 46.0 95 63.3
Nonsignaling 28 28.0 27 54.0 55 36.7
Totals 100 100.0 50 100.0 150 100.0
the independent variable). For exam- and how the results should be inter-
ple, if we wished to calculate the ap- preted. For ease of interpretation,
propriate percentages for looking at arrange the dependent variable along
the relationship between gender (in- the vertical axis, with categories of
dependent variable) and signaling the independent variables on the
(dependent variable), we would cal- horizontal axis.
culate what percentage of males sig- Step 5. Calculate tests of signicance
nal or do not signal. We would then do and measures of association. Tests of
a similar calculation for the females. signicance and measures of associ-
In this way, we can compare the per- ation may be calculated manually.
centage of males signaling with the Chapter 12 provides examples of how
percentage of females doing likewise. such computations may be done. The
The reason you would not calculate results of these computations may be
the percentage toward the dependent entered below the table, as shown in
variable is that we are not so much in- Table 16.1.
terested in the distribution of signal- Step 6. Interpret the results. In interpret-
ing by gender (we might have a differing the results of the analysis, we typi-
ent proportion of males and females cally focus on just one of the categories
in our sample and therefore more of the dependent variable; in the case of
males might signal simply because the emergency waiting room study, we
there were more male family mem- would simply compare the percentage
bers present in the waiting room). of males who are seen to signal in the
Occasionally tables are created waiting room with the percentage of fe-
when the percentages are run across males who are seen to do likewise. In the
the page; this is usually done to ac- case of Table 16.1, we would note that
commodate situations in which there although 75.9 percent of the males sig-
are a number of categories of the in- nal, 55.4 percent of the females signal.
dependent variable and, in order to It is unnecessary to report the Not Sig-
have enough space, the table is nal categories because these are sim-
turned sideways. ply the remaining cases adding up to
To calculate the percentage, divide 100.0 percent. The writer of a report
the cell value by the marginal total should also note whether the differ-
(this is the column total if the depen- ences between the gender categories
dent variable is arranged on the verti- are statistically signicant, whether the
cal axis) and multiply by 100. In Table differences observed were in the direc-
16.1, the rst cell percentage would be tion predicted in the hypotheses, and
calculated by (44 58) 100 75.9 report on any measures of the strength
percent. These percentages should be of the association. A chi-square test of
reported to an accuracy of one deci- signicance could have been performed
mal point. The total should report the on this data using the procedures out-
actual totaleither 99.9, 100.0, or lined in Chapter 12 (see Box 12.4).
100.1.
More complex tables: More complex 2. Manual Analysis of
tables containing control variables Experimental and Survey Data
may also be created from the master
table. However, before constructing Using the manual analysis procedures
such tables, the materials in Chapter outlined in Chapters 11 and 12, it should
17 should be consulted in order to es- be possible for the researcher to analyze
tablish what model is being tested experimental and survey data without us-
ing a computer. Suppose you do a be- sonal computer (PC) or Macintosh (MAC)
tween-subjects experiment; you could or on a terminal that is connected to a
compute the dependent variable means computer network. Students need not un-
and standard deviations for your treat- derstand much about computers in order
ment group and for your control group to use themjust as you do not need to
(see Chapter 11 for information on calcu- know much about engines in order to
lating means and standard deviations). drive a car.
Box 12.5 may be used as a model to cal- A simple analogy that has been found
culate a between-subjects t test. If you useful by past students with minimal
have a within-subject design, refer to Box computer experience introduce readers
12.6. If you require a one-way analysis of to computers. (Cyber space cadets may
variance, use Box 12.7 as a model. If you skip the next paragraph.)
require an ANOVA, you will probably nd Think of entering a large, strange house.
that it will save time to enter the values It is a little frightening at rst, but after a
into a computer and have the machine, while you will nd that it is fun exploring it.
equipped with appropriate software such The rst thing you will have to do is get in-
as Statistical Package for the Social Sci- side. You will need a key, such as a secret
ences (SPSS), do the calculations. password, in order to enter. Let us imagine
Although survey type data could be an- that you have been provided with a pass-
alyzed by recording the information on a word so you can get into the house. Look-
master table, generally the number of vari- ing down the hall, you realize that there are
ables measured is substantial and it would many rooms in this house. The hall repre-
usually be advisable to learn how to use a sents the computers operating system;
computer to process the information. there are many procedures and utilities
However, if there are just a few variables available to the user in the various rooms.
measured, even with a large number of Having entered the house, you now have to
cases, it is possible to transfer the infor- decide which room to explore. The rst
mation to a master table and then derive room is marked Editor; here you can make
cross-tabular tables from it. But if you and modify les. The second room is
wanted to calculate means, standard devi- marked Storage; you can save les here
ations, or correlations among the vari- and get them later. A third door is marked
ables, the chore of hand computing the re- SPSS; nursing research data are analyzed
sults would be onerous unless there were here. A fourth door is marked Word Pro-
few cases in the study. In most situations, cessing; here reports are written and mod-
use of computer technology is advised. ied. A fth door is marked Printing; here
les get copied onto paper so that they will
be available to the user. There are many
B. ANALYZING DATA USING A other doors in the house, but these need
COMPUTER not be explored immediately by beginning
research students.
The remaining sections of this chapter Before suggestions are given for the
apply if a computer is to be used to assist entry of data into a computer, we need to
in the analysis. Although this process is introduce the idea of les. When data, or
relatively straightforward, there are a the text of a report, are entered into a
number of points to keep in mind so that computer, they are stored in a le. Com-
data entry and error checking will pro- puter les are just like those found in a l-
ceed efciently. ing cabinet, they contain information on
In most cases, the contemporary re- some topic and are arranged so that they
searcher will be working either on a per- can be retrieved easily. Many computers
use a two-word naming convention for may wish to consult them as you develop
les. The rst part consists of the le your facility in working with SPSS.
name and type, and the second part con-
sists of the le extension. The two parts
are separated by a . Typically, one is al- 1. Using SPSS for Windows
lowed eight character spaces for the le (Version 10.0)
name and type with the last three char-
You will probably be using the point-and-
acters before the . representing the le
click method of issuing the various SPSS
type and the three characters after the .
commands. A few general tips for this
the le extension. For the data collected
method include:
in a study, one might use the name JOAN-
DAT.SPS. If Joan were involved in a num- After each procedure is run, the re-
ber of studies, she would need to use sults are sent to the SPSS Viewer,
unique names for each of them. However, which will be displayed on your moni-
she would always use the same le type tor. You can save, edit, and print your
label for dataperhaps SPS. Later, we results from the SPSS Viewer.
will need to know about other kinds of To select one variable. simply click on
les, so additional le types will be intro- it when it is highlighted.
duced then. To select multiple variables from a list,
Contemporary nursing students usu- hold down the Ctrl button and click on
ally use computers that are using Mi- the variables.
crosoft Windows operating systems. Typ- To start using SPSS, turn on a com-
ically students control the computer by puter that is using a Windows operat-
using keystrokes and by manipulating the ing system and is equipped with SPSS
mouse to point and click on desired op- for Windows. After you are logged on,
tions. For purposes of this book, we will you are ready to start.
assume that the computer used is also
equipped with a Windows version of SPSS.
The Statistical Package for the Social a. To Begin an SPSS Session
Sciences is undoubtedly the most popu-
lar, and widely available, statistical pack-
Click on Start.
age used by social scientists. It was de-
Point to Programs.
veloped in the 1960s, has gone through a
Point and click on SPSS.
series of embellishments over the years, A screen similar to the one shown in
and now contains a great number of sta- Figure 16.3 should now appear on your
tistical and data manipulation proce- monitor. Note that there are a number of
dures. There are versions for mainframe blank cells displayed on the screen. You
computers (SPSSx), personal computers could start a new SPSS le by entering data
(SPSS/PC), and versions that take ad- at this point or you could access an exist-
vantage of the capabilities of Microsoft ing SPSS le. We will begin by accessing an
Windows. The version featured in this SPSS le that has the data entered.
Appendix is the one designed for Win-
dows and it is known as SPSS for Win-
dows (Version 10.0). b. To Access an Existing SPSS File
A number of manuals are available from
SPSS that detail the many additional fea-
Click on File.
tures and options that are available for the
Click on Open.
various commands. Some of these manu- A screen will now appear (Fig. 16.4)
als are listed at the end of Appendix A; you and you identify the location of the le
Figure 16.3 SPSS opening image.
and its name. You may need to specify The point-and-click method, which
where the le is located as well as its simply involves choosing options from
name. For our practice sessions, we will among those provided on a screen
assume the le is called C:\internl.sav. menu and clicking on them
After you enter the name of the required The syntax method, which involves
le (C:\internl.sav), click on OK. typing in SPSS commands
The blank boxes on the screen grid will
now be lled with numbers and you are c. Using the Point-and-Click Menu
now ready to begin analyzing the data set. Method
This is a data set collected from social re-
search methods students in Canada, New Assuming you now have the SPSS le
Zealand, and Australia. The questionnaire loaded (see Fig. 16.5), you now must de-
used to collect the information is included cide what procedures you wish to run on
in Appendix B of the text. Figure 16.5 dis- the data. Here are some of the basic
plays the screen after the C:\internl.sav things you might want to do (click on that
has been accessed. standard toolbar item and then click on
There are two methods of using SPSS: each item in the order listed).
Figure 16.4 SPSS: getting the file.
Figure 16.5 SPSS: filled screen.

ANOVA: Analyze/General Linear Model/ (when you access an .sps le, you are
Univariate automatically in the Syntax Editor).
CORRELATE: Analyze/Correlate/Bivariate Syntax les are used to contain in-
CROSS-TABS: Analyze/Descriptive Statis- struction commands that are then sent
tics/Cross-tabs to SPSS for processing.
DESCRIPTIVES: Analyze/Descriptive Sta- .spo an SPSS Windows output le
tistics/Descriptives (when you access an .spo le, you are
DISCRIMINANT: Analyze/Classify/Discri- automatically in the SPSS Viewer Edi-
minant tor). An output le contains the results
FACTOR: Data Reduction/Factor of your analyses. You can modify, print,
FREQUENCIES: Analyze/Descriptive Sta- or copy these les to a word processor.
tistics/Frequencies
GRAPHS: Graphs/Scatter/Dene d. Using the Syntax Method
MANOVA: Analyze/General Linear Model/
Multivariate In the syntax method, the user types in the
MEANS: Analyze/Compare Means/Means desired commands. This method appeals
SPEARMAN CORRELATION: Analyze/Cor- to experienced SPSS users who know SPSS
relation/ Spearman commands and the command structure.
REGRESSION: Analyze/Regression/Linear New users will probably nd it easier to
(Method Backward) use the point-and-click approach. To use
RELIABILITY: Analyze/Scale/Reliability the Syntax method, follow these steps:
(ModelAlpha)
t TEST PAIRS: Analyze/Compare Means/
Click on File
Paired-Samples t Test
Click on New
t TEST GROUPS: Analyze/Compare Means/
Click on Syntax
Independent Samples t Test Now a blank screen should appear and
you simply type the commands onto the
Some other basic procedures that you
screen. Remember that after each com-
will need to know are:
mand is completed, you must end the line
COMPUTE: Transform/Compute with a period (.) Having entered the com-
IF: Transform/Compute mands, you can then position the cursor
LIST: Use Syntax Editor to issue command (the ashing vertical line) where you
RECODE: Transform/Recode/Into Differ- wish SPSS to begin processing. To have
ent Variables SPSS process the commands, do this:
SORT CASES: Data/Sort Cases
SELECT CASES: Data/Select Cases
Click on Run
Click to choose from among options
Appendix A discusses each of the above shown
procedures and provides guidelines for in-
It is a good idea to maintain a copy of
terpreting the output that results from
the syntax les used. If data errors are
each analysis. Note that the SPSS Windows
found or if additional data are added to
version uses the following le types:
the le later, then the syntax les can sim-
.sav an SPSS windows system le (when ply be resubmitted to SPSS without hav-
you access a .sav le, the data will show ing to go through the whole process a
on the SPSS Data Editor screen). A sys- second time. When you leave the pro-
tem le contains all of your data and all gram, you will be asked if you wish to
of the labels describing the variables in save a copy of the syntax le; if you re-
your le. spond yes, you will also be given an op-
.sps an SPSS Windows syntax le portunity to name the le.
2. Data Entry Guideline 2. Code any uncoded ques-

tions. If there are questions that have
Data entry is simply the process of trans- not been precoded (e.g., occupational
ferring the information collected in a codes or open-ended opinion ques-
study to a computing device. In survey re- tions), they should now be coded and
search projects, which include a large the values written on the question-
number of variables, it is usually best to naire (or form) in the margin next to
enter the data into a word processor the question. In the case of occupa-
such as Word or WordPerfect. Later you tional prestige codes, they would be
can highlight the text, and using the looked up and a value entered onto the
Edit/Copy command, paste it into an SPSS questionnaire.
syntax le. Alternatively, data may be en- Guideline 3. Do a column count. Using a
tered directly into the Syntax Editor that blank copy of the questionnaire (or
is part of SPSS for Windows. Using the form), go through it, and, opposite
Syntax Editor has the advantage of being each question, indicate where it will be
able to enter all the data on one line for entered onto the terminal screen. Typ-
each case as opposed to having to use ically, screen widths are 80 columns
two or more lines if you have quite a few wide. This means that you can enter 80
variables. The Syntax Editor approach is digits across the screen before you run
the preferable one, especially if you have out of space. Thus, if you were entering
a large survey to process. your data into a Word le, you would
We will assume that you have gured need to start a second line of data for
out how to use one of the above ap- the questionnaire if it exceeds 80 char-
proaches and you are now ready to begin acters. If you are using SPSS for Win-
data entry. The following rules provide dows, you need not worry about the
guidelines for data entry if you are creat- screen width; the Syntax Editor will al-
ing a separate raw data le. low you to use as many columns as is
necessary for your data set.
The rst three columns will be used
a. Creating a Separate Raw Data File for the ID number of the questionnaire,
and the fourth column will be used to
Guideline 1. Number the questionnaires, identify the record number for the re-
forms, and data sources. Question- spondent. (This is required only for
naires or data collection forms are num- questionnaires that require more than
bered, normally beginning with 001 (or 80 columns to enter the data. Record #1
0001 if there are more than 1000 cases). refers to the rst line of data, record #2
This ID number is written on the front the next set of data for the respondent,
page of the instrument (usually the top and so forth. The ID number is repeated
right-hand corner). This number will be on each line of data followed by the
entered into the computer and pro- record number.) If a question can have
vides the only link between the particu- values between zero and nine, then one
lar form and the data that are entered column will be required to record the
into the computer le. If an error is data; if the values range up to 99, then
found in the data, the original form can two columns will be required; if the val-
be located by using the ID number, the ues range up to 999, three columns will
information can be checked, and the be required. On the right-hand margin
error can be corrected. Keep the ques- of the questionnaire (or form), these
tionnaires or forms sorted by their ID values are recorded. The right-hand
number. margin might look something like this:
Variable Columns tion for each variable will t onto one

80-column line; each questionnaire
ID 13
will occupy one line of data. If more
Record # 4 than one record (or line) is required,
Blank 5 then a second record is used, with the
rst three columns being saved for a
Gender 6 repeat of the ID number. The fourth
Yr of Birth 78 column will contain a 2, meaning that
it is the second record for the ques-
Income 914 tionnaire. In the case of a question-
Occupation 1516 naire requiring two records, the data
might be set in a way similar to the ex-
In a short questionnaire, the informa- ample below:
0011476989232333 23222111222111 222456345644556666663339

00124534376767221122211 1122299122876321453213345222
0021456338992212 21222113222112 223341100334455654432211
00222343434445857463748 4985050595211222333423334444
Note the rst four columns. The rst partner. Inexperienced computer users
three contain the ID number (which will nd that data entry will be much
has also been written on the question- easier and will be done with fewer er-
naire), and the fourth column refers to rors if two people work together on en-
the record number. What we have here tering the data. One person can read
is respondent 001 rst record; then the the numbers off the questionnaire, and
same respondent 001, second record; the partner can enter the values into
this is followed by respondent 002, the computer.
rst record, followed by respondent Guideline 5. Leave internal blanks to
002, second record. It is not critical mark a new page. Errors can be iden-
that the ID numbers be in sequence, tied more quickly if, after the end of
but the case must be togetherrecord each page in the questionnaire, a blank
1 must be followed by record 2 for is left in the data. The column of blanks
each case The reason we enter the ID must always line up when the data are
number and then the record number is printed out; if the blanks do not line
that most computers sorting routines up, then an error has been made and
will be able to sort your le with one must be corrected. The rst six lines of
simple command. The blanks in the a data set are listed in the following ex-
data set are discussed below. ample, showing how the blanks line
Guideline 4. Enter data with the help of a up:
page 1 page 2 page 3 page 4

00112435445523 211212333448976999 123232111123212 23411112
00211231231234 322213222122232212 432235333211232 34532432
00314321234443 233445324433321212 332122341122233 23323222
00413444688822 1222221122321233219 234433221222321 34521211
00513335543221 223332221114124211 332111422111212 32121223
00614322345432 231143321122122121 112211222111231 22132454
Note that the fourth line of data con- ular columns in the data; each ques-
tains an error; by examining the listing, tion, therefore, must be in the identical
it becomes apparent that the error is position in all lines of data. (Actually,
somewhere in the data entered from most programs allow the use of a
page two of the questionnaire (there space or a comma to separate ques-
are too many digits in the space re- tions but because error detection is
served for page two). The introduction easiest when questions occupy the
of blanks to mark the end of pages is identical columns in each question, we
enormously helpful in error detection, will use the same-column method.)
so researchers should be sure to use Guideline 7. Document research deci-
them. sions. When data are being entered into
Guideline 6. Simplify missing value a computer, a number of decisions are
codes. If respondents do not answer a required. A questionnaire may have a
question or if a question is not applic- lot of missing data, and the person en-
able, a missing value code is used. tering the data might decide that it
These should be kept as simple and as would be better to discard the ques-
consistent as possible. When possible, tionnaire. When this happens, the ques-
use the values 9, 99, or 999 for vari- tionnaire or form should be set aside
ables that require one, two, or three and fellow researchers should be con-
columns, respectively. Suppose a resulted to determine whether that data
spondent is asked to indicate gender should be used. Cases may arise in-
as either male or female and leaves the which a respondent has checked two
question blank. Instead of leaving the items on a question when only one was
column blank, a 9 would be inserted expected. Generally, these discrepan-
into the appropriate column. See Ap- cies are rare enough for it to be appro-
pendix A for additional details on miss- priate to ip a coin to decide which of
ing values. the two will be taken. (One would not
Although it is possible to use alter- systematically take the highest placed
nate codes (when a question is not an- item on the list because that would sys-
swered or a question is not applicable), tematically bias the results toward
such as 8 for not answered and 9 for those items listed rst.) When such a
not applicable, such discrimination decision has been made, the decision
should only be made where it is known should be marked clearly on the ques-
that the information will be used later; tionnaire and initialed by the person
if it is not going to be used, keep mat- making the decision. With such docu-
ters simple by using the single code 9 mentation, it is then possible for others
to cover both cases. In occupational to check through the data and under-
codes, it may be necessary to provide stand what coding decisions were made
special codes for housewife if values and by whom.
are not provided in the occupational Guideline 8. Code for information not
code system being used. For the nine- on the questionnaire. Frequently,
point Likert attitudinal scales, 0 is used there will be information that is not
for the missing value code. coded on the questionnaire that
A code must be entered for every should be appended to the data. It is
question, even if it is not answered. always recommended that provision
The reason for this is that when in- be made to include a code to identify
structions are given to the computer, who collected the original informa-
each question is identied with partic- tion, who did the coding on the ques-
tionnaire, who did the data entry, and b. Point-and-Click Raw Data Entry
any other information that may be
useful in the later data analysis or in If the data are being entered directly into
error checking. The reason why it is SPSS, click on File/New/Data, move the
helpful to code who did the coding is cursor to the cell you wish to start with
that if there are systematic differences (the cell will be highlighted), and enter
between coders in dealing with differ- the data for each variable moving across
ent questions, then the cases dealt the row. The value will show on the
with by each coder can quickly be screen just above the matrix. To move
identied and compared. Appropriate from variable to variable after the value
corrections can then be made without has been entered, either hit the Tab key
too much trouble. or the right arrow key to move from case
Guideline 9. Use double data entry. When to case and use the down arrow key after
resources allow, it is recommended that the value has been entered.
different individuals enter data twice. A ( i ) Inserting a New Case
computer program can then be used to If you wish to add a new case, position the
compare the two les, agging any dif- cursor on the case below where you want
ferences between the les. This tech- to insert the new case. On the toolbar,
nique is extremely helpful in reducing click on Data and then Insert Case. Enter
data entry errors. the data for the new case.
Guideline 10. Save the data. When data
entry is complete, save the le as an ( ii ) Inserting a New Variable
ASCII or DOS le ( if using WordPerfect, If you wish to add a new variable, position
use Ctrl F5 and Save; if using Microsoft the cursor on the variable after the spot
Word, use the Save As command; on where you want to insert the new vari-
the Save As window, click on Text able. On the toolbar, click on Data and
Only). Alternatively, you can highlight then Insert Variable. Enter the data for
the data and do an Edit/Copy on it and the new variable for all the cases. Note
then do an Edit/Paste of the data into a that if you do not enter any value for a
syntax le. If you have entered the data case, the system will add a .indicating
into a syntax le in SPSS, save it as an that this case has a MISSING VALUE.
.SPS le, which will allow for easy entry
into the Syntax Editor. (Saving the le 3. Creating and Saving an
is easy; when you exit from the pro- SPSS.sav File
gram, SPSS will ask you if you wish to
save the le. Simply respond yes and If you have a modest to large survey to
provide a name for the le. The ma- process, it is probably worth learning how
chine will automatically append .SPS to use the Syntax Editor to enter the com-
to the le name, indicating that this is mands to dene the variables, attach la-
an SPSS syntax le.) bels, and indicate the missing value codes.
Guideline 11. Error checking. This is a Variable lists may be used to give many
process used to locate and correct er- variables the same labels (e.g., yes or no
rors in the data before data analysis. questions). Lists may also be used to iden-
Programs are available for identifying tify variables with the same missing values.
out-of-range entries and for locating This approach to data entry is described
non-numerics, such as O for 0, or a l below under the section called The Syn-
(lowercase letter l) instead of a 1 tax Method. If the data being analyzed in-
(one). If you have access to such a util- volve just a few variables ( 10 or 15), the
ity program, use it. direct entry method may be used comfort-
ably. This approach will be described un- Hint: In studies with a lot of variables
der the Point-and-Click Method section. (as in most surveys), it is a good idea to
use the question numbers for variable
names. Because the names must begin
a. Point-and-Click Method for with an alphabetical character, it makes
Creating and Saving an SPSS sense to call them v1, v2, v3, and so on.
System File For multipart questions, use v4.1, v4.2,
v4.3, and so on. When analyzing data, you
Hit File, then New, then Data. A screen
can quickly nd the variable names by
with blank cells will now appear. Click on
just looking at the questionnaire.
Variable View tab at the bottom of the
data screen. Your rst task will be to ( ii ) Missing Values
name each variable. Usually we start with See Guideline 6 for suggestions on the
an ID as the rst variable name and it will treatment of missing values. It is best to
give us a place to enter the number for use 9, 99, 99.9, and so on to indicate val-
the questionnaire (or observation form); ues to indicate a refusal to answer a ques-
these should correspond to the number tion or to indicate a question that does
written in the top left-hand corner of the not apply to a respondent. If it is going to
questionnaire. Figure 16.6 shows the ap- be relevant to distinguish the reason for
propriate window. the missing value, other code numbers
may be applied. If you are using the syn-
( i ) Dening and Labeling Variables tax approach to dening the missing val-
At the bottom of the Data Editor screen, ues, a variable list is made up for each
there are tabs for Variable View and for missing value. If the point-and-click ap-
Data View. Click on the Variable View tab, proach is used, declare the missing value
and the screen will display the variables when the variable labels are attached.
along the vertical axis of the screen. The Hint: In large surveys, if the point-and-
denable characteristics are arranged click approach is used to dene the vari-
along the top of the screen. By clicking on able, it speeds up the process to put on
one of the characteristics, Labels for ex- value labels and missing values after the
ample, you can then attach the labels for .sav le has been created. Simply put on
that particular variable. On the gender the variable names and labels, save the
variable, you would attach the labels le, enter the Syntax Editor, and enter the
Male and Female. VALUE LABELS and MISSING VALUES,
Figure 16.6 SPSS: defining variables.

and then save the output. Because vari- SPSS command structure, the main differ-
able lists may be used for VALUE LABELS ence between what you are familiar with
and MISSING VALUES, it is much quicker and the Windows versions is that you
to do these in the Syntax Editor. have to remember that after each com-
Hint: You can change the default set- mand is completed, you must end the line
tings for variables (type of variable, size of with a period. After entering the com-
eld) by clicking on the Variable View mands, you can then position the cursor
(bottom of screen), select the case you (the ashing vertical line) where you wish
would like to have similar settings, do an SPSS to begin processing. To have SPSS
Edit/Copy, then highlight the variables process the commands do this:
you want it applied to and do an Edit/Paste
and those denitions will be applied to the
Click on Run
cases that that have been highlighted. For
Click to choose from among options
shown
example, if you are dealing with survey
data and most variables are one character It is a good idea to maintain a copy of
wide, you would reset the defaults to show the various syntax les used. If data er-
Numeric 1, MISSING VALUES as 9. Only rors are found or if additional data are
when you encounter 2 column variables added to the le later, then the syntax les
would you have to redene the settings for can simply be resubmitted without hav-
these variables. ing to think through the whole process a
second time.
( iii ) Saving or Updating the SPSS
Box 16.1 shows the command struc-
System File: The .sav File
ture that will identify, name, and label all
Whenever you wish to save or update
the variables in a study. A key shortcut to
your SPSS system le, click on the toolbar
remember is that data lists may be used
File and click on Save. You will be able
when several variables have the same la-
then to specify where you want it saved
bels. This saves the researcher from en-
and change the name of the le if you
tering labels more than once. Examples of
wish. Anytime you compute new vari-
the various commands necessary to de-
ables or transform old ones, you may
ne an SPSS system le are shown in Box
want to save the changes, so remember
16.1. This method is probably the quick-
to update the le with the Save proce-
est one for data sets involving many vari-
dure. Be certain to maintain backup
ables and many cases. Smaller data sets
copies of your .sav le on disks as well as
can be entered directly using the point-
on the hard drive of your computer.
and-click method discussed in the pre-
ceding section.
b. The Syntax Method of Creating
and Saving an SPSS System File ( i ) Saving or Updating the SPSS
System File: The .sav File
In order to use the syntax procedure to Whenever you wish to save or update
create a system le, access the Syntax Ed- your SPSS system le, click on the toolbar
itor: File and then on Save. Then you will be
able to specify where you want it saved
Click on File
and change the name of the le if you
Click on New
wish. Anytime you compute new vari-
Click on Syntax
ables or transform old ones, you may
Now a blank screen should appear. Sim- want to save the changes, so remember
ply type the commands onto the screen. to update the le with the Save proce-
For researchers who are familiar with the dure. Be certain to maintain backup
BOX 16.1 Sample Syntax Commands to Create an SPSS System File
Title system file creation, Nursing Faculty Review, Winston

Data list file C:\oia\nursing\nursedat.sps Records 1
/id 15 v1 8 v2 9 v3 10 v4 12 v4.1 to v4.7 1319 v5 20 v6 22 v8 23 v9.1 to v9.6 2429 v10 to v12
3032 v13 34 v16 35 v17 36 v18.1 to v18.5 3741.
Variable labels id identification number
/v1 Year of Graduation
/v2 Program
/v3 Honors Thesis
/v4 Further Education
/v5 Current Employment
/v6 Employed in Field of Choice
/v8 Different Field of Study
/v9.1 Quality of Content for Nursing/ v9.2 Quality of Instruction for Nursing
/v18.4 Audio-Visual Aids
/v18.5 Quiet Study Space
value labels /v1 0 1995 or earlier 1 1996 2 1997
31998 4 1999 5 2000 6 2001 7 2002 8 2003
/v2 1 Major 2 Advanced Major 3 Honors
/v4 0 Yes 1 No
/v4.1 to v4.5 0 No 1 Yes
missing values v1, v2, v3, v4, v4.1 to v4.5, v5, v6, v8, v13 (9)
/v4.5, v4.6, v9.1 to v9.6, v10 to v12, v16, v17, v18.1 to v18.5 (0).
Frequencies var v1.
copies of your .sav le on disks as well as that le on a printer. The printout

on the hard drive of your computer. should be examined and any irregular-
ities circled. Check the following:
All lines must end in the same col-
4. Data Cleaning Using SPSS umn. If there are too few or too many
columns in a line, it means that ques-
After the data have been entered into the
tions were either missed or entered
computer and saved in a le, it is time to
twice. Mark discrepant lines so that
begin the systematic search for errors so
they can be checked against the orig-
they can be corrected before analysis of
inal questionnaire and the errors lo-
the data begins. This process is referred to
cated. The le must be rectangular
as cleaning the data. If you have special-
with no ragged edges.
ized software, then you would proceed by
using this program. (The instructors man-
Any internal blanks must line up ver-
tically. If end-of-page blanks have
ual contains a program for detecting non-
been used, one can quickly identify
numerics, and out-of-range entries.) If you
situations in which there have been
do not have access to such data cleaning
too many or too few entries made
programs, the following steps will help
from a particular page on the ques-
you detect errors in the data.
tionnaire. Once again, mark any dis-
Step 1. Assuming you have entered the crepancies.
data into a le, the rst job is to list When there are two or more les to
be checked, ensure that the le values included, the case must be iden-
lengths are equal. For example, if 165 tied and the data corrected before
questionnaires have been completed, you proceed with further analysis. To
there must be 165 lines in each of the locate the problem case (or cases), go
les. to the SPSS grid with your data listed
Proofread ID numbers (columns one and click on the column (top of grid)
through three) to ensure that there where the variable name is identied;
are no repeated or missing ID num- the whole column should then be high-
bers. Again, mark any errors. (Some lighted. If you want to nd a 3 in your
cases may have been withdrawn, so gender variable, click on the icon for
those ID values will be missing.) nd (binoculars) and tell the machine
Finally, there must be no blank lines to nd 3. The machine will now high-
or partial lines left in the data set. light the problem case. Note the case
(SPSS reads a blank line as a case and number listed under ID, go to your
assigns a missing value to all vari- sorted pile of questionnaires, and nd
ables. If you nd that you always the questionnaire with the same ID
have one or two missing cases even number and note the value that should
though you know there should be be included for gender. Correct the
none on certain variables, your prob- value by highlighting the gender cell
lem may be that you have blank lines for the case in error by entering a 2 or
somewhere in your data le.) a 1 as appropriate. While doing this
Step 2. When the le that includes the vari- check, make certain that the missing
able labels and other le denition com- values are properly agged as missing.
mands is run, SPSS will alert you to any In the case of the gender variable, 9s
inappropriate non-numerics in your probably were used as the missing
raw data le. The output will indicate value code, make certain any 9s are
which case has the problem. Note the agged as missing on the frequency
case numbers, examine your data, and distribution. But note that if there are
make corrections to the data. Some- no missing cases for a variable, the fre-
times you will have included non- quencies will not list the category for
numeric data intentionally but may missing values.
have forgotten to indicate this on your Continue checking the frequency dis-
DATA LIST commands. Remember, you tributions looking for out-of-range en-
can use non-numerics, but when identi- tries, correcting them as you go along. It
ed on the DATA LIST command you is a good idea to keep a record of these
must put an (A) after the column num- changes and then go back to your raw
bers for the variable, as in v27 1718, data le and correct that le as well.
v28 1940 (A), v29 to v31 4143. Step 4. The ratio variables should be
Step 3. On all nominal and ordinal data, listed on the descriptives window (An-
run the frequencies procedure. Exam- alyze/Descriptive Statistics/Descrip-
ine the output (you need not print it tives). Check the lowest and highest
out at this point because you are sim- values for each variable. Are they
ply scanning it for errors), looking for within range? Note especially that the
any categories that are not labeled or MISSING value is not listed as the high-
any out-of-range values for that vari- est value. If it is, you have made an er-
able. For example, on the gender vari- ror in your missing values. Problem
able, there should be only two cate- cases may be identied by bringing up
gories, one for male respondents and the SPSS data grid, highlight the vari-
one for females. If there are any other able that has an error, and use the
binoculars icon to nd the value you testing the models. Do the procedures
wish to nd. Corrections may be made to test the various relationships indi-
directly on the SPSS grid, but should cated by the model. Note that you may
also be made in the raw data le. need to RECODE some variables. If
Hint: If you make the corrections in the most of the relationships are examined
raw data le, you can resubmit the using one method (say MEANS), you
MAKER le (i.e., the one containing the should probably do all of them using
data list, variable labels, and so on) and that procedure, even though a few
this will update your SPSS System le. could be examined using correlations.
Step 5. After correcting the errors, you Enter the commands to do the basic
are now ready to proceed with your runs and, if recoding of variables has
analysis. been done, retain the recodes for fu-
ture use by updating your system le
by clicking on File/Save.
5. Steps in Analyzing Data for a Hint: When doing RECODES, create a
Project new variable to contain the informa-
tion and add an r to the name of the
When doing your analysis, it is a good
original variable. For example, if you re-
idea to use the Paste command to save
code v9, call the new variable v9r; us-
the procedures you have run in a Syntax
ing this method will remind you that
le. The reason for this is that if you have
any variable name with an r ap-
errors in your data and it becomes nec-
pended to it is a recoded version of the
essary to re-run a job after corrections
original variable.
have been made, you can then quickly re-
Step 4. If you have any intervening or
submit your jobs without having to re-
sources of spuriousness models, these
think your whole analysis.
should be processed. Review Chapter
Step 1. Be certain to have the question- 17 for suggestions on interpreting re-
naire or the recording form on which sults.
you have recorded the column num- Step 5. After you have assembled the
bers for each variable, printouts of computer output, you should consider
your FREQUENCIES (nominal and ordi- how you will present the results. (See
nal variables) and DESCRIPTIVES (ra- various summary tables in Chapter 19
tio variables) with you. If you have not for possible ways of setting up your ta-
run these analyses, run them. bles for presentation.) Note that it is
Step 2. If you need to construct any in- helpful, particularly in small sample
dexes, it should be done at this point. studies, to indicate whether the trends
Review the procedures for creating in- in various relationships go in the di-
dexes (see Chapter 13); if you are using rections predicted.
RELIABILITY, click on the appropriate Step 6. Write your report in sections, fol-
icons to do the job, and determine the lowing the outline suggested in Chap-
items that are to go into the index. Do ter 19.
a COMPUTE to create the index. Add
VARIABLE LABELS (and VALUE LA- 6. The 3M Approach to Selecting
BELS, if appropriate). Save the SPSS Analysis Procedure
(.sav) le so that the next time you sign
on, the new variable (or variables) will As a way to remember how to decide
be available. which procedure to use to analyze any re-
Step 3. Examine your causal models. Se- lationship, use the three steps in the 3M
lect the appropriate procedures for approach.
Step 1. Model. The rst step is to decide are three levels of measurement: nom-
which variable is to be treated as the inal, ordinal, and ratio.
dependent variable and which one as Step 3. Method. Having identied the
the independent variable. Because the measurement levels, the third step is to
researcher typically is trying to under- examine Table 16.3 to determine which
stand what causes variations in a de- method should be used to examine the
pendent variable, common sense relationship. The methods listed are
alone can generally determine which the names of SPSS procedures.
variable should be designated as the Having identied the appropriate
dependent variable. However, there procedure to use, you are now ready to
are other cases where it is not obvious. proceed with the analysis. The reader
In such cases, try to decide which vari- may wish to review Table 16.3, which
able occurs last in a temporal se- indicates the appropriate analysis pro-
quence. It is entirely possible that the cedure for given levels of measurement
two variables mutually inuence one in the independent and dependent vari-
another. If this is the case, one will ables. Appendix A presents the basic
nonetheless have to be designated as procedures along with the most com-
the dependent variable. It is a good monly used options and statistics. For
idea to diagram the relationship using full details of SPSS procedures, consult
greater than () and less than () the appropriate SPSS manual listed in
symbols as well as a line with an arrow the suggested readings section at the
pointing toward the dependent vari- end of the Appendix A.
able.
7. When You Get Errors and
Warnings
>A >B If you are using the Syntax method of en-
tering SPSS commands, expect to get er-
ror and warning messages when you run
Step 2. Measurement. The second step is SPSS jobs. Some of these errors lead to
to identify the level of measurement at- the termination of the job; other less se-
tained in the dependent and in the in- rious ones (warnings) simply lead SPSS to
dependent variable. Recall that there ignore an instruction. SPSS has error
Table 16.3 SPSS Procedures for Multivariate Analysis

Independent Variable
Dependent Nominal Ordinal Ratio
DISCRIMINANT
SPEARMAN SPEARMAN
DISCRIMINANT
Ratio MEANS MEANS CORRELATION GRAPH
ANOVA ANOVA PARTIAL CORRELATION
T TEST T TEST MANOVA
REGRESSION FACTOR
checking routines that identify the error ginning many of the subsequent errors
immediately after the line on which the will be xed. For example, you do a
error occurs. COMPUTE to create a new variable
Later in this section, Table 16.4 pre- known as TOTALS, but on all your sub-
sents a listing of some of the more com- sequent commands you refer to the
mon errors and warnings and what pro- variable as TOTAL. Each time the ma-
cedures might be taken to solve them. chine encounters TOTAL, it will give
But before you examine this table, we will you an error. If you change the name of
review some basic strategies in error de- the variable on the COMPUTE state-
tection. ment to TOTAL and resubmit the job,
all the subsequent references to TO-
Tip 1. Expect errors. The rst thing to re- TAL will be correct. Sometimes it is a
alize is that you will make errors. Even good idea to resubmit a job after you
after many years of using SPSS, you have xed the rst few errors to see
will continue to make errors. SPSS er- how many have been eliminated. Pro-
rors are just part of your life as a data ceed then by again xing the rst er-
analyst. rors that show up and then resubmit
Tip 2. Examine error and warning mes- the job until you get an error-free run.
sages carefully. The errors and warn- Tip 5. Stuck? Reenter the command line.
ings are listed on the output imme- Sometimes you can stare at an error
diately after the problem has been line and cannot see what is wrong with
encountered. To assist the researcher, it. Perhaps you used a letter O rather
the characters or symbols creating the than a number 0, as in v1O when you
problem are quoted. A key point is to meant v10. SPSS does not have a vari-
carefully examine the description SPSS able called v1O, but it is hard to see the
provides of the error that will assist difference on some monitors. Do not
you in identifying your mistake. waste a lot of time; just reenter the
Tip 3. Make certain SPSS is accessing the command line.
necessary les. A common error a new Tip 6. Examine results on screen before
users often encounter is that SPSS is sending to a printer. Sometimes you
not able to access the le (or les) re- may inadvertently do something very
quired. This means that you may have silly and produce an enormous, mean-
an error on your Get File command ingless output le. Suppose, for exam-
line. In SPSS, the computer will do an ple, that you are running CROSSTABS
error search and identify all the prob- and you are examining the relationship
lems it has with the instructions. Of between two variables with a control for
course, if the machine has not been occupational prestige of the respon-
able to nd your system le (*.sav), dent. In error, you failed to use the
every variable you mention will pro- recoded version of the occupational
duce an error because the computer prestige variable and you accidentally
has not found a le containing the produced a monster output le because
named variables. The rst thing to do for each occupational code (there may
when you get an error at the beginning be 70 or 80 distinct numbers), the com-
of a job is to check to see that the ma- puter generated a table showing the re-
chine was able to access the necessary lationship between your main variables.
les for the task. If you look at the results on the screen
Tip 4. Fix rst errors rst. When correct- before you print them, you will avoid
ing errors in a syntax command set, wasting paper on foolish output.
begin with the rst errors identied. Tip 7. Double check variable list. When
Often when you x an error at the be- the output reflects an inappropriate
analysis, check the list of variables output is completed, check to make

submitted. A common mistake is to sure that you have not left a FINISH
RECODE a variable for use in an analy- command in the middle of your job. Of-
sis but then use the original form of ten when previously used .sps jobs are
the variable rather than the recoded modied and resubmitted, we forget to
one when the variable is mentioned take out the old FINISH command.
on the command line. This mistake SPSS will ignore any commands after
will not generate an error or a warn- the FINISH command.
ing, but it will generate a lot of useless
output! The left side of Table 16.4 lists a few
Tip 8. Check for a premature FINISH common errors and warnings, and the
command. If part of your job is run but right side of the table suggests possible
is terminated before all the expected corrections.
Table 16.4 Common SPSS Syntax Errors and Warnings and How to Fix Them
SPSS Error or Warning Double Check This
Error: Unexpected end of le encountered Check data le (.DAT le) eliminate blank lines
(watch end and beginning of le particularly
Error: Variable name not recognized 1. If this happens on the rst and subsequent
variables mentioned, make certain system le
was accessed (check for correct le names).
2. Check the spelling of the variable name.
Error: Unrecognized commands due to A common error is a misspelled command such
incorrectly spelled words as STTSTICS for STATISTICS
Warning: The (ADD) VALUE LABELS command. Correct:
Be certain to include the values for each VALUE LABELS rv43 1 No 2 Yes.
category
Warning: Incorrectly enclosed literal Incorrect:
VALUE LABELS rv43 No 2 Yes.
Error: Text: Dependent. In regression analysis, Correct:
this warning occurs if you fail to include the VARIABLE LABEL v43 Place of Birth.
slash before the dependent variable is identied. Incorrect:
The machine will try to read DEPENDENT as a VARIABLE LABEL v43 Place of Birth.
variable if the / is omitted. If the / is missed, Correct:
other consequent errors may result, such as the REGRESSION VARIABLESv34 v12 V56
machines attempting to do a BACKWARD /DEPENDENTv34
solution but with no dependent variable /BACKWARD.
identied. Incorrect:
REGRESSION VARIABLESv34 v12 V56
DEPENDENTv34/
BACKWARD.
E X E R C I S E S
1. Using the information presented in Fig- 2. The following gure shows a portion of
ure 16.2, set up a properly formatted the data for a study. Reviewing the sec-
table to test the null hypothesis for the tion of the chapter dealing with error
relationship between signaling and detection methods may help you lo-
whether the person is a parent or cate some of the eight errors that can
spouse (grouped together) compared be found in scanning the data le. Two
with child or other (grouped together). of the errors are hard to see, so it is un-
Hand compute a test of signicance for derstandable if you only nd six. The
the relationship. What do you conclude computer could nd the others in a
about the null hypothesis? ash. Hint: check for non-numerical en-
tries. Good luck.
Display 16.3 Partial Printout for Data
00112435445523 211212333448976999 123232111123212 23411112
00211231231234 322213222122232212 432235333211232 3453O432
00314321234443 233445324433321212 332l22341122233 23323222
00413444688822 1222221122321233219 234433221222321 34521211
00513335543221 223332221114124211 332111422111212 321212
00513335543221 223332221114124211 332111422111212 32121223
00614322345432 231143321122122121 112211222111231 22132454
00812234342334 6545645646-0988777 793787987987423 98798777
00917879879877 878777778723442343 789987987777987 79879879
01018977772221 987798798778778987 798798777A97777 77788777
George, D., and Mallery, P. (2000). SPSS for Win- SPSS. (1999). SPSS Advanced Models 10.0. Chi-
dows Step by Step: A Simple Guide and Ref- cago: SPSS Inc. This advanced techniques
erence 9.0 Update (2nd ed). Boston: Allyn & manual will be of interest to those doing
Bacon. GLM repeated measures, loglinear, and
Kinnear, P., and Gray, C. (1999). SPSS for Win- other advanced techniques.
dows Made Simple (3rd ed). East Sussex, SPSS. (1999). SPSS 10.0 Syntax Reference Guide.
UK: Psychology Press. Chicago: SPSS Inc. For those who prefer to
SPSS. (1999). SPSS Base 10.0 Applications Guide. use syntax commands, this is the complete
Chicago: SPSS Inc. This basic manual con- reference guidesome 1412 pages are in-
tains the basic procedures covered in Doing cluded!
Nursing Research.
Chapter 17
Testing Simple Causal

Models
CHAPTER OUTLINE
A. Testing Three-Variable Causal Models 2. Testing for Sources of Spuriousness

1. Testing for Intervening Variables B. Testing Four-Variable Causal Models
KEY TERMS
Four-variable causal models Jacksons rule of thirds Source of spuriousness

Intervening variable model model
544
TESTING SIMPLE CAUSAL MODELS 545
I
n nonexperimental research (correla- will provide an answer. For example, it
tional and comparative surveys, case would be foolish to argue that the size of
studies, and so on), how do we test your present community inuences the
whether a proposed explanation for some size of the community in which you were
relationship has any merit? Suppose, for born. The causal sequencing is wrong;
example, that you want to see if your data the present cannot inuence the past. On
support the idea that expectations among the other hand, the size of community one
parents who value health explains the has chosen to live in as an adult may, of
observed link between health value and course, be inuenced by the size of com-
adolescents engagement in healthful life- munity in which one grew up. To demon-
styles. This chapter describes procedures strate that a relationship is not spurious is
for evaluating causal models such as the always a challenge, one that can never
one suggested above. First, the type of fully be met. (Recall that a spurious rela-
model being dealt with must be deter- tionship is one in which a third factor is in-
minedthat is, is it a source of spurious- uencing both the independent and the
ness, an intervening variable, or a candi- dependent variables; thus, their covaria-
date variable model? (See Chapter 3 for a tion may be the result of the common con-
description of these models.) Second, the nection to the source of spuriousness.) A
appropriate statistical procedures for the critic may always point to some potential
analysis must be determined. This deter- source of spuriousness for the relationship
mination is based on the level of measure- between the variables. The best the rst-
ment of the variables involved in each pro- time researcher can hope to achieve is to
posed analysis. deal with the more obvious potential
To establish a causal relationship, three sources of spuriousness.
conditions must be met: We begin this chapter with some tech-
niques for testing three-variable causal
1. The variables must be associated.
models. Later in the chapter, we examine
2. They must be in a plausible causal se-
causal models with four variables.
quence.
3. They must not be spuriously connected.
To show that two variables are associ- A. TESTING THREE-VARIABLE
ated, one has to demonstrate that they CAUSAL MODELS
vary together: To argue that one variable
is producing changes in another, one has In nonexperimental research, identical
to demonstrate that as one changes, so analyses may be used to test different
does the other. Empirical association is three-variable causal models. Such mod-
reected through cross-tabulation table els generally attempt to explain or elabo-
analysis (CROSSTABS), differences in rate on a relationship that is known to ex-
means across categories (MEANS, t TEST, ist or that a research project expects to
ANOVA, MANOVA), and various correla- demonstrate. These models introduce a
tional techniques (CORRELATION, RE- third variable to explain the relationship
GRESSION). between two other variables. We will use
To demonstrate a plausible causal se- some sample data to illustrate. Suppose
quence is largely a matter of theory or of we have done a survey on 395 senior high
common sense. What is meant here is not school students concerning factors that
only that the independent variable pre- inuence healthful lifestyle choices. Table
cedes the dependent variable in time but 17.1 presents the results.
that the ordering is believable. Usually, Table 17.1 is a standard cross-tabula-
giving a little thought to the causal order tion table whose computations could be
Table 17.1 Percent of Senior High School Students Who Make Healthful Lifestyle Choices by Health
Value Background*
Low Health Value High Health Value
Type of Background Background Total
Lifestyle Choices N % N % N %
Healthful choices 144 73.1 176 88.9 320 81.0
Unhealthful choices 53 26.9 22 11.1 75 19.0
Total 197 100.0 198 100.0 395 100.0
*X 2 16.021; df 1; signicant at the 0.001 level.
done with the CROSSTABS procedure. A 73.1 percent of those with low health value
shorter version of the table is also possi- backgrounds make similar choices. Note
ble and will be used to illustrate the model that there is a 15.8 percentage point differ-
testing to be presented later. The shorter ence by health value categories in those
version is illustrated in Table 17.2. making healthful lifestyle choices. We will
With the information provided, it is use this table in discussing the rst causal
possible to reconstruct the original table. model, the intervening variable model.
Because 73.1 percent of 197 students
with low health value backgrounds re- 1. Testing for Intervening
port healthful lifestyle choices, it is pos- Variables
sible to determine the number of stu-
dents who fall into the category (0.731 In an intervening variable model, the in-
197 144). Similar calculations could be terest is in understanding the relationship
done to reconstruct Table 17.1. between X and Yunderstanding the
The much simplied Table 17.2 is easy mechanism by which X is connected to Y.
to read and focuses attention on the two Frequently, a researcher will be testing a
percentage gures that are to be com- number of alternative explanations of
pared. Although some 88.9 percent of the how X inuences Y. In the case of one in-
students with high health value back- tervening variable, the relationship could
grounds report healthful lifestyle choices, be diagrammed as follows:
> > >

X I Y
Table 17.2 Percent of Senior High School Students Who Make Healthful Lifestyle Choices by Health
Value Background*
Low Health Value Background High Health Value Background
Healthful lifestyle choices, % 73.1 88.9
Cases, n 197 198
*X 2 16.021; df 1; signicant at the 0.001 level.

In this diagram, I is the intervening vari- ences Y. Typically, one would propose a
able, or the linking variable between X and number of possible intervening variables,
Y. The hypothesis is that variations in X so the following diagram would be more
cause variations in I, which, in turn, inu- appropriate:
In this diagram, three alternative ex- 1. Students with high health value back-
planations are suggested for the connec- grounds associate with peers who make
tion between X and Y. The researcher healthful lifestyle choices.
would collect data that measure each of 2. Students with high health value back-
the variables involved and conduct the grounds have parents who role model
appropriate statistical tests to determine, healthful lifestyle choices.
which, if any, of the proposed alternative 3. High health valuing parents are more
explanations, or intervening variables, ex- likely to put pressure on their children
plains the connection between X and Y. to make healthful lifestyle choices.
Let us suppose, for example, that three
alternative explanations are proposed for a. The Rationale Behind the Tests
the relationship between parental health
value and likelihood of making healthful The thinking behind the test is as follows:
lifestyle choices: if we have a causal relationship between X
and Y (health value and healthful lifestyle whether the original 15.8 percent differ-
choices) and propose a link to explain how ence in the relationship between health
X inuences Y, then X (health value) value and healthful lifestyle choices (1) in-
should not be able to inuence Y (lifestyle creases, (2) stays the same, (3) decreases
choices) if we hold the linking variable con- or disappears, or (4) is mixed.
stant. The argument is that X inuences Y
through I. A plumbing analogy may be
helpful. Water can only ow from X to Y b. Jacksons Rule of Thirds
through a pipe. If you turn off a valve lo-
cated between X and Y, then increasing the According to Jacksons rule of thirds, if
volume of water owing into the pipe at the original difference between the cate-
point X will have no inuence on Y because gories increases by one third or more, we
the valve has been turned off. However, if will interpret this as an increase, or a
we open the valve, then changes in the strengthening, of the original relationship;
pressure at X will inuence the ow at Y. if the difference remains within one third
Keeping the analogy in mind, let us now of the original, we will interpret this as an
see if we can control for the intervening indication that the relationship has re-
variable. mained the same; if the difference de-
Our rst explanation for the relation- creases by more than one third, we will in-
ship between parental health value and terpret this as a decrease or disappearance
likelihood of making healthful lifestyle of the relationship; nally, if the relation-
choices is that students with parents who ship is markedly different when different
have high health values tend to associate control categories are compared with one
more with peers who make healthful another (e.g., it disappears in one cate-
lifestyle choices. How can we analyze the gory, but stays the same in the other), the
data to see if the results are consistent result is mixed (Jackson, 1999).
with this model? Using a cross-tabulation To apply the rule of thirds to the case
table analysis (CROSSTABS), we run the re- under examination, we must rst decide
lationship between health value and where the cutpoints are between the
lifestyle choices, controlling for the inter- thirds. To do this, we take the original dif-
vening variable, association with peers ference of 15.8 and divide by 3; this yields
who make healthful lifestyle choices. We a value of 5.3. Table 17.3 presents these
wish to see what happens when the con- values and shows how the differences
trol is applied. In this case, we wish to see would be interpreted.
Table 17.3 Applying Jacksons Rule of Thirds

Sample Data Results Interpretation
Original difference: 88.9 73.1 15.8
Determining thirds: 15.8 3 5.3
Outcomes
1. If new difference is greater than 21.1 (15.8 5.3 21.1) Increased
2. If new difference is between 10.5 and 21.1 (15.8 5.3) Stayed the same
3. If new difference is less than 10.5 Decreased or disappeared
4. If new differences vary markedly across categories of the Mixed
control variable
c. Using CROSSTABS to Test for an To interpret the outcomes, it is neces-

Intervening Variable sary to determine whether the original re-
lationship has increased, stayed the same,
But what are our expectations? If the decreased or disappeared, or is mixed. The
model being tested is correct, we would beginning researcher should keep the in-
expect the relationship between health terpretation of the data as simple as pos-
value and healthful lifestyle choices to de- sible. Only when the difference decreases
crease or disappear when the relationship or disappears do we have possible sup-
is run, controlling for the intervening vari- port for an intervening variable model.
able. If health value inuences healthful Let us look at the ve outcomes and sug-
lifestyle choices through the linking vari- gest an interpretation for each one.
able (peers lifestyle choices), then if we
hold the linking variable constant, there Outcome 1. According to the rule of
should be no relationship between health thirds, the relationship has decreased
value and lifestyle choices. Differences in or disappeared (i.e., the new difference
healthful lifestyle choices by health value is less than 10.5, so it has decreased by
category should decrease or disappear more than one third of the original).
when the control is applied. All other out- The original difference has been re-
comes are interpreted as not supportive duced to one percentage point in the
of the model. Table 17.4 presents sum- case of those whose best friends en-
mary data for ve possible outcomes. gage in healthful lifestyle choices and to
Table 17.4 Percent of Senior High School Students Who Make Healthful Lifestyle Choices by
Health Value Background, Controlling for Healthful Lifestyle Choices of Best Friend, with Five
Possible Outcomes
Best Friend with Best Friend with
Healthful Choices Unhealthful Choices*
Low Health High Health Low Health High Health
Outcomes Value Background Value Background Value Background Value Background
First outcome 92.0 93.0 71.0 69.0
Difference 1.0 2.0
Second outcome 74.0 92.0 71.0 86.0
Difference 18.0 15.0
Third outcome 85.0 92.0 68.0 76.0
Difference 7.0 8.0
Fourth outcome 74.0 96.0 61.0 82.0
Fifth outcome 90.0 92.0 60.0 82.0
Difference 2.0 22.0
*The original difference, with no control for healthful lifestyle choices of best friend, is shown below:
Low Health High Health

Value Background Value Background
Percent who make 73.1 88.9

healthful lifestyle choices
Difference 15.8
two percentage points for those whose make healthful lifestyle choices because
best friends do not engage in healthful their parents role model healthful behav-
lifestyle choices. This is the only out- iors. We will assume for this example that
comeone in which the original rela- healthful lifestyle choices are measured in
tionship decreases or disappearsthat terms of the number of healthful behaviors
we consider to be consistent with the selected on the adolescent lifestyle ques-
intervening variable causal model. tionnaire (ratio level measurement, Gillis,
Outcome 2. The relationship stays the 1997), permitting the use of the MEANS
same (the new difference is between procedure.
10.5 and 21.1), so the intervening vari- The logic is identical to that used in the
able model is to be rejected. previous procedure. We will examine the
Outcome 3. This outcome supports the difference in the number of healthful
intervening variable model. The differ- choices between health value categories.
ence decreases or disappears ( i.e., the We will then re-run that relationship, con-
new difference is less than 10.5, so it trolling for parents role modeling of
has decreased by more than one third healthful lifestyle choices. Once again, we
of the original); the interpretation is will apply the rule of thirds to provide a
that the independent variable inu- guideline for the interpretation of the
ences the dependent variable through data. Table 17.5 indicates that there is a
the tested intervening variable. 14.0 difference in the number of healthful
Outcome 4. This outcome would lead us lifestyle behaviors selected by health
to reject the intervening variable value categories. The question is will this
model; the relationship is strengthened difference increase, stay the same, de-
( i.e., the new difference is greater than crease, disappear, or be mixed when the
21.1), so this suggests that the pro- control for parental role modeling of
posed alternative explanation is hav- healthful life style choices is applied?
ing an independent inuence on the Table 17.5 shows different outcomes.
dependent variable. Once again, we look at the control table,
Outcome 5. This outcome would also examine the difference between the lev-
lead us to reject the intervening vari- els of parental role modeling of healthful
able model. The difference decreases lifestyle choices, and contrast this differ-
or disappears in one of the control cat- ence with the original difference of 14.0
egories but increases in the other, sug- healthful behaviors.
gesting a conditional effectthe inter-
vening variable is probably having an Outcome 1. For students whose parents
independent inuence but only at cer- do not role model healthful lifestyle
tain levels of the intervening variable. choices, the data indicate a 19.1 health-
This is an example of a mixed result. ful behavior difference between health
value categories in total number of
healthful lifestyle choices. Among those
the CROSSTABS procedure in Appendix A.
who perceive their parents to role
model healthful lifestyle choices, the
d. Using MEANS to test for an difference between the categories is
Intervening Variable 19.6 healthful behaviors. By applying
the rule of thirds, we see that the differ-
The second explanation proposed for the ence in both health value categories has
connection between health value and increased by more than one third (i.e.,
healthful lifestyle choices is that the stu- the new difference is greater than 18.66
dents with high health value backgrounds [14.0 3 4.66]; 14.04.66 18.66).
Table 17.5 Mean Number of Healthful Lifestyle Choices by Health Value Background, Controlling for
Parents Role Modeling of Healthful Lifestyle Choices, with Five Possible Outcomes
No Parental Role Modeling Parental Role Modeling*
Low Health High Health Low Health High Health
Outcomes Value Background Value Background Value Background Value Background
First outcome 14.9 34.0 33.0 52.6
Second outcome 22.3 36.6 25.5 39.8
Third outcome 15.6 22.7 34.2 41.0
Difference 7.1 6.8
Fourth outcome 23.5 24.8 37.7 39.2
Difference 1.3 1.5
Fifth outcome 16.4 31.3 38.9 40.1
Difference 14.9 1.2
*The original difference, with no control for parental role modeling, is shown below:
Healthful Lifestyle Choices Low Health Value Background High Health Value Background
Mean number of healthful lifestyle choices 24.7 38.7
Difference 14.0
Therefore, we argue that the relation- supports the intervening variable

ship has been intensied; we must re- model.
ject the intervening variable model. Outcome 5. The nal outcome suggests a
Outcome 2. In the second outcome, in mixed result. The difference decreases
both the No Healthful Role Modeling or disappears within the Healthful
and Healthful Role Modeling cate- Role Modeling category but remains
gories, the difference between the stu- the same within the No Healthful Role
dents with low and high health value Modeling category. We reject the in-
backgrounds remains almost the same tervening variable model. The data
as the original difference of 14.0 ( i.e., here suggest that parental role model-
the new difference is between 9.34 and ing of healthful choices has a condi-
18.66; 14.0 4.66). Therefore, we re- tional impact on the number of health-
ject the intervening variable model. ful lifestyle choices made by students.
Outcome 3. In the third outcome, the dif-
ference has been reduced by more
the MEANS procedure in Appendix A.
than one third (the new difference is
less than 9.34); therefore, we conclude
that there is some evidence to support e. Using PARTIAL CORR to Test for
the intervening variable model. an Intervening Variable
Outcome 4. In both the No Healthful Role
Modeling and Healthful Role Model- Partial correlations are measures of the
ing categories, the differences in num- strength of an association that take into
ber of healthful lifestyle choices have account one or more additional variables.
decreased or disappeared (dropped by (Refer to Chapter 11 for a more detailed
more than two thirds); this outcome discussion.) Partial correlations measure
how closely two variables are associated the correlations between three variables:
when the inuence of other variables is ad- parental pressure index (I ), health value
justed for; a rst-order partial is one that score ( X ), and number of healthful life-
takes into account one additional variable style choices (Y ).
and a second-order partial takes into ac- The rst test is to see if the magnitude
count two additional variables. Partial cor- of the correlations is consistent with the
relations are presented as r12.3. This indi- model being tested. The prediction was
cates that you are measuring the that adjacent correlations would be
correlations between variables 1 and 2 higher than nonadjacent ones. The adja-
with the effect of variable 3 removed from cent correlations are I-X and I-Y and the
both the variables being correlated. By correlations are 0.31 and 0.46, respec-
combining CORRELATIONS and PARTIAL tively. The nonadjacent correlation is X-Y
CORR, we can test for an intervening vari- and the correlation is 0.22. So far, the
able. data are consistent with the intervening
In this case, we wish to test whether variable model.
parental inuence intervenes between The next question is whether the par-
health value and healthful lifestyle choices tial correlation will increase, will stay the
by adolescents. Using correlational tech- same, or will decrease/disappear when I
niques, we would rst establish that there is controlled. For this analysis, a partial
is an association between health value and correlation would be computed. When
adolescents healthful lifestyle choices. If this calculation is done (see Chapter 11
there is an association, we would then pro- for the formula), we nd that the partial
ceed with the analysis to test whether the (rXY.I ) is 0.10. By applying the rule of
data are consistent with an intervening thirds, we see that the original relation-
variable model. ship between X and Y is 0.22 and because
If the intervening variable model is cor- the partial is about one half of the original
rect, we should at least expect the fol- value, we note that the association has
lowing: decreased or disappeared. Therefore, we
nd support for the model.
That the correlation between adjacent
To perform this analysis using SPSS,
variables will be greater than between
see the PARTIAL CORR procedure in Ap-
non-adjacent categories (rXY rXI or rIY )
pendix A.
That if I is controlled, the relationship
between X and Y should be reduced or
disappear; this could be tested with a
partial correlation coefcient. 2. Testing for Sources of
Spuriousness
The model will be tested using two cor-
relational techniques, CORRELATIONS The next major type of causal model is the
and PARTIAL CORR. Table 17.6 presents source of spuriousness model. Here the
Table 17.6 Correlations Between Variables in Model

Number of Healthful
Health Value Lifestyle Choices
Parental Pressure (I) (X ) (Y )
Parental pressure ( I ) 1.00
Health value (X ) 0.31 1.00
Healthful lifestyle choices (Y ) 0.46 0.22 1.00
researcher proposes that although there is tor is inuencing both of them. Having ob-
a statistically signicant relationship be- served a statistically signicant relation-
tween the variables X and Y, this relationship, the researcher will want to ensure
ship may be a noncausal one, only existing that the relationship is not spurious and,
because some third variable is inuencing therefore, will run a number of spurious-
both X and Y. The argument is that the only ness checks. The source of spuriousness
reason X is related to Y is that a third fac- model may be diagrammed as follows:
X Y
S/S
a. The Rationale Behind the Tests come mixed. In order to conclude that
the original relationship is spurious,
How do we go about testing a source of the difference between the categories
spuriousness model? The idea is that if X must decrease/disappear.
and Y are spuriously associated, the rea-
son they vary together is that a third vari-
able (a source of spuriousness) is inu- b. The Dilemma: The Models Are Not
encing both X and Y. Therefore, if we Empirically Distinguishable
control for the source of spuriousness
(S/S), there should no longer be any asso- The difculty is that it is possible for two
ciation between X and Y. This suggests researchers working with the same data
that we need the same kind of analysis and with the same three variables to es-
that we used for testing intervening vari- tablish two different causal models, one
ables. To test for a potential source of spu- proposing that the variables are con-
riousness, the steps are: nected in an intervening variable model
and the other proposing that the variables
Test the original relationship between may be associated spuriously. They might
X and Y; if this demonstrates a robust then do the identical analysis but come to
relationship (probably statistically sig- totally different conclusions. Let us sup-
nicant); then, pose that when the control is applied, the
Controlling for the source of spurious- original difference decreases/disappears.
ness, re-run the relationship between In this case, one researcher would con-
X and Y. As in the intervening variable clude that the intervening variable model
model, we can then apply Jacksons has found support in the data and the
rule of thirds to determine if the rela- other would conclude that the relation-
tionship has increased, stayed the ship is spurious. Both would be correct.
same, decreased/disappeared, or be- Because the two models are not empir-
ically distinguishable, the importance of health value level that is inuencing the
precisely specifying models in advance students lifestyle choices.
becomes clear. If we develop the models Table 17.8 reports ve different out-
after analyzing the data, the interpreta- comes for this analysis. These should be
tion of the data is little more than a ight examined carefully, the rule of thirds
of fancy; if the difference disappeared, we should be applied, and a decision should
could argue either for an intervening vari- be made as to which outcome lends sup-
able explanation or for a source of spuri- port to the spuriousness model of the re-
ousness explanation. Because the test relationship between the variables. To apply
sults equally support both models, we this rule, simply divide the original differ-
must specify in advance which model we ence by 3 (14.0 3 4.66). Next, compare
are, if fact, testing. the difference in healthful lifestyle choices
by health value categories for each of the
rural and urban categories.
c. Using MEANS to Test for
Spuriousness Outcome 1. Here the differences (21.1
and 19.3) have both grown by more
To illustrate a test for spuriousness, we than a third; therefore, we reject the
will use the same variables as in the pre- source of spuriousness model.
vious example. Table 17.7 presents sam- Outcome 2. The original difference be-
ple data showing how the analysis turned tween the number of healthful lifestyle
out for the relationship between health choices by students with low versus
value and healthful lifestyle choices. high health value backgrounds was
Readers should note that there is a 14.0 choices. When the control for
14.0 difference between the students with rural versus urban background is ap-
low and high health value backgrounds in plied, the difference remains much the
the number of healthful lifestyle choices same; therefore, we reject rural versus
made. Now suppose that we wanted to urban background as a source of spu-
make certain that this relationship was riousness.
not caused spuriously by the rural or ur- Outcome 3. The original difference in this
ban backgrounds of these students. The result has been reduced to 6.1 for both
type of home community (rural versus categories. Because the difference has
urban) may be inuencing the health decreased/disappeared, we nd sup-
value level achieved by the families and port for the source of spuriousness
may also be inuencing the lifestyle model.
choices of the students. It would, there- Outcome 4. In this case, the original dif-
fore, be the urban or rural location that is ference has been reduced to less than
inuencing both variables rather than the one third of its original value, so we can-
Table 17.7 Number of Healthful Lifestyle Choices by Senior High School Students, by Health Value
Background
Healthful Lifestyle Choices Low Health Value Background High Health Value Background
Mean number of healthful
lifestyle choices 24.7 38.7
Difference 14.0
Cases, n 197 198
Table 17.8 Mean Number of Healthful Lifestyle Choices by Health Value Background, Controlling for
Rural and Urban Backgrounds, with Five Possible Outcomes
Rural Background Urban Background
Low Health High Health Low Health Health Value
Outcomes Value Background Value Background Value Background High Background
First outcome 8.7 29.8 33.3 52.6
Second outcome 22.7 36.7 25.8 39.6
Third outcome 23.0 29.1 37.1 43.2
Difference 6.1 6.1
Fourth outcome 17.3 19.1 37.4 39.7
Difference 1.8 2.3
Fifth outcome 21.6 22.7 27.4 40.4
Difference 1.1 13.0
not reject the source of spuriousness To perform this analysis using SPSS, see
model. Outcomes 3 and 4 both lend the CROSSTABS procedure in Appendix A.
support to the spuriousness model.
Outcome 5. Here the result is mixed. The
difference disappears among the rural e. Using PARTIAL CORR to Test for
students but is only slightly reduced Spuriousness
among the urban students. We reject
the source of spuriousness model. This test requires ratio level variables. It
simply involves running the zero-order
In order to do the analysis, we need a correlation followed by a partial correla-
table showing the difference in healthful tion controlling for the potential source
lifestyle choices by health value cate- of spuriousness. The analysis should be
gories. Second, we will need to re-run this run and the data interpreted according to
relationship, controlling for rural and ur- the rule of thirds. We only consider the
ban backgrounds. original relationship to be spurious if the
To perform this analysis using SPSS, see relationship decreases/disappears.
the MEANS procedure in Appendix A. When procedures are being selected
for evaluating causal models and the
d. Using CROSSTABS to Test for a researcher believes that the vast majority
Source of Spuriousness will only require one type of procedure, it
is sometimes best, in the interest of simpli-
To use CROSSTABS in a test for a source city, to use the same procedure through-
of spuriousness, one runs the original re- out rather than shift back and forth be-
lationship and then re-runs the relation- tween techniques in a way that may
ship, controlling for the source of spuri- confuse readers.This usually means some
ousness. The rule of thirds would then be underutilization of the data because the
applied and we only consider the original procedure selected must meet the mea-
relationship to be spurious if the differ- surement requirements of the variable
ence decreases/disappears. with the lowest level of measurement. For
example, if you have only one test that

could be done using the MEANS proce- A B
dure but all the rest require CROSSTABS,
then you would probably use CROSSTABS
throughout even though you could use the
MEANS procedure on one of the tests. D C
Your decision would be based on keeping
the analysis as simple as possible for your
readers.
To perform this analysis using SPSS,
see the PARTIAL CORR procedure in Ap-
pendix A.
A B
B. TESTING FOUR-VARIABLE
CAUSAL MODELS
D C
In the classic work entitled Causal Infer-
Figure 17.1 Four-variable causal model.
ences in Nonexperimental Research (1964),
Hubert M. Blalock, Jr., presented a number
of ideas on testing causal models. The pro-
cedures for testing four-variable causal furthest apart causally. If the model is
models parallel the ideas presented in the correct, the following should hold:
earlier section of the chapter and should
be seen as an extension of them. In the rAD rAB, rAC
case of four-variable causal models, there
rAC rAB
may be two intervening variables (plus an
independent and dependent variable). Fig- rBD rAB, rBC
ure 17.1 illustrates a four-variable model.
Suppose the model that you wish to ex- The strength of relationships should
amine is the one presented in Figure 17.1. diminish if intervening variables are
This model suggests a specic causal or- controlled for; if the model is correct,
dering of the variables. Or A B C D. the following partial correlation re-
What could one do to test this model? A sults should hold:
number of properties that should hold true
if the model is accurate are: rAC.B 0
Correlations between adjacent vari- rBD.C 0

ables should be higher than between
rAD.BC 0
nonadjacent variables; if the model is
correct, the following should hold:
Lets look at the steps necessary to
test the simplest model. First, we will
rAB rAC , rAD
examine the magnitude of the various
rBC rBD correlations. Are they consistent with
the four-variable model? Namely:
rCD rAC
Are the correlations between causally
The weakest correlation in a causal adjacent variables higher than be-
chain should be between the variables tween nonadjacent variables?
Observed
Do correlations drop when intervening Partial
variables are controlled? Model Predicts Correlation Conclusion
Suppose that we have the following 1. Testing AC link:
correlation matrix between the variables rAC.B 0* 0.10 Reduced by
more than one
A, B, C, and D.
third: Supports
Correlation Matrix model
Variables A B C 2. Testing BD link:
rBD.C 0 0.33 Reduced by
B 0.56
more than one
C 0.21 0.61 third: Supports
model
D 0.10 0.18 0.63
3. Testing AD link:
rAD.BC 0 0.01 Reduced by
Step 1. Does the causal order make more than one
sense? The rst step in assessing a third: Supports
model
causal model is to ensure that the
causal ordering is plausible. The model *The formulas used to hand compute the partials are as follows:
should pass a commonsensical test, in- rAC (rAC)(rBC)

dicating that the variables are in a tem- First order: rAC.B 1r
AC 1 rBC
2 2
poral sequence that is possible. For ex-

rAD.B (rAC.B)(rCD.B)
ample, the rst variable in the model Second-order: rAD.BC
1r 2 1r 2
should occur before other variables and AC.B CD.B
should represent a possible cause of

The causal inference uses the rule of thirds used in this text.
variation in variables that occur later in Hubert Blalock bears no responsibility on this point!
the model.
Step 2. Compare the correlations. Inspect
the correlations to see if correlations The evidence in this case is consis-
between adjacent variables are higher tent with the proposed model. If any of
than those between nonadjacent vari- the results had failed the test, then the
ables. In this case, we note the following: interpretation would be that the evi-
Observed dence is not consistent with the pro-
Model Predicts Correlations Conclusions posed model.
rAB rAC , rAD 0.56 0.21 0.10 Supports model Although it is possible to test complex
rBC rBD 0.61 0.18 Supports model causal models using cross-tabulation ta-
bles, the procedure requires a large num-
rCD rAC 0.63 0.21 Supports model
ber of cases when additional control vari-
ables are added. Once again, however,
Step 3. Eliminate causal links not im- the principle remains the same: observe
plied by the model. There are three what happens to the original relationship
causal links that are not implied by the when the control variables are applied si-
model. These include the AC link, the multaneously. Does the original differ-
BD link, and the AD link. (These are ence increase, stay the same, decrease/
marked with thinner lines in Figure disappear, or is it mixed?
17.1.) The rst two are tested using To perform this analysis using SPSS, see
rst-order (one control) partial corre- the CORRELATION and PARTIAL CORR
lations, and the last one is tested by procedures in Appendix A.
means of a second-order partial corre- Chapter 18 introduces some other ba-
lation. The tests reveal the following: sic techniques that are used when at-
tempts are made to deal with multiple cludes regression, discriminant, analysis
variables simultaneously. The chapter in- of variance, and factor analyses.
E X E R C I S E S
1. Working with a two-variable relation- 2. Compare the procedures used by a re-
ship that you think would be relatively searcher using an experimental design
strong, diagram four models (two in- to arrive at causal inferences with those
tervening variables and two possible used by a nonexperimental researcher.
sources of spuriousness) that may ex- Using the same three variables, illus-
plain the original two-variable rela- trate an experimental and a nonexperi-
tionship. Describe in detail how you mental approach by outlining a possible
would go about testing the four pro- design each might use in exploring the
posed models. What outcomes would relationship between the same three
lend support to each of your models, variables. For the particular relation-
and what kinds of evidence would lead ship you have explored, discuss the rea-
you to reject them? sons why you prefer one of the designs.
Blalock, H.M., Jr. (1964). Causal Inferences in the classic statement on interpreting three-
Nonexperimental Research. Durham: The variable causal models. Because most un-
University of North Carolina Press. Under- dergraduates nd the terminology difcult
graduates will nd this book challenging but to sort out, the authors of this text do not
nonetheless rewarding to work through the follow the Lazarsfeld approach.
methods for testing models with four or Munroe, B., and Page, E. (1993). Statistical
more variables. Methods for Health Care Research (2nd ed).
Brink, P., and Wood, M. (1998). Advanced De- Philadelphia: J.B. Lippincott. Several help-
sign in Survey Research. Thousand Oaks, ful chapters present theory testing ap-
CA: Sage. This book contains a compre- proaches using path models and structural
hensive chapter on structural equation equation models.
modeling, a means of testing hypothesized Musial, C.M., Jones, S.L., and Warner, C.D.
relationships between concepts. (1998). Structural equation modeling and
Hyman, H. (1955). Survey Design and Analysis. its relationship to multiple regression and
New York: The Free Press. Paul Lazars- factor analysis. Research in Nursing and
felds foreword to Hymans book contains Health, 21, 271281.
Chapter 18
Four Multivariate
Techniques
CHAPTER OUTLINE
A. Multiple Regression C. Multivariate Analysis of Variance

1. The Rationale 1. The Rationale
2. The Linear Regression Equation 2. Comparison with Multiple Regression
3. The Strength of the Association, R2 Analysis
4. Using Variables Not Meeting the 3. Presenting and Interpreting Results
Measurement Assumption D. Factor Analysis
5. Some Tips for Using Regression Analysis
1. The Rationale
6. Presenting and Interpreting Regression
2. Basic Terminology of Factor Analysis
Results
3. Presenting and Interpreting Results
B. Discriminant Function Analysis
4. Comparison with Cronbachs Alpha
1. The Rationale Procedure
2. Comparison with Multiple Regression
5. Comparison with Multiple Regression
Analysis
Analysis
3. Presenting and Interpreting Results
E. The Four Techniques Compared
KEY TERMS
b Coefficient Eigenvalue Multicollinearity

BACKWARD (stepwise) Exploratory factor analysis Multivariate analysis of
solution Factorable variance (ANOVA or
Beta weight MANOVA)
Factor analysis
Confirmatory factor analysis Multiple regression analysis
Factor extraction
Correlation matrix Principal components
Factor loadings
method
Covariates Hierarchical regression
Principal factors method
Data reductioin Instrument development
R2
Discriminant function analysis Instrument validation
Stepwise entry solution
Dummy coding Linear regression equation
Theory development
Dummy variable Marker variable
559
T
his chapter extends some of the proaches featured in this chapter: dis-
techniques for analysis introduced criminant analysis, multivariate analysis
in the previous chapters. The four of variance, and factor analysis. An exami-
sections of this chapter cover regression nation of these basic multivariate tech-
analysis (an extension of the correlation niques reveals that they have much in
discussion in Chapter 11); discriminant common.
function analysis, which is a multivariate
extension used in the analysis of nominal
dependent variables (see Chapter 11);
A. MULTIPLE REGRESSION
multivariate analysis of variance (an ex-
tension of analysis of variance in Chapter
Multiple regression analysis attempts to
12); and factor analysis, which is used pri-
predict variations in a dependent vari-
marily in index construction (an exten-
able from two or more independent vari-
sion of the discussion in Chapter 13). In
ables. And just as we did with correlation,
all likelihood, you have already encoun-
there are two basic elementsa measure
tered these four techniques in reading
of the strength of the association and an
the nursing literature and, therefore, you
equation that describes the relationship.
need to have an understanding of each of
Multiple regression involves two or more
them. Fundamentally, they all rely on the
independent variables in contrast to the
basic ideas surrounding the concept of
simple linear regression that involved
correlation.
one independent variable.
As discussed throughout this text, phe-
nomena nursing researchers wish to un-
derstand are complex and multivariate in 1. The Rationale
nature. This means that the analytical
techniques we use must allow us to take Regression is a powerful tool in the hands
into account many variables simultane- of nurse researchers because of its ability
ously. With the exception of factor analy- to provide:
sis, these techniques provide for the analy-
sis of multiple independent variables.
An estimate of the relative importance
of independent variables in inuencing
Factor analysis allows one to determine
a dependent variable
the similarities and differences among sets
of separate measures; therefore, it is an im-
A mathematical equation that describes
the relationship between the indepen-
portant technique used in the develop-
dent variables and the dependent vari-
ment of indexes. Although bivariate (two-
able
variable) analyses have their place, nurse
researchers ultimately need to use basic
A measure of how much variance is ex-
plained by the combination of inde-
multivariate techniques if they are to
pendent variables
tackle the complexity inherent in their sub-
ject matter. For example, if you are trying to esti-
The selection of the appropriate tech- mate the relative contribution of three
nique should be guided by a consideration variables when predicting cigarette smok-
of (1) the level of measurement attained in ing behavior, a regression analysis might
measuring the variables and (2) what the indicate that exposure to cigarette adver-
analysis is attempting to reveal. We will tising accounts for 22 percent of the vari-
begin our consideration of multivariate ation, that friends smoking status ac-
analysis by exploring multiple regression counts for 37 percent of the variation, and
analysis. An understanding of regression that annual income contributes another
is helpful when examining the other ap- 14 percent of the variation in smoking
FOUR MULTIVARIATE TECHNIQUES 561
behavior. Taken together, the three vari- (see Box 13.1 for an example of the impact
ables account for 73 percent of the varia- of reduced levels of measurement).
tion in cigarette smoking.
Alternatively, the relationship between 2. The Linear Regression Equation
smoking behavior and exposure to ciga-
rette advertising could be analyzed simply In Chapter 11, we examined the relation-
by comparing the average number of ciga- ship between two variables using a corre-
rettes smoked across categories of adver- lation approach. The equation describing
tisement exposure (MEANS analysis). This the relationship between X (independent
would permit the researcher to show a sig- variable) and Y (dependent variable) con-
nicant relationship between cigarette sisted of an a value and a b coefcient as in:
smoking and advertisement exposure but
would not enable the researcher to say Y a bX
that exposure to cigarette advertising is
about one half as important as friends To take into account two or more in-
smoking status. Because regression analy- dependent variables, this basic equation
sis provides this additional information, it is extended as follows:
is a powerful tool. Besides allowing the re-
searcher to examine the relative impor- Y a b1X1 b2 X2 . . . bk Xk
tance of the various factors producing
variation in a dependent variable, it also In the equation, the a is a constant and,
allows the researcher to express the rela- if the values for each case were calculated
tionship in the form of an equation, giving and plotted, would represent the point
the researcher the ability to predict values where the regression line crosses the Y
for a dependent variable, given values for axis. The b coefcients refer to the slopes
the independent variables. of the regression lines. If small increases in
As in correlational analysis (see Chap- the X variable lead to large increases in the
ter 11), we are interested in both the equa- Y variable, the b value will be higher (see
tion that describes the relationship as well the relationship between cigarette smok-
as a measure of the strength of the associ- ing and number of cigarette advertise-
ation. We will only consider the simplest ments shown in Fig. 18.1); on the other
version of regressionthat of an additive hand, if it takes large increases in X to pro-
or linear relationship among the variables. duce an increase in Y, the b value will be
Regression analysis assumes that the vari- smaller (see the relationship between cig-
ables are normally distributed and that arette smoking and annual income shown
measurement is at the ratio level. Special in Fig. 18.2). In the two-variable case (Figs.
procedures, however, do permit the inclu- 18.1 and 18.2), the b value in the case of
sion of variables not achieving ratio mea- number of cigarette advertisements is
surement. higher than the b value for the case of in-
Although regression analysis assumes come predicting cigarette smoking behav-
ratio level measurement and normally dis- ior. The reason is that the values for num-
tributed variables, researchers sometimes ber of cigarette advertisements have a
include ordinal and even nominal vari- smaller range, perhaps from 0 to 50 per
ables. However, the price one pays for re- year approximately, but the range in the
duced levels of measurement is almost values for income is considerable, perhaps
certainly a weakened ability to predict from $18,000 to $80,000. Thus, the b values
variation in the dependent variable, as well tend to be much lower. Even though the
as a greater instability in the coefcients correlation between cigarette smoking
associated with the independent variables and number of cigarette advertisements
Figure 18.1 Prestige and years of education. (continued on next page)
and between cigarette smoking and in- advertisements). Each independent vari-
come may be similar (e.g., 0.90 in both able is standardized (reassigned values so
cases), the b value for income will be much that each has a mean of 0 and a standard
lower than the b value for number of ciga- deviation of 1), which then allows us to
rette advertisements. compare the beta values directly. The s
If standardized slopes are of interest, represent the amount of change in Y (the
these will be referred to as weights or dependent variable) that can be associ-
beta weights. In this case, think of all the ated with a given change in one of the Xs,
variables in the equation as being stan- when the inuences of the other indepen-
dardizedthink of them as Z scoresso dent variables are held constant. Regres-
that it does not matter if the independent sion programs provide the researcher
variables have different ranges (as in the with both the b and the (beta) coef-
case of income and number of cigarette cients (see Fig. 18.2).
Figure 18.1 Continued
The strategy of multiple regression in- 3. The Strength of the

volves determining the slopes for each of Association, R2
the independent variables while simulta-
neously holding constant, or adjusting The R 2 is a measure of the amount of vari-
for, the other independent variables. The ation in the dependent variable that is ex-
slopes (b coefcients) are determined to plained by the combination of indepen-
maximize our ability to predict variations dent variables. Recall that when two
in the dependent variable. Thus, we may variables are involved, the measure is r2
dene a linear regression equation as (see Chapter 11). The two statistics are di-
one that describes a relationship between rectly comparable and would yield identi-
a number of independent variables and a cal results if, in a case in which there are
dependent variable and that provides for multiple independent variables, one sim-
the best linear (additive) weightings of ply took the values for each variable,
the independent variables and a constant plugged them into the equation, and then
calculated so as to maximize the predic- computed the predicted value for the
tion of the dependent variable. dependent variable. If one then correlated
Figure 18.2 Prestige and annual income. (continued on next page)
the predicted and the observed values of Pure researchers, on the other hand, usu-
the dependent variable, the problem is really focus on the s because, generally,
duced to a simple two-variable correla- theorists are more concerned with the
tion, and the r2 would equal R2. Both of relative impact of each independent vari-
these statistics vary from 0 to 1. The able on the dependent variable.
higher the value, the higher the explained
variance; the higher the value, the higher
the predictability of the dependent vari- 4. Using Variables Not Meeting
able by the independent variables. the Measurement Assumption
Regression analysis is an important
Researchers frequently wish to include,
technique for both applied and pure re-
along with ratio level variables, variables
search. Applied researchers are particu-
measured at either the ordinal or nominal
larly interested in identifying the inde-
levels. It is possible to do this, but one
pendent variables that are:
must exercise greater caution in inter-
Important in inuencing the depen- preting the results.
dent variable
Modiable through policy changes
a. Using Ordinal Variables
Applied researchers are most con-
cerned with the b coefcients because The price one pays in using ordinal vari-
they indicate how much change in the in- ables is generally a weakened ability to pre-
dependent variable will be required for a dict variations in a dependent variable. In
unit change in the dependent variable. general, the fewer the categories in an ordi-
Figure 18.2 Continued
nal variable, the lower its correlation with characteristic. Again, these variables are
other variables. Thus, more categories in at a disadvantage in explaining variation in
the ordinal variable are preferable. If re- the dependent variable. Indeed, with only
gression analysis is anticipated, using nine- two categories, these variables are at a
point Likert items rather than ve- or seven- particular disadvantage (e.g., when com-
point categories is recommended. When pared with another ordinal variable with
ordinal variables are placed into competi- nine values or a ratio variable). Again, the
tion with ratio level variables for explaining coefcients are likely to underestimate
variance, the resulting equation will tend to the importance of such variables when
underestimate the relative importance of they are in competition with variables
the ordinal variables (see Box 13.1 for more measured at the ordinal or ratio levels.
details on the effects of using fewer cate-
gories).
c. Using Dummy Variables for
Nominal Variables
b. Using PresenceAbsence Questions
Nominal level variables may be included
In the case of presenceabsence ques- in a regression analysis, but they must
tions, one really has an ordinal variable rst have special coding methods applied
with two values. Such questions are nor- to them so that they can be properly in-
mally coded so that 0 refers to the ab- terpreted in the regression analysis. Sup-
sence and 1 refers to the presence of the pose you have a religious afliation vari-
Table 18.1 Recoding the Religion Variable into Three Dummy Variables
Coding for Three New Dummy Variables
Case # Original Coding, Religion Variable Catholic Protestant Jewish
1 1 (Catholic) 1 0 0
2 2 (Protestant) 0 1 0
3 3 (Jewish) 0 0 1
4 3 (Jewish) 0 0 1
6 1 (Catholic) 1 0 0
7 4 (atheist) 0 0 0
8 9 (no answer) 0 0 0
9 4 (Mormon) 0 0 0
178 4 (agnostic) 0 0 0
able (nominal) and wish to include it in a be accounted for by the combined values
regression analysis. For purposes of illus- in the absence category. It should be
tration, suppose religion is arbitrarily noted that you always need an other in
coded into four categories: Protestant this type of situation. In other words,
1; Catholic 2; Jewish 3; and other even if your variable has 20 categories in
4. It would not be advisable to enter this it, you would still need a nal category
variable into the regression equation in called other. Note that if you use gender
this manner because it would be treated as a dummy variable, either the males
as if the numbers really meant that 2 was or the females will be assigned to the
twice as big as 1, 4 was twice as big as 2, other category.
and so on. Rather, the researcher should The dummy-variable coding procedure
apply a special coding method referred uses 1s and 0s. Thus, we will have a
to as dummy coding. In dummy coding, Protestant variable, and it will be coded
new variables are created out of the orig- into presence (1) or absence (0); a Cath-
inal variable in such a way that there will olic variable using 1s for Catholics and 0s
be one fewer new variables than the num- for non-Catholics; and a Jewish variable
ber of categories in the original variable. similarly coded 1 and 0 (Table 18.1). There
In this example, we would create three is no distinction among members of a cate-
new variables (there were four categories gorythat is, all 1s are considered equally
in our nominal religious afliation vari- Catholic and all 0s are considered equally
able). The new variables are coded as ei- non-Catholic. These three new variables
ther presence or absence of the charac- would then be entered into the regression
teristic by assigning the value of 1 to analysis as independent variables along
the cases with the characteristic and 0 with the other relevant independent vari-
to those without the characteristic. The ables.
reason we can fully describe the religious Another example of a dummy variable
afliation variable with three rather than is the inclusion of a gender variable into a
four new variables is because the fourth regression analysis. In this case, either
( in this case) is other, and other will the males or the females would be as-
signed a value of 1, and those not assigned ample, if you have a variable such as age
would be given a 0. Thus, if females were coded as age at last birthday, then use
assigned the value of 1, then the males the raw age variable data (e.g., 24, 19, 37,
would be given the value 0. In this case, 54, and so on) rather than recoded age
only one new variable is needed to repre- categories that may have used three
sent the two categories of gender (recall categories to represent those under age
that we use one fewer dummy variable 30 years, those age 30 to 49 years, and
than categories of the original variable). those age 50 years of age or older.
Tip 5. Use the BACKWARD solution in the
5. Some Tips for Using Regression REGRESSION procedure. In nursing re-
Analysis search, a great deal of intercorrelation
exists among variables. In regression,
Tip 1. Ensure that variables are theoreti- one wishes to nd the smallest group of
cally independent of one another. variables that best predicts the depen-
What this means is that you cannot use dent variable or accounts for the largest
aspects of the dependent variable as proportion of variance in the depen-
independent variables. Ensure that you dent variable. By using the BACKWARD
are including only meaningful potential (stepwise) solution, all the variables are
causes of the dependent variable, not included in the regression equation;
alternate measures of it. then the least important variable is
Tip 2. Watch out for highly correlated in- dropped and the equation is recalcu-
dependent variables. The weighting lated. This procedure is repeated until
that is attached to the variables will be only signicant variables remain. The
unstable if the independent variables advantage of this format is that vari-
are highly correlated with one another. ables that are important when in com-
The program will print out a warning if bination with other variables will re-
there is a problem in this area. The term main in the equation; in other formats,
multicollinearity is used to refer to the they might never be included. This is
extent of the correlation among the in- the method used in the computer analy-
dependent variables. To achieve high sis example presented in this chapter
predictability of a dependent variable, it (see Table 18.2).
is preferable to have the independent Other commonly used formats for en-
variables correlated with the depen- tering variables include hierarchical
dent variable but not with one another. and stepwise with the forward solution.
If you have a number of highly intercor- In hierarchical regression, the re-
related independent variables, it is usu- searcher determines the order of entry,
ally advisable either to develop an index or hierarchy, of the variables into the
out of them (if it makes sense to do so) regression equation. This technique is
or to select one of the measures and use used in creating path models (see
it to represent the others. Chapters 2 and 3). There is usually a
Tip 3. Try to achieve ratio level measure- theoretical rationale for the order in
ment. If you intend to use regression which the variables are entered. Hier-
analysis, attempt to collect data in as archical entry is often used when you
raw a form as possible, at the ratio level have a set of given variables that are be-
if possible; if ordinal data are collected, yond the control of the researcher (e.g.,
use more categories, rather than fewer. such as age, gender, or socioeconomic
Tip 4. Use raw data. Do not use recoded status [SES]) and you are interested in
variables (dummy variables are an ex- determining if a specic nursing inter-
ception) in regression analysis. For ex- vention will impact a nursing outcome,
above and beyond the factors that you ables are substantially correlated with
cannot change. For example, if you one another. (In such situations, with
wanted to know if a relaxation class re- another sample, it is quite likely that
duces the number of cigarettes smoked different variables will be selected as
by participants, you would enter the signicant predictors of the depen-
given variables (age, gender, SES) rst dent variable.)
followed by your intervention last. Tip 7. Monitor number of cases care-
With the stepwise entry solution, the fully. By default ( i.e., if you do not pro-
independent variable that has the great- vide a specic instruction to the con-
est correlation with the dependent vari- trary), SPSS deletes a case if it has a
able is entered rst. The next variable missing value in any of the variables in
entered is the one that will increase the the equation. Thus, if there are a lot of
R2 the most above the rst variables variables in an analysis, there is a dan-
contribution. Because there are inter- ger of losing many cases. And as the
correlations among the independent number of cases drops close to the
variables, we cannot simply select the number of variables, the R2 increases
independent variable with the next dramatically. To determine the number
highest correlation with the dependent of cases used in the analysis, add one to
variable; rather, we must rst calculate the total degrees of freedom reported in
partial correlations between the re- the table.
maining independent variables and the If a large number of cases have been
dependent variable. The partial correla- dropped, one should attempt to detect
tion removes the effect of the rst vari- if there is a pattern to the missing cases.
able from the correlation. The variable If one subcategory of respondents is
that has the highest partial correlation more likely to respond to a question,
with the dependent variable enters the one might wish to analyze the data
regression next and so on until no re- these respondents provide separately
maining independent variables con- from those subcategories of respon-
tribute signicantly to the R2 value. At dents whose responses are less com-
this point, the analysis is ended. plete. However, if it appears that the
At times, the stepwise forward solu- missing values are random, consider
tion is combined with the backward so- using one or more of the following tech-
lution to overcome the weakness of niques:
each approach. When this happens, the Repeat analysis: After an initial re-
variables are entered into the regres- gression analysis (using the BACK-
sion according to the forward solution WARD option) has identied the sig-
method but they are assessed at each nicant variables, rerun the analysis,
step using the backward method to de- naming only the signicant variables,
termine whether their contribution re- plus perhaps two or three that were
mains signicant given the inuence of dropped in the last few steps. This
the other variables in the equation. will preserve the cases that were
Tip 6. Interpret weightings with care. dropped because of missing values in
Understand that the weightings are de- variables that are not in the nal
signed for the particular combination equation. Frequently, many fewer
of independent variables in a particu- cases will be dropped if this proce-
lar sample and that they may not be re- dure is followed.
liable if applied to other samples. One Pairwise solution: Try running the
has to be particularly cautious in situ- analysis using PAIRWISE treatment
ations in which the independent vari- of missing cases. In this solution, the
correlations are determined for all dependent variable, as in the linear re-
the pairs of variables for which data gression equation.
are available. (On the linear regres-
sion screen, click on the Options 6. Presenting and Interpreting
button to choose the PAIRWISE Regression Results
treatment).
Means solution: A third approach is Table 18.2 presents a sample of a regres-
to try the MEANS treatment of miss- sion results table. Note that both the b co-
ing values, which will substitute the efcients and the beta weights are re-
mean of the variable for any missing ported. It is possible to hand compute an
cases. (On the linear regression estimate of the impact of each indepen-
screen, click on the Options button dent variables by using the following for-
to choose the MEANS solution). mula (Hamblin, 1966):
Tip 8. Deal with interactions among in-
dependent variables. If you have rea- % Variance explained 1 R2
son to suspect that the joint effect of by each variable 100
s
two independent variables is important
but that, individually, they may not be This estimate represents the impact of
signicant predictors of the dependent each variable in the equation. However, if
variable, there are two relatively easy other variables were included, the per-
approaches to this problem: centages would change.
Create a new variable by multiplying The regression equation is included in
the values of the two variables sus- Table 18.2. This equation will help you
pected of interaction and include identify where each of its elements are
the new variable in the regression found in a printout; note that the b coef-
analysis along with the variables cients (not betas) are used along with the
from which it was constructed. If the constant term.
new variable is a signicant predic- Pure researchers have particular in-
tor, it will remain in the equation terest in the beta weights because they
when the analysis is done; if it is not, provide a basis for directly comparing
it will be dropped. the impact of the different independent
Convert all the variables to log func- variables on the dependent one. Applied
tion variables (this converts the researchers are usually more concerned
equation to a multiplicative power with b coefcients, especially those that
function). By doing so, any interac- can be changed through policy alter-
tions will be taken into account in the ations. They provide the basis for under-
weightings of the independent vari- standing how much change in the depen-
ables. In effect, what this procedure dent variable may be produced for each
does is to raise the weightings to a change in the independent variable. Vari-
power. This is accomplished by ables can be grouped for presentation to
transforming the elements to log illuminate the degree to which different
functions, as in: variable types inuence the dependent
variable (e.g., how much of the variation
log Y log a log X1 log X2 log Xk in the dependent variable is caused by
socioeconomic variables in comparison
The consequence of using log trans- to the variety of experience variables?).
formations is that each independent A second method for estimating the rel-
variable is raised to a power rather ative importance of variables is to calcu-
than multiplied by the value of the in- late a part correlation coefcient. This coef-
Table 18.2 Multiple Regression Analysis for Registered Nurses Salaries

Variable b Coefficient Beta Coefficient Percent Explained
Qualications 794.0 0.038 3.1
Contract status 1326.0 0.065 5.3
Age in current year 94.0 0.092 7.5
Years in position 263.0 0.153 12.5
Professional age 210.0 0.172 14.0
Years at hospital 264.0 0.250 20.4
Professional rank -4570.0 -0.372 30.3
Constant 38504.0 % Explained 93.1
Multiple R 0.964
R2 0.930
The variables are:

Qualications: Baccalaureate degree 1; no degree 0
Contract status: Permanent contract 1; nonpermanent contract 0
Age Age in 2001
Years in position: Years in position in 2001
Professional age: Years since RN license earned
Years at hospital: Years working as RN at institution
Rank: General practice nurse 1; specialty practice nurse 2; advanced practice nurse 3; clinical nurse practitioner 4
Salary 38504 794 (qual) 1326 (contract status) 94 (age) 263 (rank) 210 (proage) 264 (years) 4570
(rank)
cient is calculated for each independent leisure time activities; and psychosocial
variable and represents the difference in variables of self-esteem and coping mech-
the R2 when the variable is included versus anisms. Based on a study of school-aged
when it is not included. The variables may youths, discriminant analysis would then
then be ordered in terms of the unique be used to predict whether each case
contribution of each independent variable. would fall into the smoking or no
To get part correlations in SPSS, click on smoking categories of cigarette smoking
Statistics and click on the Part and Partial behavior.
Correlations window.
1. The Rationale
B. DISCRIMINANT FUNCTION Discriminant function analysis has some

ANALYSIS similarity to regression analysis. This al-
ternative is used in situations in which:
Discriminant function analysis attempts
to predict the category of the dependent
The measurement of the dependent
variable is at the nominal level.
variable into which each case falls by us-
ing the combined information from the
Regression analysis is inappropriate
because the assumption that the de-
independent variables. For example, sup-
pendent variable is normally distrib-
pose you wished to predict which chil-
uted is not met.
dren would initiate a smoking habit in
adolescence. The independent variables Discriminant analysis is a valuable tool
included tobacco use by parents, sib- for nurse researchers and has applica-
lings, friends, and teachers; students tions in many situations. To illustrate
some of these situations, let us consider The coefcients computed are based
applications that might be made of the on a regression-like linear equation:
technique in various areas of nursing re-
search: D B0 B1 X1 B2 X2 . . . Bk Xk
Epidemiology: Studies of the distribu- The X values are the values of the inde-
tion of residence patterns of families pendent variables, and the B values are
who have tested positive and those that the coefcients associated with each in-
have tested negative for tuberculosis dependent variable, weighted to maxi-
Community health: Studies of rates of mize the prediction of D, the categories of
participation and nonparticipation in the dependent variable. The B values are
pre- and postnatal education, of par- weighted to maximize the ratio of the be-
ticipation in a parenting class, or of the tween-groups sum of squares to the
rates at which adolescents drop out of within-groups sum of squares. The output
school or stay in school for the analysis will display the discrimi-
Health promotion: Attempts to deter- nant function score for each case. Plug-
mine whether individuals are exercis- ging in the observed values for each vari-
ers or nonexercisers or smokers or able and multiplying it by the coefcient
nonsmokers and the distribution of derives the score for each case. Suppose
clients who favor the use of nurses, the discriminant equation was as follows:
nurse practitioners, or physicians for
annual health assessments D 0.013 0.003(X1) 0.004(X2 )
Psychiatric nursing: Classifying clients 0.078(X3 ) 0.056(X4 )
as mentally ill or mentally healthy or to
differentiate between leaders and non- Table 18.3 displays the values for each
leaders in group therapy sessions variable, the coefcients, and the result-
School health: Attempts to distinguish ing discriminant score for the rst case in
between healthful and nonhealthful the le.
behavior in school-aged children Discriminant scores are calculated for
each case, and each case is then classi-
ed into one of the groups ( in this exam-
2. Comparison with Multiple
Regression Analysis ple, participating or not participating in
prenatal education). The printed output
Similar to multiple regression analysis, indicates the actual group each case be-
discriminant analysis has the ability to si- longs to, and asterisks are used to indi-
multaneously deal with multiple indepen- cate the cases that were misclassied.
dent variables. So, if your dependent vari-
able has two, three, or four categories; 3. Presenting and Interpreting
your independent variables are ratio vari- Results
ables; and you wish to assess the extent
to which you can correctly classify cate- A basic statistic provided by discriminant
gory membership on the dependent mea- analysis is the percentage of cases that
sure, consider using discriminant func- can be classied correctly using informa-
tion analysis. And as with regression tion from the combination of independent
analysis, if you wish to use nominal or or- variables. The statistic calculated is simi-
dinal independent variables, it is possible lar to that of Lambda, which was de-
to do so (using dummy-coded variables), scribed in the section dealing with cross-
as long as caution is exercised in inter- tabular table analysis (see Chapter 11).
preting the results. You may recall that Lambda computes the
Table 18.3 Calculating the Discriminant Score, Case 1

Variable Observed Value Coefficient Discriminant Score
D dependent variable* 1
Constant 0.013 0.013
X1 maternal identity score 72.6 0.003 0.218
X2 family socioeconomic score 47.0 0.004 0.188
X3 family size 3 0.078 0.234
X4 number of previous pregnancies 0 0.056 0.000
Sum of values, case #1 0.653
*Dependent variable is Participating or Not Participating in prenatal education.
error reduction in estimating a dependent variable, the results are different. Table
variable, given knowledge of an indepen- 18.4 provides a sample of the way in which
dent variable. In our example, using the discriminant analyses may be presented.
computed discriminant function equation, Bertrand and Abernathy (1993) used
we determine that 85.2 percent of the discriminant function analyses to predict
cases were correctly grouped according three categories of smoking behavior
to the discriminant function analysis. among school-aged children. Data were
As with regression analysis, the com- collected over a 3-year period and included
puted coefcients are available in stan- the predictor variables of peer inuence,
dardized and unstandardized forms. The self-esteem, mental health, leisure time ac-
unstandardized coefcients are used when tivities, and parentchild relationships.
calculating the discriminant scores for The investigators hoped to identify groups
each case. When the researcher wishes to of children at risk of initiating smoking so
compare the relative impact of each vari- that they could be targeted early and of-
able, standardized coefcients are used fered effective smoking prevention pro-
(this is to take into account the different grams. Children were classied into the
ranges of the variables). Indeed, if there categories of never smoked, tried but
were two categories in the dependent vari- quit, and current smoker. Results indi-
able, we would have achieved similar re- cate that the highest rates of correct clas-
sults if we had used regression analysis. sication were with students who fell into
The b coefcients in regression analysis the categories of never smoked and cur-
have a similar ratio to the B discriminant rent smoker. The group tried but quit
coefcients. However, when there are was less likely to be correctly classied.
three or more categories in the dependent Table 18.5 presents the hit rate for clas-
Table 18.4 Discriminant Analysis, Sample Presentation

Predicted Group Membership
Actual Group Cases, n 1 2
Participate (1) 261 221 40*
Not participate (2) 83 11* 72
Total 344 232 112
Percent of grouped cases correctly classied: 293 out of 344 cases 85.2%. (221 + 72 = 293)
*Cases misclassied
Table 18.5 Discriminant Analysis, Number of Children Correctly Classified into Smoking Categories
Predicting from Sixth Grade
Never Smoked, Tried and Quit, Current Smoker,
N % N % N %
Grade 7 1074 64.5 72 20.9 33 52.4
Grade 8 792 59.7 70 15.4 126 49.6
Grade 9 613 53.4 152 30.4 148 40.9
These percentages reect the number of students correctly classied out of the total number actually falling within that
smoking category (hit rate).
sication among the three smoking cate- someone watched TV or went to the
gories produced from the analyses. When movies, you again would probably opt for
the accuracy of prediction is examined for analysis of variance. In short, if an inde-
each category, we see that approximately pendent variable is made up of values
53 percent of students who reported they that differ in kind rather than in quantity,
had never smoked in ninth grade were cor- you would opt for analysis of variance.
rectly classied on the basis of their sixth A second situation that is better han-
grade responses. The percentages rise to dled with analysis of variance is one in
65 and 58 percent for seventh and eighth which the relationship between the inde-
grades, respectively. pendent and the dependent variable
changes over the continuum. Perhaps the
relationship is nonlinear and, once again,
C. MULTIVARIATE ANALYSIS analysis of variance should be considered
OF VARIANCE as an alternative to regression. For exam-
ple, suppose you were examining the rela-
Multivariate analysis of variance tionship between job satisfaction and size
(MANOVA) is a method used in examin- of the institution in which respondents
ing the relationship between ratio level work. Perhaps satisfaction increases as
dependent variable (or variables) and one moves up the size scale, levels off, and
two or more nominal, ordinal, or ratio then declines in work places above 500
level independent variables, treatments, employeesa nonlinear relationship. In
or covariates. this caseand particularly if other covari-
ates such as workers age and size of home
community are included in the analysis
1. The Rationale one would perhaps choose a multivariate
analysis of variance technique.
Although regression analysis is appropri-
It should be noted that multivariate
ate in many situations, there are several
analysis of variance is appropriate under
situations in which multivariate analysis
the following conditions:
of variance (ANOVA or MANOVA) would
be more appropriate than regression When the dependent variable is mea-
analysis. For example, as Kachigan (1986) sured at the ratio level
points out, if you had predictor variables When one or more of the treatment
that have qualitative differences (e.g., variables is measured at the ratio level
three different countries) or if, rather and others are measured at the nomi-
than having a ratio variable measuring nal or ordinal levels
the amount of time spent watching TV, If you have multiple dependent mea-
you had a variable that indicates whether sures that you wish to examine simul-
taneously; this is often the case in signicant interactions among inde-

nursing and health-related research in pendent variables
which one is interested in the impact
of an intervention on more than one In these cases, you would consider
outcome variable using MANOVA. The ratio level treat-
In nonexperimental designs in which ment variables are called covariates. Box
you wish to examine whether there are 18.1 presents sample results. Note how
BOX 18.1 MANOVA Analysis, Sample Presentation: Analyzing

Egalitarianism
Analysis of Variance
Source of Variation Sums of Squares Mean Square df F Significance of F
Within cells [error] 11539.00 69.10 167
Regression 127.71 63.85 2 0.92 50.399
Country [A] 932.68 466.34 2 6.75 0.002
Gender [B] 2.18 2.18 1 0.03 0.859
Country by gender[A B] 87.92 43.96 2 0.64 0.531
Regression analysis for within cells error term:

Dependent variable: Egalitarianism
Covariate Beta T Value Significance of T

Socioeconomic status 0.024 0.311 0.756
Family size 0.100 1.290 0.199
The variables are as follows:

Dependent variable: Egalitarianism
Independent variables: Country (Canada, United States, Australia); gender (males, females)
Covariates: Socioeconomic status (occupational prestige rating); family size (number of children in family)
Understanding the Values in the Summary uct of the dfs making up the interaction
Table (in this case: 2 1 2)
a. The degrees of freedom are calculated Error term: df total number of cases
as follows: minus the product of number of levels
Regression: df number of covariates of A and B (in this case, 173 [3 2]
in analysis (in this case: 2) 167)
Main effect (A): df 1 less than number b. To compute the mean square for an
of levels in that factor (in this case: 3 effect, divide the sum of squares for the
1 2) effect by its df (for the effect of Country:
Main effect (B): df l less than number 932.68 2 466.34)
of levels in that factor (in this case: 2 c. To compute the F value for each ef-
1 1) fect, divide the mean square by the mean
Interaction effect (A B): df prod- square error term (the F for Country:
466.34 69.10 6.75)
BBOX 18.1 MANOVA Analysis, Sample Presentation: Analyzing

Egalitarianism (Continued )
Figure A
Males
EQUALITARIANISM
Females
NEW CANADA AUSTRALIA

ZEALAND
Figure B
Males
EQUALITARIANISM
Females

ZEALAND
Figure C
Females
EQUALITARIANISM
Males

ZEALAND
A No interaction. This chart illustrates a situation in which there is a significant main effect of Country on
Egalitarianism, but the plot of the means of egalitarianism by Gender categories indicates that because the males in each
country remain about the same distance above the mean scores for the females, there is no interaction between
gender and country. B Nonsignificant interaction. Similar to the previous chart, this one features a significant main
effect (Country) and shows a weak but not significant interaction of Gender by Country. (This is the data from those
reported in Figure 18.1.) C Significant interaction. This chart indicates a significant Gender by Country interaction, as
well as a significant main effect (Country). In this case, there appears to be no difference between the genders in terms
of egalitarian scores in the United States, but the gap between the genders in egalitarianism increases as one moves
from the Canadian data to the Australian data.
MANOVA results combine standard an- It is possible to analyze a data set us-
alysis of variance results along with re- ing both analysis of variance and regres-
gression analysis. (See the discussion of sion techniques. The regression analysis
analysis of variance in Chapter 12.) would simply use dummy-coded vari-
ables to match the categorization used in
the analysis of variance. The key differ-
2. Comparison with Multiple ence between the two techniques is
Regression Analysis shown in Box 18.2.
Although the residuals in regression
Although the strength of regression analy-
analysis are measured in terms of devia-
sis lies in its ability to take many variables
tions from the regression line (see Chap-
into account simultaneously and deter-
ter 11), the residuals in analysis of vari-
mine the mathematical equation that best
ance are measured as deviations from
describes the relationship between inde-
group means. As Iversen and Norpoth
pendent variables and the dependent vari-
point out:
able, the strength of analysis of variance
lies in its ability to assess interactions If the regression line passes through all
among variables and estimate the extent group means, the residual sum of
to which the results of a study might be squares will turn out the same for the re-
caused by random sampling uctuations. gression analysis as for analysis of
Hence, analysis of variance is most asso- variance. But if the relationship be-
ciated with tests of signicance (see Chap- tween X and Y is not linear, then the re-
ter 12), and regression analysis focuses on gression line will not pass through all
predictive equations. Regression analysis group means, and as a result, the re-
uses analysis of variance as a test of sig- gression residual sum of squares will
nicance for the inuence of the indepen- exceed the analysis of variance resid-
dent variables in the equation. ual sum of squares. (1976, p. 91)
BOX 18.2 Comparing the Correlation Coefficient and Analysis

of Variance
A. Correlation coefficient. In the case of In the case of analysis of variance, treatment ef-
the correlation coefficient, the sum of the fects are reflected by the sum of squared devi-
squared deviations in the numerator refer ations between treatment categories; random
to deviations from predicted values; the de- error refers to the sum of squared deviations
nominator uses sum of squared deviations within the treatment categories. The F repre-
from the mean of Y. The denominator mea- sents the ratio between the two types of devi-
sures within category variation. Note that ations.
the correlation coefficient refers to the ra- Note that in both cases, we use the sum of
tio of the two types of variation measured squared deviations from relevant means. For
subtracted from 1. both statistics, the denominator reflects vari-
Variations around regression ations within a category. The numerators are
r2 1 different. For the correlation coefficient, the
value reflects deviations from predicted val-
B. Analysis of variance (f score). ues; in the case of analysis of variance, the nu-
F Variation within category merator reflects variations between cate-
Variation between categories gories.
Variation within category
Thus, if there is a substantial difference nicant main effect, you should then
in the residual sums of squares between inspect your results for any interac-
analysis of variance and regression analy- tions between the effects. Your job is
sis, it would indicate a nonlinear relation- simple if there is no signicant interac-
ship between the variables. If there is not tion effect: you simply report that
much difference in the residual sums of none was present. It is more compli-
squares, it would lead us to conclude that cated with an interaction; in this case,
the relationship is, indeed, linear. the effect of a treatment is not indepen-
Although generally seen as the tech- dent of the other treatment. Rather, it
nique associated with experimental and may be that under the high condition
quasi-experimental designs, analysis of of treatment A, we nd that treatment
variance can also be used in a variety of B does not enhance the impact on the
nonexperimental situations, particularly dependent variable. Here you need to
those in which there are a limited number indicate the conditions in which the in-
of independent variables. In all cases, teraction occurs. You might, for exam-
however, the dependent variable should ple, create a plot of the relationship be-
be measured at the ratio level. tween the variables, such as one of the
ones shown in Box 18.1 gures A, B, C.
3. Presenting and Interpreting Some aspects of analysis of variance

Results were also discussed in Chapters 11 and
12.
In inspecting your results shown in Box
18.1, you should examine two issues as
you begin to interpret your results. First,
D. FACTOR ANALYSIS
you will want to know if your treatment
(or independent) variables had a signi-
Variables that are correlated can be
cant effect; if there are covariates, is the
grouped together into a factor. A factor is a
regression effect signicant? If there are
hypothetical entity that represents related
signicant treatment effects, you should
concrete measures taken on research par-
see if there is any interaction between the
ticipants. For example, in a study of the ef-
treatments.
fectiveness of a variety of nursing inter-
a. Main effects. To decide if a treatment ventions in reducing postchemotherapy
or covariate was statistically signi- pain in patients with breast cancer, the re-
cant, you will need to examine the sig- ported excellence of the experience with
nicance of F column to see if the meditation, guided imagery, and chanting
value is less than 0.05. (If this value is may be grouped together into a factor
not shown, you will need to compare called spirit-centered interventions. Fac-
the F value with the values found in tor analysis, then, is a multivariate empiri-
Table 12.4 using the appropriate de- cal procedure used to determine the un-
grees of freedom. If the value you ob- derlying structure of a set of variables
tained is larger than the one reported that is, to show how variables cluster to-
in Table 12.4, you have a statistically gether to form a unidimensional construct.
signicant effect. To use Table 12.4, It determines the structure of a set of vari-
you will need values for the required ables by analyzing the intercorrelations
degrees of freedom. Use the smaller among them. It sorts the variables into
value across the top of the table and categories according to how closely re-
the larger one on the vertical axis.) lated they are to other variables. Within a
b. Interaction effects. If you have a sig- given data set, a varying number of factors
may emerge. For example, in this example, ther using procedures such as regres-
we may have spirit-centered interventions, sion or analysis of variance.
body-centered interventions, family and Instrument development. The devel-
social interventions, and so on. opment of new instruments is a major
use of factor analysis. Instrument de-
velopment usually proceeds with a
1. The Rationale large number of items used to repre-
sent the research phenomenon. The
Factor analysis provides a measure of the item pool can be factor analyzed to de-
amount of variance in the data set ex- termine which items cluster together
plained by a particular factor as well as the and should be retained and which
amount of variance in the factor explained items should be discarded. Gillis (1997)
by a particular variable. It is a somewhat used factor analysis to reduce 66 items
controversial procedure because it in- into a 43-item measure that isolated
volves a greater degree of subjectivity seven dimensions to a healthy lifestyle.
than is associated with most statistical Instrument validation. Factor analysis
analyses. After a factor is identied math- is an important tool for validating new
ematically, the researcher provides a research instruments (see Chapter 13).
name for it (this is the subjective part) by In nursing research, it is used most fre-
examining the variables included in the quently for this reason. Factor analysis
factor and deciding what the underlying offers an empirical method of identify-
factor is all about. He or she then labels it ing the underlying factors that compose
accordingly. Another researcher may, of a construct. Although the researcher
course, apply a somewhat different label may a priori identify various dimensions
to the identied factor. to the research construct, only factor
The major applications of factor analy- analysis can offer empirical support for
sis include: the researchers conclusions.
Theory development. Factor analysis is
Data reduction. A researcher may col- used to identify structures of variables
lect a large amount of data, particularly or constructs that can be interpreted
in survey research, and then reduce the meaningfully by the researcher. The re-
data through factor analysis. A single searcher explains why variables that
composite variable or several variables have been dened mathematically are
measuring different dimensions of a clustered together the way that they
concept may be created out of a con- have emerged. It can also lead to the
glomerate of variables through factor conceptualization of research phenom-
analysis. It may be used to sort out ena and their underlying dimensions
meaning from a large number of survey and show important intercorrelations
items on a questionnaire and reduce a and relationships among constructs.
large number of variables into a smaller, These are important steps in theory de-
more manageable set of measures. For velopment.
example, a survey may collect data on
50 indicators of an individuals per- There are two types of factor analysis,
ceived health status. Through factor exploratory and conrmatory analysis.
analysis, the data may be reduced to Exploratory factor analysis is used when
three dimensions or factors, namely a a researcher does not a priori identify the
physical dimension, a psychological di- factor structure of the variables in the
mension, and a social dimension. The data set. It is similar to stepwise regres-
reduced data can then be analyzed fur- sion analysis in that the variance of the
rst factor identied is partialed out be- ned mathematically as a linear combi-
fore analysis of the second factor pro- nation of variables in a data matrix and is
ceeds. Conrmatory factor analysis is represented by the following equation:
usually conducted after an exploratory
analysis has occurred or when a re- F b1 X1 b2 X2 b3 X3 . . . bk Xk
searcher has a hypothesis about the ex-
isting factor structure that is to be tested. where F a factor score (each factor is
It is usually based on theory. computed separately); X1 to Xk values
Computer programs provide many op- on the k original variables; and b1 to bk
tions when doing a factor analysis. Typi- weights (these are the factor loadings).
cally, a researcher can specify the vari- An eigenvalue is the sum of the
ables to be included; the basis on which squared factor loadings for any one factor.
variables are selected for inclusion; and Thus, the eigenvalue is a measure of how
how the second and additional factors much of the total variance is explained by
should be identied to be independent of a given factor. The higher the eigenvalue
the rst (orthogonal rotation) or simply relative to other ones, the greater the vari-
the second-best factor (oblique rotation). ance explained by that factor.
Factor analysis is a complex, widely A correlation matrix is said to be fac-
used technique in nursing research. A full torable if there are reasonably robust
discussion of the method is beyond the correlations among the items. Ordinarily,
scope of this chapter. Interested readers one would expect the interitem correla-
are referred to statistical texts referenced tions to average above 0.25 to consider
at the end of this chapter and to the work the variables to be worthy of factor analy-
of Ferketich and Muller (1990), who pro- sis. Any lower than that and you may not
vide a comprehensive decision tree that nd that any factors would be extracted.
describes the key decision points and op- In factor extraction, the analyst seeks to
tions available to researchers who wish to determine the number of factors that need
conduct factor analyses. to be extracted in order to capture or ex-
plain the variation in the data set. To pro-
2. Basic Terminology of Factor ceed, the analyst will then select a method
Analysis for how additional factors will be identied
through a process of factor rotation.
Factor analysis begins by computing a A marker variable is one that pro-
correlation matrix (a table showing the duces the highest correlation with the un-
correlation coefcients between all desig- derlying factor that is identied in the
nated variables). As such, the analysis is analysis. The marker variable is useful to
assuming ratio level measurement of vari- researchers in attempting to label the un-
ables and normal distributions of vari- derlying factor.
ables. Ordinarily correlations between In determining the second factor, a ro-
variables need to average 0.25 or higher in tation is required, either (1) an orthogonal
order for the data set to be considered fac- (varimax is the most common version of
torable. orthogonal rotations) forces a 90 degree
Factor loadings are the b weights (sim- rotation, producing an entirely indepen-
ilar to the bs in regression analysis) in the dent solution or (2) an oblique rotation
factor analysis equation. Factor loadings that selects the second best solution but
may be combined to yield a factor score, does not insist on the 90 degree rotation.
which measures the extent that factor There are two main ways in which fac-
scores are similar thus reecting the un- tors are extracted: (1) the principal com-
derlying common factor(s). A factor is de- ponents method, which includes an analy-
sis of all the variance in the observed vari- factor (1, 2, 3, 4, 7). The second identied
ables, not just the shared variance among factor included items 5 and 6. In this case,
the variables; and (2) the principal factors the marker variable is item 5, with a fac-
method (known in SPSS as the Principal tor loading of 0.610, Abortion should be
Axis Factoring method). This approach re- entirely the personal choice of the
lies on the shared variances of the vari- woman involved. The other variable that
ables included in the analysis and appears loads on this factor is item 6, There is
to be the most commonly used approach. nothing wrong with being homosexual.
At this point, the researchers task is
3. Presenting and Interpreting to identify and label the factor that seems
Results to underlie the two identied liberalism
dimensions. Examining the marker vari-
Tables 18.6 to 18.8 report results from a ables and the other variables that loaded
factor analysis of variables designed to on the two factors, the researcher might
measure liberalism. There were seven label the rst one as a public policy liber-
variables included, and these were based alism factor because all of the variables
on nine-point Likert-type questions have to do with government programs
(Strongly Disagree to Strongly Agree) and spending that might be used to in-
posed to students in three countries. The crease the equality of participation of mi-
items were as follows: nority groups. The second factor made
up of two variables appears to have more
1. My government should increase its
to do with individual rights liberalism.
provision for foreign aid.
If the researcher wishes to compute
2. Quotas in job hiring should be used to
factor scores for individual cases, one
increase the numbers of visible mi-
could proceed in one of two ways:
norities in good jobs.
3. Visible minorities in this country are 1. Simply add together the individual raw
not given an equal chance in education scores on the 5 variables in factor 1
compared with those in the majority. (public policy liberalism).
4. The immigration policies of this gov- 2. Multiply the raw scores by the factor
ernment are too strict. loading for each variable and add
5. Abortion should be entirely the per- these together.
sonal choice of the woman involved.
Suppose the rst two individuals in the
6. There is nothing wrong with being ho-
le had the following raw values on the
mosexual.
seven questions:
7. The government does not pay enough
to welfare recipients.
Question # 1 2 3 4 5 6 7
Using the principal factors method, the
Respondent 1: 7 8 7 6 3 4 9
analysis extracted two factors, which ac-
count for a total of 55.4 percent of the Respondent 2: 3 4 3 5 8 9 5
variance (see Total Variance Explained in
Tables 18.6 to 18.8). The marker variable 1. Simple Addition Method for Deter-
( i.e., the one with the highest correlation mining the Factor Score: Here we sim-
to the underlying dimension measured) ply added together the raw scores for
for the rst factor, with a factor loading of the appropriate variables. To get the
0.636, is item 7, The government does result on the public policy liberalism
not pay enough to welfare recipients. As factor, we would add the following
indicated by the factor loadings, ve numbers together to get the result for
items loaded predominantly on this rst the two respondents:
Table 18.6 Factor Analysis of Liberalism Measures

Extraction Sums of Rotation Sums of
Initial Eigenvalues Squared Loadings Squared Loadings
% of Cumulative % of Cumulative % of Cumulative
Factor Total Variance % Total Variance % Total Variance %
1 2.471 35.295 35.295 1.842 26.320 26.320 1.622 23.165 23.165
2 1.405 20.075 55.370 0.839 11.993 38.313 1.060 15.148 38.313
3 0.801 11.449 66.810
4 0.650 9.283 76.102
5 0.642 9.167 85.269
6 0.549 7.841 93.109
7 0.482 6.891 100.00
Table 18.7 Factor Matrix*: Factor 1: Public Policy Liberalism

Factor 1 Factor 2
Government increase foreign aid 0.546 0.360
Increase number of visible minorities in jobs 0.535 0.179
Visible minorities are not given equal education 0.550 4.76E-02
Immigrant policies too strict 0.513 0.243
Abortion is a womans personal choice 0.310 0.610
Homosexuality is okay 0.438 0.490
More money should be given to welfare 0.636 6.161E-02
Extraction method: principal axis factoring.

*Two factors extracted; 16 iterations required.
Table 18.8 Rotated Factor Matrix*: Factor 2: Individualism Rights Liberalism

Factor 1 Factor 2
Government increase foreign aid 0.651 6.20E-02
Increase number of visible minorities in jobs 0.556 9.243E-02
Visible minorities are not given equal education 0.508 0.216
Immigrant policies too strict 0.567 2.612E-02
Abortion is a womans personal choice 1.20E-02 0.684
Homosexuality is okay 0.157 0.639
More money should be given to welfare 0.533 3.53
Extraction method: principal axis factoring; rotation method: varimax with Kaiser normalization.
*Rotation converged in three iterations.
a. Public Policy Liberalism Factor b. Individual Rights Liberalism Fac-

Score tor Score
Respondent 1: 7 8 7 6 9 37 Respondent 1: 3 4 7
Respondent 2: 3 4 3 5 5 20 Respondent 2: 8 9 17
2. Multiply Factor Loadings by Raw analysis and reliability analysis led to

Scores to Determine Factor Scores. similar conclusions regarding the appro-
This is the preferred method because it priate items for the rst factor in mea-
weights the individual scores accord- suring liberalism.
ing to their relationship to the underly-
ing factor. In this case, we use the fac-
tor scores to weight the raw scores. 5. Comparison with Regression
Some analysts use all of the factor Analysis
weightings to determine these scores,
but others simply use the ones related Factor analysis is similar to regression
to the underlying factor. We will use analysis in that it involves the formation of
the latter method because it more linear combinations of variables. In re-
clearly represents the factor alone. Us- gression analysis, the researcher develops
ing the information in Factor Loadings equations that test hypotheses composed
in Tables 18.6 to 18.8, the following cal- of independent and dependent variables.
culations are made to demonstrate In factor analysis, equations are developed
how these factor scores can be com- that test the interrelationships among a
puted. (If you were using SPSS, these large number of variables and produce
could be done quickly with a COM- clusters of variables (factors) that are
PUTE procedure.) closely related to each other. Factor analy-
a. Public Policy Liberalism Factor sis requires larger sample sizes than re-
Score (Weighted by Factor 1 Load- gression analysis. Nunnally (1978) sug-
ing) gests a minimum of 10 participants for
Respondent 1: (7 0.546) (8 each variable. Thus, if there are 50 vari-
0.535) (7 0.550) (6 0.513) ables being factor analyzed, the sample
(9 0.636) 20.754 should be approximately 500 cases.
Respondent 2: (3 0.546) (4 The equation for computing the factor
0.535) (3 0.550) (5 0.513) loadings is similar to the one used for lin-
(5 0.636) 11.173 ear regression analysis. Indeed, the b co-
b. Individual Rights Liberalism Fac- efcients are used as the factor loadings
tor Score (Weighted by Factor 2 for factor analysis. If you prefer, the re-
Loading) gression equation is the same as the
Respondent 1: (3 0.684) (4 equation for the rst factor in factor
0.639) 4.608 analysis. One should also note that both
Respondent 2: (8 0.684) (9 regression analysis and factor analysis
0.639) 11.223 use the correlation matrix as a starting
point and both share common assump-
tions about ratio measurement of vari-
4. Comparison with Cronbachs ables and normally distributed variables.
Alpha Procedure The major difference between the two
methods is that in the case of regression
The seven variables that were designed analysis, we are computing weightings on
to measure liberalism were also run using independent variables to maximize the
the RELIABILITY procedure in SPSS. This prediction of a dependent variable for
analysis indicated that the abortion and which we have values. In the case of factor
homosexuality items should be dropped analysis, we do not have a measured de-
from the index in order to maximize Cron- pendent variable; instead, in the computa-
bachs alpha score. In short, both factor tion of the rst factor, we have a calcu-
lated value that produces the maximum would be most appropriate for your proj-
association between all the variables. The ect.
second and succeeding factors produce The advanced techniques presented in
the next best t and depends on whether this chapter represent the main ones likely
the researcher species an orthogonal or to be required by beginning researchers.
an oblique rotation. Particular research questions and the lev-
els of measurement achieved in the vari-
ables determine the appropriate tech-
E. THE FOUR TECHNIQUES nique that should be used. Note that most
COMPARED of these techniques rely primarily on vari-
ables measured at the ratio level. The ma-
Table 18.9 summarizes some of the char- jor exception is that in discriminant func-
acteristics of each of the techniques con- tion analysis, the dependent variable is
sidered in this chapter. assumed to be a nominal variable.
Table 18.10 identies the SPSS tech- Each technique has its unique strengths,
niques and their associated levels of mea- limitations, and appropriate usages. When
surement. This table should be consulted in doubt consult an experienced re-
when deciding which form of analysis searcher or a statistician.
Table 18.9 Characteristics of the Four Multivariate Techniques

Measurement or
Technique or Distribution
SPSS Procedure Assumptions Major Strength Major Weakness
Multiple Regression Ratio level on all Dealing with multiple Instability of weight-
REGRESSION variables; use dummy variables simultane- ings from study to
variables for nominal ously, predictive study if independent
independent variables; equations variables share
assumes normal much common
distribution of all variance
variables
Discriminant Function Nominal dependent; Does regression-type Similar to regression,
Analysis ratio independent analysis using a nomi- an instability of
DISCRIMINANT variables; similar to nal dependent variable; weightings assigned
regression, can use provides a measure of if the independent
dummy independent predictive accuracy of variables share
variables; normal model much common
distribution of inde- variance
pendent variables;
population covariance
matrices must be equal
Multivariate Analysis Ratio level dependent Deals well with interac- Deals well with inter-
of Variance variable; handles treat- tions among variables actions among
MANOVA ment and control variables
variables measured at
all measurement levels
Factor Analysis Ratio level variables; Identifying factors with Labeling of factors is
FACTOR however, Likert a common variability subjective
measures commonly
used
Table 18.10 SPSS Procedures for Multivariate Analysis

INDEPENDENT VARIABLE
DEPENDENT Nominal Ordinal Ratio

DISCRIMINANT
SPEARMAN SPEARMAN
DISCRIMINANT
Ratio MEANS MEANS CORRELATION GRAPH
ANOVA ANOVA PARTIAL CORR
t TEST t TEST MANOVA
REGRESSION FACTOR
E X E R C I S E S
1. Suppose you wish to predict exercise 2. Suppose you are examining variations
behavior in a group of sedentary adult in serum cholesterol levels using so-
women. You have nonexperimental cioeconomic status and region of coun-
data on a number of psychosocial and try as independent variables. Also sup-
biometric variables such as age, moti- pose that you suspect that there is an
vation levels, attitude toward exercise, interaction between region of country
belief in ability to successfully adhere and socioeconomic status. What fac-
to an exercise regimen, body weight, tors would you need to take into ac-
and percent of body fat. Of the meth- count in deciding which method of
ods outlined in this chapter, which one analysis you would use? How might you
would seem to be most appropriate for deal with the suspected interaction be-
this problem? Provide the rationale for tween the two independent variables?
your choice.
Ferketich, S., and Muller, M. (1990). Factor discussion of the techniques included in
analysis revisited. Nursing Research, 39(1), this chapter.
5962. Munro, B., and Page, E. (2000). Statistical Meth-
Iverson, G.R., and Norpoth, H. (1976). Analysis ods for Health Care Research (2nd ed).
of Variance. Beverly Hills: Sage. This is a Philadelphia: J.B. Lippincott. This is a good
ne introduction to analysis of variance review of the statistical techniques most
emphasizing the link between regression commonly used in health research. It con-
and analysis of variance techniques. tains clear explanations of computer print-
Kachigan, S.K. (1986). Statistical Analysis: An outs of various statistical procedures.
Interdisciplinary Introduction to Univariate Neter, J., Wasserman, W., and Kutner, M. (1985).
and Multivariate Methods. New York: Ra- Applied Linear Statistical Models. Home-
dius Press. This text includes a detailed wood, IL: Irwin. This is a more advanced pre-
sentation that is, however, excellent in ex- Guide. Chicago: SPSS Inc. This manual cov-
plaining the principles of regression and ers all the basic procedures plus discrimi-
analysis of variance techniques. nant function analysis, factor analysis, reli-
Polit, D. (1996). Data Analysis and Statistics for ability, regression and analysis of variance.
Nursing Research. Stamford, CT: Appleton SPSS Inc. (1999). SPSS Base 10.0 Syntax Refer-
& Lange. This practical statistics textbook ence Guide. Chicago: SPSS Inc. This manual
for nursing students focuses on how to un- provides the syntax for all SPSS 10.0 proce-
derstand and interpret statistics rather dures.
than how to calculate them. SPSS Inc. (1999). SPSS Base 10.0 Users Guide.
SPSS Inc. (1999). SPSS Base 10.0 Advanced Chicago: SPSS Inc. This manual provides
Models. Chicago: SPSS Inc. This manual in- the basic instructions for using SPSS 10.0
cludes details on using MANOVA as well and includes many of the same procedures
advanced regression techniques. presented in the SPSS Base 10.0 Application
SPSS Inc. (1999). SPSS Base 10.0 Applications Guide.

Part 7
REPORTING,
CRITIQUING,
AND USING
RESEARCH
FINDINGS IN
NURSING
T
he three chapters in this section deal with aspects of pre-
senting, critically evaluating, and using research ndings in
practice. Chapter 19 presents guidelines for organizing and
presenting quantitative and qualitative research reports,
Chapter 20 discusses the process of critiquing both quantitative and
qualitative research methods, and Chapter 21 concludes the text with
a discussion of what research means for nursing. You will explore
how to use research ndings to improve practice and create a re-
search-mindedness in the profession. You may recall that in Chapter
1 we encouraged you to take a walk on the wild side of nursing re-
search. We now challenge you to apply some of the things that you
learned along that walk to improve care, excite the profession, and be
agents of change in the new millennium.
587

Chapter 19
The Research Report
CHAPTER OUTLINE
A. General Orientation 3. Hypotheses and Research Questions

1. Audience 4. Methods
2. Style 5. Presenting Results
3. Type of Report 6. Discussion
4. Avoiding Plagiarism 7. Conclusion
5. Avoiding Sexist Language 8. Final Checklist
B. Organization of Quantitative Reports C. Organization of Qualitative Reports
1. Introduction 1. Guidelines for Qualitative Reports
2. Review of the Literature 2. Sample Outlines of Qualitative Reports
KEY TERMS
Audience Implications Review of literature

Clinical significance Nonparallel construction Research report
External validity Plagiarism Sexist language
Generalizability Pronoun problem Summary tables
589
590 REPORTING, CRITIQUING, AND USING RESEARCH FINDINGS IN NURSING
H
owever well designed, well exe- us consider how these factors inuence
cuted, and brilliant a research pro- report writing.
ject is, its impact depends, above
all, on the quality of the written report. 1. Audience
Unless the researcher communicates
clearly and disseminates results widely, Reports are written for a variety of audi-
the research effort will yield little in terms ences, and the particular characteristics
of knowledge development or improved of your audience should be taken into
practice. This chapter presents some gen- account when you are preparing one. Au-
eral guidelines for communicating results diences may include staff nurses, other
to the research community and users of researchers, policy makers, legislators,
nursing knowledge. Suggestions for the professors, colleagues in an academic
organization and presentation of both eld, health-care administrators, allied
quantitative and qualitative research re- health professionals, or the general pub-
ports are provided. lic. Your audience will help to determine
what content to include, the level and
complexity of ideas to present, and the
A. GENERAL ORIENTATION areas to be emphasized.
If a report is intended for a professional
Writing the research report is the nal journal, then it should be organized in a
stage in the research process. No project manner similar to material found in the
is complete until the research report is journal to which the report is to be sub-
submitted. The research report is usually mitted. Often journals provide guidelines
a comprehensive description of key as- for the submission of manuscripts. In
pects of the project and includes a mini- cases in which they are not included, you
mum of four major sections: the introduc- can contact the journal editor to obtain
tion, methods, results, and discussion. guidelines and inquire about their interest
Some sections may contain subsections in the manuscript topic. Use a model arti-
depending on the nature of the project cle from the journal as a helpful guide for
and source of the publication. The most preparing your report. If the journal is a
popular types of reports include disserta- popular one, then the report should
tions or theses, reports on class-assigned avoid the use of technical terminology. For
research projects, reports to funding example, if you were writing an article on
agencies, evaluation research reports, pa- parenting styles for a professional journal,
pers presented at scientic or profes- it would be appropriate to use such terms
sional conferences, and articles in reas authoritative and laissez-faire; how-
search or professional journals. Reports ever, in a popular magazine article, such
may range in length from 15 pages for a terms would have to be explained.
journal publication to 300 pages for a doc- There are different kinds of journals
toral dissertation. In preparing a report, some are thematic, others are specialty
consider the specic audience that will re- journals, and still others are broad in
ceive the report, the intended effects you scope and publish a wide range of papers.
hope the reader will get from the report, It is important to recognize publishing op-
and the differentiated uses that will be portunities that each may present for the
made of the report. Although there is con- manuscript you are preparing. For exam-
siderable consistency in report structure, ple, thematic journals, such as Advances
variations reect intended audience, style in Nursing Science, often indicate in ad-
of writing, type of research project, and vance a schedule of topics that will be pre-
mode of publication or dissemination. Let sented in forthcoming issues of the jour-
THE RESEARCH REPORT 591
nal. Specialty journals, such as JOGNN and Do not assume specialized knowledge on
Health Care for Women International, pub- the part of your reader.
lish manuscripts that relate to womens
health. Research journals such as Nursing 2. Style
Research publish on a broad range of top-
ics. Finally, some journals are more eclec- Edit your material carefully. Read it
tic, such as the Journal of Professional slowly, perhaps out loud, and eliminate
Nursing, and regularly publish sections on redundant words, sentences, and para-
research, policy, and educational issues graphs. Editing should shorten the docu-
related to nursing. ment considerably. Do not sacrice read-
Most often, it is best to write for the ability and brevity for the sake of saving
general audience, to assume the audi- a nice turn of phrase. Keep it short.
ence has no prior knowledge of the pro- It is a good idea to provide headings
ject, and to convey ideas clearly and sim- and subheadings to help guide your read-
ply. Students who are submitting a paper ers through the material. When technical
for a course requirement are well advised details that would detract from the ow of
to write not for the professor, but rather the main text are nonetheless required,
in a manner that any intelligent person footnotes should be used. For purposes of
would be able to follow. One hint is to editing, it is easier to place footnotes at
write for your Aunt Martha or your Un- the end of the paper.
cle John, not for your professor. Your Tables should be numbered and titled
aunt and uncle have no knowledge of for easy reference. They should include
your research project, are not nurse sci- sufcient information to permit the reader
entists, and have only a high school edu- to read tables rather than text. Text should
cation, but they are very smart. If you ex- compliment important aspects of the table
plain things clearly, they will understand and not present redundant information.
your project. A side benet of this is that Generally, it is preferable to locate tables
your professor will also be able to gure and gures on separate sheets and place
it out! them on the page after the rst reference
Why would writing for an aunt or uncle to them in the text. In manuscript prepara-
be helpful? There is a tendency when tion, tables are also placed on separate
preparing a report to use too many com- sheets so that the editor may move the ma-
puter terms, to use technical jargon, and terial to the nearest convenient spot in the
to fail to explain either the logic behind text. Because tables take extensive re-
your research design or the logic behind sources to typeset, it is best to limit them
the inferences you have made from your to a maximum of four or ve tables per
data. If you write for Aunt Martha, you will manuscript, if considering a journal publi-
be less likely to fall into some of these cation.
traps. And, in the process, you will proba- Style is always inuenced by the publi-
bly write a better report, whether it is in- cation source. Most editors impose a spe-
tended for a journal, a term paper, an in- cic style that has been adopted by their
ternational research conference, or for journal. For example, the American Psy-
the president and CEO of your health chological Association Manual (1998) is
agency. widely adopted by nursing journals such
If you know your audience, you will as Nursing Research, Applied Nursing Re-
also know what questions will come to search, Research in Nursing and Health,
their minds and you will be able to ad- and the Canadian Journal of Nursing Re-
dress issues of concern to them. Above search. Authors need to consult the editor
all, explain your points clearly and fully. or the journals guidelines for authors to
determine what specic style a journal form to the following classic format or
prefers. some rendition of it:
Problem statement
3. Type of Report Theoretical framework
The classic criteria of clarity, concise-
Research question or hypotheses
ness, simplicity, and ease apply to all writ-
Methodology
ing, whether it is qualitative or quantita-
Results
tive research one is reporting. Differences
Discussion
exist in style depending on the type of re- Qualitative reports use a format that
search. Quantitative reports traditionally varies depending on the specic research
use objective stances rather than subjec- tradition. That is, the structure for re-
tive or emotionally laden statements, and porting an ethnographic study varies
impersonal pronouns rather than per- from a grounded theory report or a phe-
sonal pronouns such as I, me, or we. nomenological study. Each qualitative re-
Colloquialisms are avoided. Quantitative searcher must craft a structure that is ap-
researchers attempt to convey a dispas- propriate to the particular study. The
sionate accounting of exactly what was literature is increasingly producing more
studied and what was discovered. and more diverse ways of reporting such
Qualitative researchers, in contrast, use studies. A variety of choices are available
a variety of narrative forms and techniques to the researcher. Box 19.1 illustrates
to engage the reader. The researcher is some alternatives available for the orga-
present in the writing of qualitative reports nization of a qualitative study.
and uses specic strategies to disclose his
or her own biases, values, and the context
that have shaped the narrative. This may 4. Avoiding Plagiarism
include an epilogue, reective footnotes, in-
In writing papers, you must scrupulously
terpretive commentaries, or a section on
avoid plagiarism. Plagiarism is the unac-
the role of the researcher (Creswell, 1998).
knowledged borrowing of other authors
The researcher attempts to give the avor
ideas or words. Most academic disciplines
of the lived experiences being studied
have now adopted the method of referenc-
and to provide enlightening examples that
ing used by the American Psychological
illuminate the phenomena explored. Quo-
Association. This method requires you to
tations are frequently used to bring in the
identify the source from which the mater-
voice of participants and convey complex
ial has been taken in the body of the text
meanings and understandings. Such de-
and to include the complete bibliographic
vices result in a lively and interesting
information in the list of references at the
report. The writing style is usually per-
end of the paper. This is the referencing
sonal, familiar, up-close, highly readable,
method used throughout this text.
friendly, and applied for a broad audience
(Creswell, 1998). If the writing style is effec-
tive, it should transport the reader into the a. Short Quotations
world of the study. A careful description of
the settings, people, events, and the inter- When fewer than 40 words are being di-
actions among them is one of the main con- rectly quoted, the material should be en-
tributions of qualitative research (Streu- closed in quotation marks. After the quo-
bert and Carpenter, 1999). tation, reference should be made to the
The type of research report inuences name of the author, year of publication,
the structure. Quantitative reports con- and the page number. For example: The
BOX 19.1 Alternative Structures for Qualitative Reports
Emami, Torres, Lipson and Ekman (2000): A cluding a subheading of Limitations); and, fi-
four-part ethnographic report consisting of an nally, the Conclusion section
Introduction, Methodology (including Data Gates (2000): A seven-part phenomenolog-
Collection, Sample, and Analysis), Find- ical report composed of an Introduction/Con-
ings, and Discussion. text, Researchers Perspective, Methodol-
Orne, Fishman, Manka, and Pagnozzi (2000): ogy (including Participants, Data Gathering,
A four-part phenomenological report com- and Data Analysis), Findings, Discussion,
posed of an Introduction, Method (includ- Research and Practice Implications, and Con-
ing Study Participants, Procedures, and clusion.
Thematic Analysis), Findings (including four Dickson (2000): A participatory action re-
separate theme clusters), and Discussion search report composed of seven sections be-
Gray and Smith (1999): A five-part ginning with an introduction that is untitled; fol-
grounded theory report starting with the - lowed by a section entitled Grandmothers
Introduction; followed by The Study, that describes the context in which the study
which is subdivided into a heading called Ma- occurred; a section called The Grandmothers
terials and Methods (containing Design, Project; followed by the Methodology sec-
Sample and Data Collection, Ethical Is- tion (Data Collection, Data Analysis, and
sues, and Data Analysis); a Findings sec- Trustworthiness), the Findings and Discus-
tion that describes the five phases that sion of Effects of Participation section, a Study
emerged from the core category of profes- Strengths and Limitations section, and con-
sional socialization; a Discussion section (including with an Implications section
difference between mores and folkways rectly, you are nonetheless required to
lies in the nature of the reaction the viola- cite your source. Once again, you should
tion of the norm produces, and not in the include the authors name and the year of
content of the rule (Jackson, 1999, p. 19). publication but not the page number.
b. Long Quotations d. Citing Secondary Sources
When more than 40 words are directly Occasionally you nd a study that is dis-
quoted, the material is indented an addi- cussed in a journal article or book (a sec-
tional ve spaces on the right side of the ondary source) but you are unable to lo-
page and the material is single-spaced, cate the original (primary source) in which
with a reference to the authors name, the study was rst published. When a sec-
date, and page at the end of the quotation ondary source is used, cite the original au-
(unless the author has been introduced thor rst, followed by a reference to the
earlier in the paragraph). For example, if secondary source. For example, Gillis,
the quote is introduced, Gillis and Jackson cited in Jackson (1999). In this case, Gillis
(2002) noted . . . then only the page num- is the primary source and Jackson is the
ber is needed at the end of the quotation. secondary one.
c. Paraphrased Material e. Multiple Authors References
When using paraphrased material by bor- When there are three, four, or ve au-
rowing ideas that you are not quoting di- thors, the rst reference includes all the
names, and subsequent references to their and nurses, secretaries, and parents as
work use only the rst authors name, fol- she. There are a number of solutions to
lowed by et al. When there are six or the pronoun problem. To illustrate, sup-
more authors, use only the rst authors pose we have the following sentence: A
name followed by et al. and the year for doctor has to be especially careful; other-
the rst and subsequent citations. wise, he can be sued for malpractice.
(The problem is characterizing the doctor
as a he.) This sentence can be changed in
f. Reference List any one of the following ways:
At the end of the paper, provide a refer- A doctor has to be especially careful;
ence list that includes all references cited otherwise, he or she can be sued for
in alphabetical order. Notice how books malpractice. (This is a somewhat
and articles are cited in this text. Addi- awkward solution but is acceptable if
tional examples may be noted in the bib- not used too frequently.)
liography that is included at the end of Doctors have to be especially careful;
this text. otherwise, they can be sued for mal-
practice. (This is a common solution,
Sample article reference: Parsons, K.
converting to the plural form avoids
(1997). The male experience of care
the use of he or she.)
giving for a family member with Alz-
heimers disease. Qualitative Health Re-
If not especially careful, doctors can
be sued for malpractice. (This sen-
search, 7(3), 391408. (The italicized
tence avoids the pronoun all together
material may be underlined if you do
by reconstructing the sentence.)
not have an italic font on your printer.)
Sample book reference: Norwood, S. L.
(2000). Research Strategies for Ad- b. The Man Problem
vanced Practice Nurses. Uppersaddle
River, NJ: Prentice Hall Health. Traditionally, many words and expres-
Sample chapter from an edited book ref- sions in the English language used man or
erence: Israel, B.A., and Schurman, S. men to refer to persons of either gender.
(1990). Social support, control and the In a manner similar to the pronoun prob-
stress process. In Glanz, R., Lewis, F., lem, such usage may unintentionally sug-
and Rimer, B. (Eds.): Health Behavior gest, for example, that a foreman should
and Health Education. San Francisco: be a man. Table 19.1 includes a few ex-
Jossey-Bass. amples and some alternative forms that
could be considered.
5. Avoiding Sexist Language

c. The Nonparallel Construction
In the past few years, there has been an Problem
increasing awareness of sexist language.
Language may frequently put one gender
We will briey look at some of the major
at a disadvantage. For example, a man
pitfalls in gender references.
may be referred to more formally than a
woman, suggesting a power or impor-
a. The Pronoun Problem tance differential. More generally, non-
parallel constructions can result in con-
The pronoun problem encourages stereo- fusion or misrepresentation by violating
typic thinking by referring to such people the principle that parts of a sentence that
as doctors, managers, and patients as he are parallel in meaning should be parallel
Table 19.1 Neutralizing Gender Terms reports associated with each perspective
Traditional Usage Alternative Forms and method (see Box 19.1). This section
Chairman Chair, chairperson, coor- relates primarily to quantitative reports.
dinator, head, leader, The nal section of the chapter presents
moderator, presiding some minimum guidelines for qualitative
ofcer reports with an example of the structure
Clergyman Cleric, member of the of a qualitative research report. An effort
clergy, minister should be made to cover the material dis-
Fisherman Fisher cussed under each of the following head-
Foreman Boss, supervisor
ings, but it should be noted that there are
somewhat different traditions for the re-
Mailman Letter carrier, postal
porting of experimental studies versus
worker
surveys and other descriptive designs.
Mankind Human beings, humanity,
people
Manmade Articial, manufactured, 1. Introduction
synthetic
Manpower Personnel, workers
The introduction should inform your
reader what the project is about, indicate
Salesman Sales agent, salesperson
the general approach that has been used
To man Operate, to staff to solve the problem it tackles, and suggest
Workman Employee, laborer, the critical problems the project raises.
worker Also mention interesting questions and un-
resolved issues that you propose to an-
swer in your research.
Table 19.2 Nonparallel and Parallel Gender
References
2. Review of the Literature
Inappropriate Usage Alternative Forms
(Nonparallel) (Parallel)
The review of the literature tries to pro-
Man and wife Husband and wife vide an overview of the state of scientic
Men and ladies Men and women knowledge in your area of study. Con-
Mens and ladies Mens and womens sider reviewing the theoretical models
teams teams that are appropriate and the empirical
Males and women Males and females
ndings that bear on the particular rela-
tionships you will be examining or, if
such material is not available, give the
reader some sense of the variables that
in structure. Table 19.2 provides some
have been related to the major depen-
illustrations of nonparallel gender refer-
dent variable in your study. The review
ences and more appropriate, parallel
should highlight the areas in which there
forms.
are inconsistencies in the conclusions of
other studies and indicate which of these
inconsistencies you intend to address.
B. ORGANIZATION OF Generally, it is best not to present sum-
QUANTITATIVE REPORTS maries of articles; instead, it is good to fo-
cus on what the consensus is on the rela-
A paper should be organized into sec- tionship between particular variables and
tions. As previously noted, there are a va- the dependent variable. For example, sup-
riety of reporting formats for qualitative pose that you were examining factors re-
lated to nurses job satisfaction. It would symbols. Figure 19.1 presents the dia-
be useful for your reader to know whether gram used by Murray (1999) in a study of
there is any agreement in the scientic unwanted intimacy.
literature on whether satisfaction is re-
lated to such variables as age, type of ed-
ucational program, nursing practice set- 4. Methods
tings, level of decision making, and years
of work experience. When there are in- The methods section details the research
consistencies, do you have any observa- plan that was used to answer the re-
tions as to why they emerged? Different search question or test the hypotheses. It
regions, different measurement or ana- includes a description of what was done
lytic procedures, or systematic variations in sufcient detail so that another re-
in the compositions of the populations searcher could replicate the study. It is
studied might all account for the varia- usually divided into subsections includ-
tions between studies. If inconsistencies ing descriptions of the sample, data col-
are present, they can be noted and you lection procedures, and indexes and mea-
can heighten your readers interest in surement procedures used. The rationale
your project by proposing to answer given for the design selected should state
some of the questions that have been the advantages the chosen design has
raised. over alternative designs. What designs
The review of the literature varies in have typically been used by other inves-
length and style depending on whether it tigators looking at similar relationships?
is being prepared for a thesis or disserta-
tion, a conference presentation, a schol-
a. Description of the Sample
arly paper, or a journal publication. Jour-
nal editors often limit the length of the Readers should be introduced to the re-
review because of cost and space limita- sults by reporting some of the back-
tions. Chapter 3 contains a discussion of ground characteristics of the individuals
how a review of literature can be devel- involved in the study. Gender distribu-
oped and presented. Readers are re- tion, rural or urban location, and average
ferred to that chapter for additional sug- age all may be reported if judged to be rel-
gestions. evant by the researcher. If efforts are to
be made to ensure the representative-
ness of the sample, this is the stage in the
3. Hypotheses and Research report that such material would be intro-
Questions duced appropriately. If you know from
census material, for example, that 54 per-
The review of literature section should cent of the regions population above age
lead into a section dening the hypothe- 15 years is female and your sample is 59
ses, questions, or relationships that are percent female, this fact should be noted.
to be examined. These should be pre- In experimental studies, a detailed de-
cisely stated and connected to the litera- scription of the method of randomization
ture of the discipline. It is almost always used should be included.
best to diagram the research models that
are being evaluated in the research: not
only does this permit the presentation of b. Data Collection Procedures
hypotheses clearly but additional preci-
sion is achieved by drawing in causal ar- After determining the questions and the
rows and greater than and less than design, the next step in the methods sec-
Background:
< socio-economic status

> community size
> broken home < self-esteem
< education
Unwanted
Intimacy
Situational: Views > feministic

> alcohol use
> male pays for date
> conductive setting for intimacy
-------------------------------------------------------------
Source of Spuriousness:
> Unwanted Intimacy > Responsibilty
< Self-esteem
-------------------------------------------------------------
Figure 19.1 Causal Model for Unwanted Intimacy Study. Adapted from Murray (1999). Unwanted
intimacy in female university students. Research methods paper, Antigonish, Nova Scotia, Canada: St.
Francis Xavier University.
tion is to describe the measurement, sam- study, the procedure for development
ple design, and data collection proce- and testing of the quality of the instru-
dures. A description of the methods used ment should be described. Also indicate
to collect the data is critical to evaluating what revisions were made and what scor-
the study. A description of the instru- ing procedures were followed. Any com-
ments and procedures and the rationale ments on problems in data collection
for their selection is usually included. In should be mentioned at this point in the
describing a questionnaire, it is not nec- report.
essary to discuss every question; how-
ever, sufcient detail should be provided c. Description of Indexes and
so that it is evident how the variables Measurement Procedures
were operationalized. It is helpful to focus
attention on nonstandard items and, if A description and evaluation of indexes
possible, include a copy of the question- or scales constructed and a preliminary
naire in an appendix to the report. If the report on the mean results should be
instrument was developed for use in the made at this point. If you have used an in-
dex that was administered in a previously cents who took peer pressure resistance
published study, comparisons may be classes engaged in healthier lifestyles
made between the mean results of your than those who did not.
study and those of other researchers. De- Clinical signicance of ndings should
tails for the measurement of other key also be discussed. Findings may or may
variables should also be included. If the not have statistical signicance (ndings
study involves repeated measures, what are unlikely to be caused by chance) and
steps were taken to ensure that the mea- yet still have importance for nursing prac-
surement process itself was not contami- tice. If the difference is great enough to
nating the results? make an impact on client care or on a phe-
nomenon of concern to nursing, then it is
clinically signicant, even if analysis pro-
5. Presenting Results cedures do not produce statistically signi-
cant results. For example, the use of a cer-
The results section is usually a combina- tain sedative in combination with an
tion of narrative description and statisti- analgesic consistently reduces the percep-
cal reports. If descriptive and inferential tion of pain in a variety of postoperative
statistics have been used, the descriptive conditions in comparison with pain per-
statistics are reported rst. When report- ception levels for those only receiving the
ing statistical ndings, three pieces of in- analgesic. Although the differences be-
formation are required: the test value, the tween the two groups may not be statisti-
level of signicance, and the degrees of cally signicant, the trend consistently fa-
freedom. For example, it may be reported vors the use of the sedative in combination
that depression was experienced at 1 with the analgesic; therefore, the results
month postpartum in new mothers and have clinical signicance because the
signicantly lower levels of depression treatment leads to reduced pain levels. In
were reported in 3-month follow-up mea- small sample studies (those with fewer
sures. The following statement reects the than 200 subjects), it is particularly impor-
three pieces of statistical information to tant to report on whether the data trend is
communicate this: The ANOVA for rein the direction predicted by the various
peated measures indicated a statistically research hypotheses.
signicant reduction in depression by Generally, the Statistical Package for
time (F 12.12; df 2; P 0.001). the Social Sciences (SPSS) output is not in
The narrative presentation of results an appropriate form for presentation. Ta-
should be written in the past tense. This is bles placed in a nal report need to be re-
because research studies do not prove or formatted. Their format should conform
conrm that hypotheses are correct; to those shown throughout this book or
rather, hypotheses are supported or not be modeled on those presented in journal
supported by the ndings. It is inappro- articles or books. Although it is possible
priate to write: Adolescents who take to control the SPSS output format, begin-
peer pressure resistance classes engage ning researchers will probably nd it eas-
in healthier lifestyles than those who do ier to retype the tables.
not. The use of the present tense implies In qualitative reports, the results sec-
the results apply to all adolescents rather tion often contains a summary of themes
than just the sample whose behavior was and narrative descriptions of phenom-
observed in the past (Thomas, 2000). The ena. Extensive quotations may be used to
verbs take and engage should be substantiate the major themes identied.
changed to took and engaged. The re- Models, gures, and diagrams are useful
sults should read as follows: Adoles- in concisely illustrating complex ideas,
theories, and relationships that emerge plored when reporting ndings. If you are
from qualitative investigations. using tests of signicance, report whether
the ndings are statistically signicant. If a
particular relationship is not statistically
a. Organizing Summary Tables signicant, is there a trend in the data? It
is misleading simply to report signicant
The variations in the dependent variable
ndings. As noted earlier, relationships
should be explored at this point. Any ba-
that are not statistically signicant may
sic analyses that explore the formal hy-
well have substantive signicance.
potheses of the study should be pre-
Let us look at some formats for report-
sented. A challenge for the researcher is to
ing single-variable information, cross-
compress this information into as few ta-
tabulation tables, differences of means,
bles as possible. Well-designed summary
and correlations into summary tables.
tables can add a lot to your report without
sacricing anything. When possible, use ( i ) Summarizing Univariate
summary tables to compress the results of Statistics
many analyses into one table. Remember Table 19.3 presents one way of compress-
to focus on the relationships being ex- ing the information on a number of vari-
Table 19.3 Summarizing the Sample Characteristics

A. Nominal or Ordinal Variables Number Percentage
Gender
Male 82 48.0
Female 89 52.0
Status
Student 26 15.2
Retired 2 1.2
Unemployed, looking for work 13 7.6
Unemployed, not looking for work 4 2.3
Employed part time 19 11.1
Employed full time 107 62.6
Population of home community
Under 5000 17 9.0
5000 to 19,999 24 12.7
20,000 to 99,999 16 8.5
100,000 to 999,999 108 57.1
1,000,000 or more 24 12.7
B. Ratio Variables Mean Standard Deviation Cases, n
Age, years 29.6 14.7 183

Income, dollars 43,257 16,419 77
Seniority 8.87 3.76 104
Children, n 1.37 1.06 78
ables into a single table. Note that nominal itative studies are often conceptual or
variables can simply have the frequencies thematic. Table 19.7 presents ve the-
listed for each category. These kinds of matic areas that represent changes found
listings are useful in summarizing descrip- in a group of native grandmothers who
tive characteristics of the cohort studied. participated in a health promotion pro-
The mean values and the standard devia- ject. Figure 19.2 visually presents some
tions can be used to summarize the ratio ndings from a grounded theory study
variables. designed to build a theory about individ-
uals perception of being psychiatric pa-
( ii ) Summarizing Cross-tabulation
tients. The gure highlights the patients
Tables
experiences before hospitalization that
Table 19.4 shows one method for report-
were shaping their responses, including
ing a series of cross-tabulation table re-
medication noncompliance, lack of social
sults (CROSSTABS procedure in SPSS).
capitol, and substance abuse.
Note that only the percentage of smokers
is reported, along with the number of
cases in the column, the chi-square prob- b. Using Graphs and Charts
ability, and whether the results have a
trend in the predicted direction () or in When feasible, it is a good idea to present
the opposite direction (). data using graphs and charts to make a
greater visual impact on readers. Figures
( iii ) Summarizing Mean Values for
19.3 and 19.4 show some alternative forms
a Dependent Variable
of reporting information. Figure 19.3 uses
Table 19.5 provides an illustration of how
a bar graph to show the relationship be-
you can compress a number of analyses
tween country of origin and having suici-
of ratio variables into one table. The
dal thoughts. Respondents countries of
analysis by Annette Fougere compares
origin are grouped into three categories:
the academic performance of grade 12
Canada, New Zealand, and Australia.
students who did or did not regularly eat
Figure 19.4 uses a plot to show the re-
breakfast. Note that Table 19.5 includes
lationship between prestige and income.
the mean grade performance, standard
The visual information conveys much
deviations, number of cases, probability
about the strength of the association be-
level, and an indication of whether the
tween the two variables. To seasoned re-
data have a trend in the predicted direc-
searchers, the correlation of 0.83 between
tion () or in the opposite one ().
the two variables reveals much, but for
( iv) Summarizing Correlations most readers, the plot of the information
Table 19.6 shows a correlation matrix for reveals additional information. Notice, for
variables related to rst-year university example, that the prestige ratings tend
performance. Note that a report on many not to increase much as incomes rise
relationships can be compressed into above the $90,000 mark.
such a table. By using asterisks, it is also
possible to indicate which of the correla-
tions are statistically signicant. c. Evaluating Hypotheses and Models
(v) Tables and Figures in The reader should now be well prepared,
Qualitative Reports and anticipating the results of the hy-
Table 19.7 and Figure 19.2 show examples pothesis and model testing. If diagrams
of the type of word items that are fre- described the original relationships, they
quently presented in qualitative research should be used again when the ndings
reports to portray results. Tables in qual- are reported. If you are testing a theory
Table 19.4 Summarizing Cross-tabulation Tables: Respondents Smoking Behavior by Selected

Independent Variables*
Cases, n Probability
Independent Variables Smokers, % Column Total (Chi-Square) Trend
Residence
On campus 19.8 101 0.93283 ?
Off campus 20.4 54
SES of father
Low SES 22.2 54 0.61304
Mid to high SES 18.8 101
Respondents age, y
19 or younger 19.7 76 0.93597
20 or older 20.3 79
Level of self-esteem
Low 19.5 77 0.87237
High 20.5 78
Level of stress
Low 19.4 67 0.90052
High 18.6 86
Home community, population
Under 30,000 17.8 118 0.22068
30,000 or more 27.0 37
Gender
Male 17.6 68 0.51734
Female 21.8 87
Type of program
Arts 23.3 60 0.48859
Science 18.7 91
Exercise/week
Less than three times 29.6 71 0.00613
Three or more times 11.9 84
Father smokes
Yes 29.3 41 0.8363
No 16.7 114
Mother smokes
Yes 45.5 33 0.00004
No 13.1 122
Source: Adapted from Michelle Lee (1992). Smoking Behaviors. Antigonish: St. Francis Xavier University, Research
Methods Paper.
Table 19.5 Summarizing Means: Grade Performance of Grade 12 Students

Standard Test of
Independent Variables Mean Deviation Cases, n Significance Trend
Eats breakfast
No 68.9 12.7 46 0.0003
Yes 76.3 10.0 95
Breakfast eaten in past 7 days
No 69.5 9.9 39 0.0049
Yes 75.4 11.5 106
Breakfast maker
Others 72.7 14.4 42 0.2731
Self 75.1 10.0 85
Lunch maker
Others 73.8 11.7 105 0.8412 *
Self 74.3 11.6 29
Supper maker
Others 74.1 11.6 117 0.6416 *
Self 72.9 10.6 27
Gender
Male 71.4 11.7 85 0.0026
Female 77.2 10.0 58
Community population
5000 73.7 11.9 103 0.8024
5000 74.2 10.2 42
Career plans
University 78.2 9.0 92 0.0000
Nonuniversity 66.2 11.1 53
Extracurricular activities
No 70.9 12.0 70 0.0015
Yes 76.8 9.9 74
After-school job
No 73.4 11.8 88 0.5825
Yes 74.5 10.8 57
Trend predicted correctly; trend predicted incorrectly; * trend not predicted.

Source: Adapted from Annette Fougere (1992). Effects of Eating Breakfast on Grade Performance. Antigonish: St. Francis
Xavier University, Research Methods Paper.
and have derived a hypothesis, then you In cases in which you are investigating
should report the nding of the test, even alternative explanations for a relation-
if the relationship is not statistically sig- ship you have initially assumed to be sta-
nicant. tistically signicant, you would only con-
Table 19.6 Correlations Between First Year University Average, Average High School Grade, and
English High School Grade (n 3617)
First Year Average High English High
Correlations University Average School Grade School Grade
First year average 1.000
Average high school grade 0.573* 1.000
English high school grade 0.464* 0.662* 1.000
*P 0.001
tinue to test the alternative explanations will not be explored because the original
if the relationship turned out to be statis- relationship was not sufciently strong.
tically signicant. You would, of course, If it turns out that the primary rela-
include the intervening variable model in tionship is not statistically signicant, it
your report; however, if no relationship is then appropriate to explore the rela-
has emerged, you would not proceed tionship of independent variables to the
with an evaluation of the alternative ex- dependent variable. If you are forced to
planationsyou have no relationship to this fall back position and have not es-
explain. All you would do is note in your tablished the various hypotheses in ad-
report that the alternative explanations vance, then you should indicate that your

Figure 19.2 Basic conditions, intervening factors, and individual responses to psychiatric hospital
admission.
explorations are being conducted with- research t into the general picture. Such
out the guidance of hypotheses. The references also help tie the paper together,
reader is then alerted to the fact that you reminding the reader of the problems that
are on a hunting expedition. Hunting is the project initially raised. In what areas
ne as long as the reader is alerted. does your research support the general
view, and in what areas does it not? When
6. Discussion there are discrepancies, what are some of
the possible explanations? This discussion
At this point, an effort should be made to should provide readers with a sense of
tie together the whole project. The discus- what has been learned and what remains
sion section accomplishes this by focusing problematic.
on the interpretation of the results; the lim- The generalizability of the ndings to
itations of the study; and the implications other contexts should be explored. This
of the ndings for practice, education, the- refers to the extent that the ndings may
ory, and future research. Again, references be extrapolated from the present study
should be made to the review of literature to other groups in general. It is com-
section, showing how the results of your monly referred to as external validity.
40
30
20
10
Percent
0
CANADA NEW ZEALAND AUSTRALIA
NAME OF COUNTRY
Figure 19.3 Sample bar graph. Percent with suicidal thoughts by country of origin.
Figure 19.4 Sample scatter plot.
The impact of threats to external and in- tative terms so they may be tested in sub-
ternal validity on results should be dis- sequent investigations.
cussed in this section (see Chapters 4
and 12). In so doing, the researcher com- 7. Conclusion
municates to readers an awareness of
the limitations of the study and dis- The nal section should try to briey
cusses their impact on the interpretation state what the central problem was and
of the study ndings. what conclusions have been identied.
The implications of the study ndings This section may also include sugges-
are explored in the discussion section. tions for how the current project may
The author highlights the importance of have been improved and what other is-
the study in terms of its potential contri- sues the researcher identies as worthy
butions to practice, education, theory, of further exploration.
and research, as appropriate. This com-
ponent is particularly critical if the re- 8. Final Checklist
sults are to be used by research con-
sumers to inform practice or to identify Table 19.8 contains a checklist of items
new and challenging research questions. that authors of quantitative reports may
Recommendations emerging from the wish to use to provide some guidance for
study should be clearly articulated in ten- when a paper is ready to submit.
Table 19.8 Checklist for Determining When Your Paper Is Ready for Submission
Section Description
Title Short, informative title should be selected
Abstract If needed, keep to one half a page
Introduction Excite interest of reader; indicate why the topic is of interest. Normally
the Introduction is short; keep literature references to a minimum.
Review of Literature Include summary table if possible; it is generally best to organize the
review around ndings by variable not article summaries; should
anticipate hypotheses (diagrams) to follow.
Statement of Hypotheses Include causal model diagrams or formal statements of hypotheses to
be tested. Diagrams force precision.
Methodology Description of variables, measurement procedures, sampling, and
analytical procedures identied
Results Provide a description of sample studied; include summary tables for
major hypotheses tested
Discussion Link ndings of study to the body of literature reviewed earlier. What
questions remain? Which direction do you see as the most productive
for researchers to pursue?
Conclusions Summarize the major ndings of the study. Keep this section short.
References Are in-text references consistent? Are all in-text references included in
the References Cited section at the end of the paper? Have you
consistently used the APA referencing format?
Spell Check Run the paper through a spell checker to catch any errors.
Format Examine the paper to be certain you are not breaking tables in the
middle; use different sized fonts for subtitles.
View Document Use word processors view document to check on page layout.
Print Print your paper!
Proofread A nal check before you submit it.
C. ORGANIZATION OF come alive for the reader. The researcher

QUALITATIVE REPORTS creates scenes or vivid descriptive ac-
counts of his or her experience over time
Although some content discussed in the with the participants rather than report-
previous section is appropriate for some ing facts and gures. Each tradition of
qualitative investigations, it would be un- qualitative research elects to do this using
wise to expect qualitative reports to con- a different narrative structure. In prepar-
form to the established format of a con- ing to write a qualitative report, one
ventional quantitative study. Qualitative should obtain a high-quality model of a re-
research emanates from a different world- port using that particular tradition. This
view (see Chapter 1), and reporting for- proves extremely helpful in guiding the re-
mats must respect this difference. Quali- searcher with the nal report. One can
tative researchers are concerned with learn a great deal about the proper struc-
capturing the essence of phenomena of ture and style for writing qualitative re-
concern to nurses and communicating the ports by a careful study of relevant phe-
meaning of human experiences, with a nomenological, ethnographic, grounded
level of detail that makes the experience theory, or historical studies.
1. Guidelines for Qualitative with numbers, the former, coming from a

Reports different paradigm, reports themes or
processes. Editors are becoming more
In searching for models of qualitative stud- sensitive to the differences in report
ies, you will notice that there is no short- styles and are less likely to force a quali-
age of styles or formats in the literature. tative investigator to conform to the re-
From this rich diversity of approaches, porting style adopted by quantitative re-
Miles and Huberman (1994, p. 304) have searchers.
identied a set of minimum guidelines to
assist beginning researchers in drafting a
qualitative report. These include: 2. Sample Outlines of Qualitative
1. The report should tell us the essence Reports
of the study.
Parsons (1997) used the phenomenologi-
2. It should describe the social and his-
cal method to investigate the male expe-
torical context of the setting (or set-
rience of caregiving for a family member
tings) where data were collected.
with Alzheimers disease. She interviewed
3. It should describe the natural history of
eight menve spouses and three sons
the inquiry, including what was done
to determine what their experiences were
and by whom and how. This is similar
like. From the analysis of the interview
to the methods section of a quantitative
data, nine themes emerged. The themes
account. It should convey how key con-
were interrelated to form a whole that
cepts emerged over time, which vari-
captured the experience of caregiving for
ables appeared and disappeared, and
the male caregivers (the essence). Below
which codes led to important insights.
is an outline of the report:
4. It should communicate raw data in the
form of vignettes, quotations, pho- 1. Abstract
tographs, and so on, so that the reader 2. Introduction and statement of topic
can draw conclusions in parallel with and purpose
the researcher. Drawing conclusions 3. Review of literature
unsubstantiated by data is the antithe- 4. Methodology
sis of research. Participants
5. Finally, in the discussion or conclusion Procedure
section, researchers should articulate Data analysis
their conclusions and the impact on 5. Findings
the nursing world. Enduring
Many qualitative investigators agree
Vigilance
that it would be unwise to develop a set
Asense of loss
of canons for reporting qualitative nd-
Aloneness and loneliness
ings. It appears at this juncture in our re-
Taking away
search history that the eld of nursing
Searching to discover
agrees. Most researchers believe the
The need for assistance
studys questions, context, and audience
Reciprocity
should drive the design of the qualitative
Overstepping the normal boundaries
report. A balance between interpretation
The essence
6. Discussion
and description is important.
A major difference in reporting quali- If you consult the original source, you
tative studies compared with quantita- will notice that the terms abstract and in-
tive ones is that whereas the latter deals troduction do not appear in the article. We
have included them in our outline, how- were conducted with 26 family caregivers of
ever, because clearly they are evident in persons with cancer or AIDS during a 4-
the structure of the report. Journal arti- month period. Data were analyzed in terms
cles often omit any heading at the begin- of the strategies, consequences, and inter-
ning of an article and consider all the ma- actions involved in the caregiving experi-
terial before the methodology section as ence. The researchers structured their re-
an introduction. The introduction of this port according to the following outline:
report identies the research topic, the
1. Abstract
statement of purpose, the review of the lit-
2. Introduction
erature, and the signicance of the prob-
3. Theoretical framework
lem for nursing.
4. Purpose
The methods section is divided into
5. Methods
subsections and identies the design
(phenomenology), the sample selection
Procedure
process (participants), and the procedure
Sample
for data collection and analysis. In qualita-
Instrument
tive reports, a great amount of detail is ex-
Data analysis
6. Findings
pected in the data collection and analysis
section because so much of the ndings
Strategies for taking care
depend on the researchers interpretation
Consequences of taking care
of the data. Because of this, it is important
Interactions and taking care
7. Limitations
to describe any procedures taken to pro-
8. Practice recommendations
tect and evaluate the quality of the data.
The results section is called Findings This outline follows a logical step-by-
and is organized according to the nine step approach to the challenging but ex-
themes that emerged and the description citing task of writing up qualitative re-
of the essence of the experience. This is search results. In most qualitative outlines
typical of the style of reporting found in a such as this one, you will notice that the
phenomenological study. Quotations are ndings section, in particular, usually por-
used to illustrate important points. The trays a range of subheadings unique to the
themes are summarized and descriptive specic study. The subheadings reect the
data are integrated into the materials. key ideas or concepts that emerge from
The discussion section discusses the im- the analysis of the qualitative data. This
portance of the ndings to nursing. section quickly provides the reader of the
You will notice that although there are report with a rich description of the criti-
similarities in the overall structure of this cal ndings of the report. For example, in
phenomenological report and some quan- their ndings section, Stetz and Brown use
titative reports, there are signicant differ- the subheadings of Strategies, Conse-
ences in the organizational sections and quences, and Interactions involved in
the content within the sections differs. the caregiving experience. These ndings
To further explore the similarities and vividly portray the family caregivers ex-
differences in quantitative and qualitative periences in caring for ill family members.
reports, lets look at the organizational In other ways, the outlines of quantitative
structure of a grounded theory study. Stetz and qualitative reports are similar because
and Brown (1997) provide an in-depth de- both use the general headings of Title,
scription of Taking Care, one of the phases Abstract, Introduction, Method, Re-
of a grounded theory of caregiving for fami- sults, Discussion, and Conclusion.
lies experiencing life-threatening illnesses It is important to remember that a re-
such as cancer or AIDS. In-depth interviews search project is not complete until the -
nal report is written and disseminated to is the researchers professional account-

appropriate audiences. The perfect re- ability to make certain that results are
search project is of little value if the re- published as soon as possible upon com-
sults are not communicated to others. It pletion of the project.
E X E R C I S E S
1. Select a nursing research journal that 3. Read the introductory section of a
publishes both quantitative and quali- nursing research study. Carefully edit
tative studies. Review an issue of it and that section for redundant words,
determine if the quantitative articles phrases, and sentences. Did you signif-
within it adhere to the organizational icantly shorten the section as a result
headings provided in this chapter. of your editing? Did you sacrice clar-
Compare and contrast the organiza- ity for brevity?
tional structure used by the quantita-
tive and qualitative researchers to
write their reports.
4. Visit your nursing library and review
the following types of research publi-
cations:
2. Select one article containing tables or Scientic nursing journal
graphs (or both). Assess them against Professional journal that publishes
the guidelines provided in the chapter. research related articles
Select a second article that does not Nursing student thesis
contain tables and create a table based Popular literature report of a re-
on the data presented in the Results search study
section. Compare and contrast the struc-
ture of each report. Who was the in-
tended audience? Did the author do
a reasonable job of communicating
the study to the intended audience?
What suggestions, if any, would you
make to improve the style of the
presentations?
American Psychological Association (1998). Special sections on health research, health

Publication Manual of the American Psycho- care, and health promotion are included.
logical Association, Washington D.C.: Amer- Creswell, J.W. (1998). Qualitative Inquiry and
ican Psychological Association. Research Design. Thousand Oaks, CA: Sage.
Bell, L. (1995). Effective Writing: A Guide for A good description of the different narra-
Health Professionals. Toronto: Copp Clarke. tive styles used in each qualitative tradi-
An excellent resource for those interested tion.
in writing for publication, either for health- Hodges, J.C., Whitten, M.E., Brown, J., and
care professionals or the general public. Flick, J. (1994). Harbrace College Handbook:
for Canadian Writers (4th ed). Toronto: Har- ments of Style. Galt, Ontario: Brett-MacMil-
court Brace, Canada. This is a comprehen- lan. Originally published by Strunk in 1935,
sive guide to grammar and style, including this short manual is a classic in the art of
a section on research papers. clear writing.
Miles, M.B,. and Huberman, A.M. (1994). Qual- Thomas, S. (2000). How to Write Health Sci-
itative Data Analysis. Thousand Oaks, CA: ences Papers, Dissertations and Theses. New
Sage. This is a helpful discussion of writing York: Harcourt. This is a short practical
qualitative research reports. guide to writing research reports in the
Strunk, W., and White, E.B. (2000). The Ele- health sciences.
Chapter 20
The Research Critique
CHAPTER OUTLINE
A. Understanding the Critiquing Process 2. Criteria for Critically Appraising

1. What Is a Research Critique? Qualitative Research Reports
2. What Is the Role of the Critic? C. Examples of Research Critiques
3. What Are the Stages of the Critique 1. Example of a Critique of a Quantitative
Process? Study
B. Critically Appraising Research Reports 2. Example of a Critique of a Qualitative
1. Criteria for Critically Appraising Study
Quantitative Research Reports
KEY TERMS
Critical appraisal Research critique Research critic

Peer review
611
T
he purpose of this chapter is to pro- should be considered seriously in your
vide beginning researchers with a practice. The outcome of a critique should
set of skills for critically appraising inuence future research and knowledge
research efforts. This is an essential skill development in nursing.
for all nurses who wish to engage in evi-
dence-based practice, use research nd- 2. What Is the Role of the Critic?
ings to contribute to improved client care,
or be responsible consumers of research. The major role of a research critic is to
Unless nurses can correctly assess the provide an objective evaluation of a
merits of a research report, they are at a studys merits. To do this, a critic must
loss in terms of using research to guide read a report comprehensively and offer
practice and will be unable to contribute constructive comments on all aspects of
to extending the eld of nursings research the report, as well as on the specic
knowledge base. This chapter discusses strengths and limitations of the study.
what a research critique is and outlines Comments should be offered in a spirit
the process for evaluating the quality of that promotes mutual respect and colle-
research studies. Guidelines are provided giality among the researchers and the re-
for conducting critiques of quantitative viewer. The intent of a critique is to con-
and qualitative research reports. Samples tribute positively to a research program.
of research report critiques are provided. Although the areas needing improvement
should be communicated clearly by the
critic, tactfulness is needed so that re-
A. UNDERSTANDING THE searchers are not offended or discour-
CRITIQUING PROCESS aged in their attempts to advance nursing
knowledge.
1. What Is a Research Critique?
A research critique is a critical appraisal of 3. What Are the Stages of the

a piece of completed research. It involves Critique Process?
a high level of critical thinking and analysis
A critique of a research report involves a
of each component of the research study.
review of the report in its totality, as well
Critiques may be conducted on research
as a microscopic examination of each
reports, manuscripts, or published arti-
component of the study. The critic ap-
cles based on research studies or they may
praises the merit of a research project in
be done on research proposals. A critique
four stages. These include:
provides constructive criticism of a spe-
cic piece of work for the purpose of pro- Stage 1. Clearly understanding the mean-
moting excellence in research. Through ing of the purpose and problem state-
specic commentary, a critique can iden- ment of a study and determining if the
tify both the strengths and limitations of a research design and methodology are
study and assist the nurse in deciding how consistent with the study purpose.
best to apply the study ndings in practice. Stage 2. Studying the manner in which
Critiques are usually three to four pages the study was conducted and deter-
long and address the major aspects of the mining if the methodology was applied
research process. Critical appraisal in- properly.
volves judging whether or not a research Stage 3. Studying the ndings of a study
study is described clearly and comprehen- and assessing if the outcomes and con-
sively enough to decide if the ndings and clusions are believable and supported
implications are logical and believable and by the ndings (NHS, 1998).
THE RESEARCH CRITIQUE 613
Stage 4. Summarizing the overall quality lines for evaluating studies that emanate
of the study, identifying the strengths from different paradigms separately. The
and limitations of the study, evaluating critique of studies from the positivistem-
the contribution of the study to nurs- piricist paradigm, which uses quantitative
ing, and identifying suggestions for im- research methods, is discussed rst, fol-
proving the work. This is usually the lowed by a discussion of the critique
last paragraph or two of a critique. process for qualitative studies.
Each of the stages evaluates critical
components of the research report unique 1. Criteria for Critically
to that stage of the critiquing process. Box Appraising Quantitative
20.1 identies the components of a quan- Research Reports
titative research report that are evaluated
A number of nursing scholars have written
at each stage.
comprehensive lists of guidelines and ac-
companying questions that are helpful in
critically appraising quantitative research
B. CRITICALLY APPRAISING reports (Field, 1983; Fain, 1999, Parhoo,
RESEARCH REPORTS 1997; LoBionda-Wood and Haber, 1998).
The authors of this text provide a set of
Some researchers suggest that critical ap- criteria and a series of related questions
praisal of a research study is based on the based on ideas gleaned from the literature
match between the study purpose and the and their own research experience. The
other elements of the study (Kna and guidelines and questions should be ap-
Howard, 1984; Cobb and Hagemaster, plied with exibility and in various combi-
1987; Forchuk and Roberts, 1993). Because nations when evaluating quantitative nurs-
research that evolves from various para- ing studies.
digms (i.e., positivistempiricist or natu- The criteria for evaluating quantitative
ralisticinductive worldviews) differs con- research studies are presented rst, fol-
siderably in its purpose and underlying lowed by questions appropriate to each of
nature, it is important to discuss guide- the four stages of the critiquing process.
BOX 20.1 Stages of the Critiquing Process and Components of a Quantitative

Research Report
Stage 1: Purpose, Problem Statement, and Stage 2: Conduct of the Research

Congruency with Design and Methodology
Protection of participants rights
Study purpose Ethics of research
Problem statement Consistency and integrity of procedures
Theoretical framework
Stage 3: Outcomes of the Research Process
Literature review
Hypotheses or research questions Statistical significance of results
Research design Clinical significance of results
Sampling procedures Discussion of findings
Data collection procedures Implications for practice, education, and re-
search
Instruments
Data analysis Stage 4: Overall Quality of Study
Conclusions
Regardless of whether you are doing the section of the critique, you can suggest
critique of a research study or reading improvements to the researcher for
a critique published by someone else, strengthening the study in the future.
the criteria and questions you pose are Step 3. As you go through the list of ques-
the same. The following criteria are sug- tions and respond positively to each
gested for evaluating quantitative re- item, make a note of what you consider
search studies: the real strengths of the study. These
can then be summarized in the
1. Relevant and clearly articulated state-
strengths and Limitations section of
ment of purpose
the critique. It is important, however,
2. Consistency between explicitly stated
not to unfairly criticize the researcher
purpose and problem statement or
for something that was never a part of
question or hypothesis
the original purpose (Wilson, 1993).
3. Comprehensive literature review iden-
Step 4. As you apply the list of questions to
ties gaps in the research area and log-
your critique, you are advised to review
ically leads to the research questions
the chapters in the text that relate to the
under study
particular design used in the study, as
4. Theoretical framework provides a ra-
well as chapters on the various compo-
tionale for the study
nents of the research process. For your
5. Congruent match between the pur-
convenience, these are noted in paren-
pose, design, and method
theses after each heading.
6. Appropriate sample and sample selec-
tion procedures
7. Statistical procedures appropriate to
the type of data collected and research 2. Criteria for Critically
questions posed Appraising Qualitative Research
8. Adequate reliability and validity to ac- Reports
cept ndings and generalize to appro-
As more nurses engage in qualitative in-
priate populations
vestigations, it is important to know how
9. Signicance of study for nursing is ap-
to evaluate their contributions to nursing
parent
pracice. What general standards exist for
The suggested questions in Box 20.2 critically appraising the quality of a quali-
are listed for your consideration when tative investigation? Qualitative studies
conducting each stage of the critique. It is emerge from the naturalisticinductive
helpful if you use the following steps to paradigm and as such are based on a dif-
move through the stages of the critiquing ferent set of assumptions and a different
process: worldview and have a different purpose
than quantitative studies. These differ-
Step 1. Answer the questions in Box 20.2 ences make it difcult to evaluate qualita-
by providing content from the study to tive studies using the same criteria as
show how the researcher specically quantitative investigations. Qualitative re-
addresses or fails to address the search hopes to promote understanding
essence of each item. Avoid answering of phenomenathat is, that deep struc-
questions with a simple yes or no. ture of knowledge that comes from pro-
This will do little to enhance the re- longed engagement and persistent obser-
searchers ability to improve the study. vation in the eld, from visiting personally
Step 2. In situations in which you must with participants, and probing to obtain
answer no to an item, make a memo detailed meanings (Creswell, 1998). Be-
to yourself so that in the Limitations cause a critique of a study is based on the
BOX 20.2 Stage 1: Key Questions for Critiquing Quantitative Designs
Stage 1 Questions: Critical Appraisal of

Purpose, Problem Statement, and Congruence
Is the population to whom results will be
generalized described?
with Design and Methodology
Purpose or Problem Statement (see Chapter 3)
Are precautions taken to avoid collecting a
biased sample (see Chapter 9) that would
Is the study problem and purpose statement limit generalizability of findings?
clearly articulated? Are the demographic characteristics of the
Are the reasons for conducting the study sample described?
stated? Is the sample representative of the population?
Is the studys potential contribution to nurs- Are inclusion and exclusion criteria identified?
ing knowledge stated? Is the sample size appropriate to meet as-
Are the research objectives or research ques- sumptions of statistical tests?
tions or hypotheses stated clearly and re-
Data Collection (see Chapters 13 and 14)
searchable (answerable through the collec-
tion of empirical data)? Are data collection methods appropriate to
meet the study purpose and answer the
Are terms defined conceptually and opera-
questions or hypotheses?
tionally?
Literature Review (Chapter 3)
What evidence is provided that data collec-
tion procedures are valid and reliable?
Does the literature review or theoretical Are adaptations to data collection tools de-
framework provide evidence that the re- scribed?
searcher has synthesized the classic and cur- Are data collection instruments described in
rent literature and placed the research ques- sufficient detail to enable readers to ascer-
tion in the proper context? tain method of scoring and range of values
Does the literature review identify gaps in and what a particular score means?
knowledge, suggest how the current study ex-
Data Analysis (see Chapters 11, 12, and 16 to 18)
tends the knowledge base in this area, and
point out contradictions in the current knowl- Are data analysis procedures described?
edge base? Are the statistical techniques appropriate for
the study methodology (i.e., the type of data
Does the researcher summarize the litera-
collected and analysis)?
ture review, provide a rationale for the cur-
rent study, and show how this study will ex- Do the statistical tests answer the research
tend previous research? questions and specify the level of significance?
Design (see Chapters 4 to 8)
If results are nonsignificant, is a power analy-
sis conducted to explore nonsignificant find-
Is the study design specified, including its ad- ings (see Chapter 15)?
vantages and limitations for the research
Stage 2 Questions: Critical Appraisal of the
problem?
Conduct of Research
Is there evidence that a pilot study had been
Human Rights (see Chapter 10)
conducted and the findings were used to en-
hance the design? How are rights of research participants pro-
Is the design (overall plan of research) ap- tected?
propriate to the research purpose and capa- Are ethical issues anticipated and handled ap-
ble of answering the research question? propriately?
How does the design control for extraneous Procedures
variables?
Sample (see Chapter 15)
Are techniques used to ensure that there is
consistency in the data collection process?
Was probability or nonprobability sampling What procedures were used to keep research
used and the reason for the choice specified? conditions the same for all participants?
BOX 20.2 Stage 1: Key Questions for Critiquing Quantitative

Designs (Continued )
Are there strategies to limit errors in data Does the researcher discuss both clinical and
collection, recording, and analysis? statistical significance of findings?
Did any unplanned circumstances influence Does the researcher overgeneralize the find-
the results? ings beyond the appropriate population?
In experimental designs, is there evidence of Are limitations of the study such as sample
manipulation of independent variables, ran- size, inadequate instruments, sources of bias,
domization in selection of sample and assign- and so on identified and their implications
ment to experimental and control groups, discussed?
and control of extraneous variables (Wilson, Implications and Conclusions (see Chapter 19)
1993)?
Stage 3 Questions: Critical Appraisal of
Does the researcher identify important im-
plications of the study for practice, educa-
Research Outcomes
tion, or research (if appropriate)?
Findings (see Chapters 9 and 19) How do the findings of the study advance
Are the findings presented clearly and cor- nursing knowledge?
rectly and are they related to the theoretical Do new research questions emerge from the
framework? study?
Is there a clear statement of whether or not Stage 4 Questions: Critical Appraisal of the
the data support the hypotheses or answer Overall Quality of the Study
each research question? Overall Quality
Are tables and graphs clearly labeled, easy to
comprehend, and congruent with results What are the major strengths of the study?
presented in text form (see Chapter 19 for What are the major limitations of the study?
table construction)? Was the study described in sufficient detail
to facilitate a replication study?
Are findings presented in an unbiased man-
ner (see Chapter 9)? What are the major contributions of this
study to knowledge development in nurs-
Discussion (see Chapter 19) ing?
Are alternative explanations offered for the What suggestions might enhance the study
findings? and correct the limitations?
match between its purpose and other ele- nomenon that is recognized easily by
ments of the study, it makes sense that those who experienced it?
qualitative studies that have a different As with quantitative research, a num-
purpose from quantitative investigations ber of authors have proposed guidelines
should have a different set of criteria for for critiquing qualitative studies; how-
appraising their merit. As noted in Chap- ever, they are not as numerous. Box 20.3
ters 6 to 8, qualitative research includes a contains examples of criteria for critiquing
wide array of methods and designs, each qualitative studies. Before applying any of
reecting a different tradition and per- these criteria, you should rst read the re-
spective. Criteria used to evaluate qualita- port of the study in its entirety to get a feel
tive research must take this into account. for the essence of the study and its contri-
In critiquing a qualitative study, the critic bution to nursing. After this, you should
asks, Did the investigators get it right? read the report, again paying attention to
That is, did they publish an accurate and each stage of the critiquing process.
authentic account of the research phe- Creswell (1998) notes there is a gulf of dis-
tance among the authors discussing cri- quantitative critiques, but the questions
teria and standards for evaluating qualita- forming the content of each stage ad-
tive studies. The criteria must be general dress concerns of qualitative studies.
enough to be applied across the different The questions in Box 20.4 are intended as
qualitative traditions (i.e., grounded the- guidelines to help you apply the criteria
ory, phenomenology, ethnography, his- to qualitative investigations and provide
torical research). Considerable diversity a framework from which to analyze re-
and richness are evident in the criteria search reports. You are advised to con-
represented in Box 20.3. sult Chapters 6 to 8 for applicable aspects
Having examined the criteria present of qualitative research designs as you re-
in Box 20.3, you are now ready to respond spond to questions in this section.
to specic questions for each stage of the In the nal analysis, when one critiques a
critiquing process of qualitative reports qualitative study, the central question that
(Box 20.4). The stages of the process do must be addressed is: How do we know
not differ greatly from those used in that the qualitative study is believable,
BOX 20.3 Criteria for Critiquing Qualitative Research Reports
Leiningers Criteria (1990) Reciprocity: Evidence that an intense sharing,

Credibility: The truth value or believability of trust, and mutuality exists
the findings to the participants Sacredness of relationships: The researcher
Confirmability: Direct evidence from the in- respects the collaborative and egalitarian as-
formants, including mutual agreement pects of research and makes room for the
Meaning in context: Interpretations fit within lifeways of others
broader understandings of the experience Sharing of the privileges: The researcher
Recurrent patterning: Themes recur in se- shares rewards (e.g., royalties or publication
quence or regular patterns rights) with the participants whose lives he
Saturation: New data are redundant; no new or she portrays
themes emerge when additional information
Forchuk and Roberts (1993)
is collected
Transferability: Consistency or generalizabil- The research domain, topic, or question is
ity across settings specified and the appropriate design selected
All research procedures are in keeping with
Lincoln (1995) the stated purpose
Positionality: The text should reflect authen- The research method is described and the
ticity and honesty about its own stance and rationale is provided for its selection
the position of the author Appropriate review of the literature is re-
Community rubric: Research should be ad- corded and referred to when it is relevant to
dressed to and serve the purposes of the the research
community it takes place in Participants, informants, context, and re-
Voice: Research should give voice to partici- searcher are described in relevant detail
pants, and multiple voices should be heard in Data gathering and analysis are described in
the text detail and the researcher should discuss how
Critical subjectivity: The researcher shows these are appropriate for the method se-
evidence of developing heightened self- lected
awareness that enables one to understand The researchers interpretations and con-
his or her psychological and emotional states clusions must be consistent with the data
before, during, and after the research expe- and the importance and relevance of the re-
rience search for nursing are clearly addressed
BOX 20.4 Key Questions for Critiquing Qualitative Designs
Stage 1 Questions: Critical Appraisal of Data Collection

Purpose, Statement of Research Phenomenon,
and Congruence with Design and Methodology
Is data collection congruent with the study
purpose, research question, and qualitative
Purpose or Statement of the Phenomenon of tradition selected?
Interest
Are prolonged engagement and persistent
Is the research phenomenon of interest clearly observation in the field used to build trust
stated? with participants and ensure validity of data
Is rationale provided for approaching the collection?
study in an inductive, qualitative manner? Are data collection strategies and proce-
Is the philosophy of the research tradition dures described in sufficient detail?
described?
Data Analysis
Is a statement of self-understanding included?
Is there a single broad research question? Are data analysis procedures clearly described
Are their subquestions? and appropriate to the research tradition?
Does the initial question become more fo- What evidence is provided that data collec-
cused as data are collected and analyzed? tion and analysis are concurrent and ongoing?
Is the study purpose clearly stated (e.g., dis- Is there evidence of decision rules for ana-
covery, description, theory building, and so lyzing data, and does the researcher remain
on)? true to the rules?
What evidence is there of narrowing the
Literature Review coding as categories are systematically dis-
Does the qualitative method used call for a carded when they are unsupported by the
literature review before data collection? data and the researcher becomes more fo-
Does the review indicate that the researcher cused on data collection and analysis?
has expertise in the chosen area, know Is there evidence of theoretical saturation?
where gaps exist, and show how this study Stage 2 Questions: Critical Appraisal of the
will eliminate such gaps? Conduct of the Research
If a review is appropriate only after data col-
Human Rights
lection, is there evidence that this is done?
Is a framework appropriate? If so, is it pre- How are the rights of research participants
sented clearly? protected?
Are ethical issues anticipated and handled ap-
Design propriately?
Is the design (i.e., overall plan for the re- Procedures
search) appropriate to the research pur-
pose? What evidence is provided that research
Is there congruency between the methodol- meets the criteria of rigor (credibility: find-
ogy and the research question? ings must be understood and viewed as cred-
Is the context for the study adequately de- ible by the informants; trustworthiness: one
scribed? can believe the findings are true; and useful-
Is the researcherparticipant relationship un- ness: findings shed light on an important phe-
derstood? nomenon)?
Creswell (1998) identifies eight procedures
Sample for testing the truth value (appropriate rep-
Is purposive sampling used? resentation of multiple realities) and notes
Are informants capable of informing the study? that researchers should use at least two of
Is the selection of participants appropriate to these in any given study. The critic should ask
allow for saturation of data? which of these the researcher uses.

BOX 20.4 Key Questions for Critiquing Qualitative

Designs (Continued )
1. Prolonged engagement and persistent ob- Stage 3 Questions: Critical Appraisal of the
servation in the field to build trust with Research Outcomes
participants, learn about the culture and Findings
context, and correct distortions intro- Are the findings contextualized (i.e., pre-
duced by the researchers presence sented within the context of the circum-
2. Triangulation for purpose of corroborat- stances that influence their interpretation)?
ing evidence from different sources to Do readers of the report recognize the phe-
shed light on a theme or perspective nomenon or vicariously experience it?
3. Peer review as an external check of the Are the findings true to the data?
research process; this person asks hard Do the themes, categories, or theoretical
questions about the method, meanings, statements present a comprehensive, plausi-
and interpretations and keeps the re- ble, and meaningful picture of the phenome-
searcher honest non?
4. Negative case analysis, in which the re- If models, diagrams, or figures are used, are
searcher reworks the hypothesis until all they effective?
cases fit it Are the findings compatible with the field of
5. The researcher articulates researcher bias nursings knowledge base?
so that the reader knows what prejudices, Are reasons for incompatible findings ex-
biases, or past experiences have shaped the plored by the researcher?
researchers approach to the study
Discussion, Implications, and Conclusions
6. Member checks to solicit informants opin-
ions of the accuracy and credibility of the Does the researcher identify implications of
findings and conclusions. Some methods, the study for nursing (i.e., in practice, theory
however, such as hermeneutic analysis, do building, instrument development, educa-
not always conduct member checks be- tion, and research)?
cause their emphasis is on the interpreta- Is a context provided in which to use the
tion of meaning by the research team of the findings?
experience of the participants in the re- Do the implications and conclusions follow
search project. logically from the findings?
7. External audits in which a consultant ex- Do new research questions emerge from the
amines both the process and product of findings?
the qualitative study to assess accuracy Stage 4 Questions: Critical Appraisal of the
and determine if the findings, interpreta- Overall Quality of the Study
tions, and conclusions are supported by What are the major strengths of the study?
the data What are the major limitations of the study?
8. Rich, in-depth descriptions of partici- What are the major contributions of this
pants and setting are provided so that the study to knowledge development in the field
reader can make decisions regarding of nursing?
transferability of findings to other set- What suggestions might enhance the study
tings and correct the limitations?
accurate, and right? (Creswell, 1998, p. sible within the scope of this text to de-
193). Application to the qualitative report of scribe specic critiquing details for all of the
the criteria and questions provided in different qualitative approaches, we have
Boxes 20.3 and 20.4 should help the critic to summarized general standards for judging
answer this question. Although it is not pos- the quality of a qualitative study and
pointed out signicant differences where 1. Example of a Critique of a

appropriate. Quantitative Study
The research critique requires a system-

atic analysis of each section of the quan-
C. EXAMPLES OF RESEARCH titative report. The following is a critique
CRITIQUES of a research study by Asch (1996) titled
The Role of Critical Care Nurses in Eu-
Reading systematic reviews of nursing re- thanasia and Assisted Suicide. Excerpts
search studies is an excellent means for from the original study are provided in
developing skill in carrying out a critical italics to give the reader the essence of
appraisal of a project as well as learning each component of the research study.
about an area of nursing practice. Re- This is followed by the critique of that as-
sources are available to beginning re- pect of the report. Interested readers
searchers and nurses interested in locat- may wish to refer to a published critique
ing critical appraisals of research studies of this report by Mawdsley (1997). Both
applicable to nursing. Several sources of the research report and Mawdsleys cri-
critical appraisals of nursing research tique may be read in their entirety by
studies include: consulting the references in the recom-
1. The Western Journal of Nursing Research mended readings for the original source.
regularly publishes critiques and com-
mentaries of published research arti- a. Stage 1: Purpose, Problem
cles after the main study report. This Statement, and Congruence with
mechanism provides a rich source of Design
feedback to the author on the original
study and gives the opportunity for the ( i ) Purpose
researcher to respond to the critique. . . . Evidence from the literature sug-
2. Scholarly Inquiry for Nursing Practice: gests that nurses outside the U.S.A. may
An International Journal also publishes be as willing as physicians to engage in
original research studies followed by euthanasia. The purpose of this study
critical responses to research studies was to examine the role of U.S. critical
by experts in the eld. Both of these care nurses in acts of euthanasia and
journals are highly recommended as assisted suicide.
quality examples of critiques for any-
Critique. The purpose of the study is
one interested in learning more about
clearly and explicitly stated. It is consis-
the skill of critiquing.
tent with the background information
3. Many research texts provide critiques
presented. The purpose is important be-
of nursing studies in their chapters on
cause critical care nurses are in key posi-
the critique process. Examples include
tions related to end-of-life and life-saving
Fain (1999), Brockopp and Hastings-
care.
Tolsma (1995), and LoBionda-Wood
and Haber (1998). ( ii ) Problem Statement
4. Other resources available include ex- Little is known about the actual ex-
perts in the eld such as directors of periences and practices of critical care
nursing research units, research con- nurses when end-of-life decisions are
sultants with professional associations, imminent. In a survey of 943 Australian
and individuals designated to give ad- nurses, 218 reported being asked by a
vice on clinical effectiveness in health physician to engage in euthanasia; of
care institutions. these, 85% reported complying with the
request. Moreover, 16 nurses reported lic discussion, and proposed legisla-

complying with a patients request for tion. Almost all the discussion in this
euthanasia without having been asked area has focused on the role of physi-
to do so by a physician. In a survey of cians . . . In surveys of British and Aus-
278 Australian nurses, 52 (19%) re- tralian physicians, 7 to 29% admit hav-
ported taking active steps to bring ing performed euthanasia. In a recent
about the death of a patient, often with- survey, 218 of 828 physicians in Wash-
out being asked to do so by the patient ington State reported receiving requests
or the patients family . . . for assisted suicide or euthanasia.
Critical care nurses frequently care These physicians satised 38 of the 156
for patients who wish to die, and these requests for assisted suicide and 14 of
nurses are often in a position to hasten 58 requests for euthanasia . . . In one
their deaths. And, like physicians they study of 1210 oncology nurses in the
may also be in a position to engage in United States, 47% indicated that they
such activities outside the practice set- would vote to legalize physician-
ting, on behalf of friends or relatives assisted death, and 16% indicated that
who wish to die. they would under a physicians order,
. . . for purposes of this study eu- administer a lethal injection to a com-
thanasia and assisted suicide are de- petent, terminally ill patient who re-
ned as events in which someone per- quested such assistance . . .
forms an act with the specic intent of
causing or hastening a patients death, Critique. The literature review in this
but excluded are those acts that reect study is limited. It focuses on the role of
the withholding or withdrawing of life- health professionals in euthanasia and as-
sustaining treatment. By this denition, sisted suicide. Most of the literature in this
we include such acts as providing an in- area has focused on the role of physicians;
tentional overdose of narcotics or potas- hence, this gap in the literature is pre-
sium chloride or providing explicit ad- sented as the rationale for conducting this
vice to patients about how to commit study. References suggest that nurses may
suicide but exclude such acts as with- be an important group to study because of
drawing a mechanical ventilator, even their unique position in understanding the
though all these acts might result in the wishes of patients. The researcher cites
patients death. current studies conducted in this area and
notes that little is known about the actual
Critique. The problem statement is clearly experiences of nurses, particularly critical
articulated and identies the need to ex- care nurses. The researcher provides no
plore the role of nurses in euthanasia and summary of the literature.
assisted suicide. This is an important area
to consider because nurses are often in a (v) Methods: Design
position to understand and act on pa- A survey was mailed to 1600 critical
tients wishes. The research question is care nurses in the United States, asking
not specically stated but rather implied in them to describe anonymously any re-
the statement of purpose and problem. quests from patients, family members
The problem is researchable and the or others acting for patients, or physi-
terms are dened conceptually. cians to perform euthanasia or assisted
suicide, as well as their own practices.
( iii ) Literature Review
Euthanasia and assisted suicide Critique. The study used a descriptive
have received considerable attention survey design. The design was not explic-
recently in the medical literature, pub- itly stated; rather, it was implied by the
mailed survey instrument sent to U.S. lected. Furthermore, the denition of crit-
nurses. This design is appropriate to the ical care unit used by the researcher to se-
research purpose and is capable of an- lect the 1600 nurses was not specied.
swering the research question. A strength This resulted in eliminating more than 200
of the survey is its ability to tap many vari- questionnaires because nurses did not
ables, include many cases, and measure work exclusively in adult critical care set-
attitudes and perceptions of respon- tings (Mawdsley, 2000). A clear denition
dents. The researcher failed to identify of critical care unit is required because it
both the advantages and limitations of the may mean different settings to different
survey design to this study. nurses. For example, some may include
coronary care, step-down units, post-
(vi ) Methods: Sample
anesthetic units, and so on. The type or
The subjects in this study were nurses
size of critical care unit in which the
practicing in the United States in inten-
nurses practiced was not specied. This is
sive care units for adults. A random
a serious omission because previous re-
sample of 1600 subscribers to Nursing
search has shown a relationship between
magazine who practiced in critical care
withdrawing life support and the size of
settings was selected in order to repre-
unit and the type of patient (Mawdsley,
sent a broad spectrum of attitudes and
1997). The sample size and the response
experiences. Nursing is the largest nurs-
rate were adequate.
ing journal in the world, with a circula-
tion of nearly 500,000. Forty surveys
(vii ) Methods: Data Collection
were returned as undeliverable and
Data were collected by means of an
1139 completed surveys were returned
eight-page survey instrument that re-
by the cutoff date, a response rate of
quired 10 minutes to complete. The
71%. Subjects were excluded if they
terms euthanasia and assisted suicide
were not critical care nurses or did not
were dened for participants (see prob-
practice in clinical care settings (165
lem statement). To address the sub-
subjects), or if they practiced in pedi-
jects experience with requests that they
atric intensive care (46 subjects) or
perform euthanasia or assist in a sui-
emergency departments (9 subjects). A
cide, they were asked: Have you ever
sample of 852 nurses who practiced ex-
been asked by a patient, family mem-
clusively in critical care units for adults
ber, or other surrogate to administer a
remained. Table 1 (p. 1375 of the arti-
medicine to a patient or perform some
cle) shows selected demographic char-
other act with the intent of causing that
acteristics of the sample (age, gender,
patients deathother than withhold-
clinical experience, and practice site).
ing or withdrawing life-sustaining treat-
Critique. Signicant concerns center on ment? To address the subjects actual
the procedure for randomly selecting the practice, they were asked: While a crit-
sample, the denition of critical care unit ical care nurse, have you ever adminis-
used as part of the inclusion criteria, and tered a medicine to a patient or per-
the representativeness of the sample to formed some other act with the intent of
the population of practicing critical care causing or hastening that persons
nurses. Although Nursing magazine is a deathother than withholding or with-
popular journal containing articles for drawing life-sustaining treatment? Most
generic as well as critical care nurses, of the questions asked subjects to quan-
there are many critical care journals from tify their experiences during their ca-
which a more representative sample of reers and during the previous 12
critical care nurses could have been se- months. They were asked whether their
actions were undertaken at the request because common practice in many critical
of patients, patients families, other care settings is to withdraw ventilation
nurses, or physicians or with the ad- while providing opioids and sedatives to
vanced knowledge of these persons. make the patient comfortable (Ignatavicus
The initial surveys were mailed dur- et al., 1999). The study participants were
ing January and February 1995, and not given any validation of this common
two additional mailings of the same in- practice. Such lack of clarity in terms un-
strument to the same respondents fol- dermines the validity of the data collection
lowed at one-month intervals. As part of instrument and the strength of the results.
a related study of survey techniques, po-
(viii ) Methods: Data Analysis
tential subjects were randomly as-
Critique. The researcher does not pro-
signed to one of two groups. One group
vide a description of the data analysis
received a coded postcard with the sur-
procedures. From the Results section,
vey instrument, to be returned sepa-
one assumes that descriptive statistics
rately with the questionnaire. This min-
and measures of central tendency and dis-
imized the mailing of new copies of the
persion were used to describe the data.
instrument to nurses who had already
Sufcient information is not provided on
responded, while keeping specic re-
the analysis of qualitative data to deter-
sponses anonymous. The second group
mine how the handwritten comments and
of nurses did not receive postcards, so
explanations from the respondents were
each subject in this group received
treated. The quantitative measures ap-
three complete, identical packets. All
pear appropriate given the type of data
mailings included stamped, addressed
collected and the research purpose.
envelopes for return of the instruments
with instructions not to return more
than one completed questionnaire. b. Stage 2: Conduct of the Research
Critique. There is no validity or reliability ( i ) Human Rights

information provided about the study in- The completed surveys were anony-
strument. This makes it difcult to deter- mous. No identifying information was
mine if the instrument really does measure placed on the instruments or return en-
the nurses role in euthanasia and assisted velopes. The protocol was approved by
suicide. The researcher did provide den- the human subjects committees of the
itions of euthanasia and assisted suicide, University of Pennsylvania and the
but no critical care case examples differ- Philadelphia Veterans Affairs Medical
entiating between euthanasia and with- Center. The study was reviewed by an
drawing life support were provided, nor independent board selected by a major
was the concept of specic intent dened. professional nursing society and by two
This is a serious omission because such in- experts in the areas of bioethics and
formation is essential to understanding nursing.
what is meant by euthanasia and assisted
Critique. The researcher clearly ad-
suicide and differentiating between eu-
dresses the rights of participants and
thanasia and withdrawing life support
protects their anonymity and conden-
(Mawdsley, 1997). The researcher further
tiality. The study received ethical ap-
confuses these terms by referring to the
proval from an appropriate body.
withdrawal of ventilation as an example of
withdrawing life support, but the adminis- ( ii ) Procedures
tration of large doses of opioids to clients Refer to the Methods: Data Collec-
is dened as euthanasia. This is disturbing tion section.
Critique. No details are provided by the ( ii ) Discussion

researcher on strategies to limit errors in As public debate continues over the
data collection, recording, or analysis. social, moral, and professional issues
Research conditions were not the same surrounding euthanasia and assisted
for all participants because one group re- suicide, 19% of the nurses in this study
ceived a coded postcard and the other reported engaging in these practices.
group received three complete identical Some reported doing so on several oc-
packets of questionnaires. However, all casions, and some without the knowl-
respondents did receive the same ques- edge of physicians, patients, or surro-
tionnaire by mail. The researcher does gates and without their request.
verify that the characteristics of the re- If these practices had been sporadic,
spondents and their responses did not they might be attributed to a few lone
differ between the two groups. practitioners, operating beyond the
margins of their profession. Although
the moral appropriateness of an action
c. Stage 3: Research Outcomes is not measured by its pervasiveness,
some will nd it hard to accept the con-
( i ) Findings clusion that so many nurses in this sam-
Of the 1139 nurses who responded ple were acting inappropriately. We
(71 percent), 852 said they practiced ex- need to nd another explanation.
clusively in intensive care units for adults One possible explanation is that al-
in the United States. Of these 852 nurses, though these activities may have been
141 (17 percent) reported that they had undertaken with the intent to hasten
received requests from patients or family death, they may nevertheless reect a
members to perform euthanasia or as- continuum of moral acceptability and
sisted suicide; 129 (16 percent of those professional practice. At one end of the
for whom data were available) reported continuum, perhaps, are nurses who re-
that they had engaged in such practices; port hastening death in hidden ways
and an additional 35 (4 percent) re- for example, deliberately giving lethal
ported hastening a patients death by overdoses of medications. Others prac-
pretending to provide life-sustaining ticed at the limits of their authorityfor
treatment ordered by a physician. Some example, titrating intravenous drips
nurses engaged in these practices with- within prescribed ranges but beyond re-
out the request or advance knowledge of quired doses. Some nurses appealed
physicians or others. The method most fully or in part to the doctrine of double
frequently used for euthanasia was the effect, arguing either that their intent
administration of a large dose of an opi- was only to relieve suffering or that their
ate to a terminally ill patient. intent was both to relieve suffering and
to hasten death. Finally, some nurses re-
Critique. Findings were accurately pre- ported hastening death by administering
sented. Tables were easy to read and high doses of opiates while withdrawing
clearly presented but were only briey patients from mechanical ventilation. In
discussed in the text. Textual discussion these cases the death was imminent and
of the handwritten comments of respon- was an accepted goal.
dents was brief and weak. A more detailed Because so many different kinds of
discussion of the nurses description of activities are reected in the responses
their activities would have enhanced the of nurses it is difcult to ascribe a single
readers understanding of the nurses role meaning to the results or to take a sin-
in euthanasia and assisted suicide. gle moral stance toward them. Further-
more, a weakness of the study was the These are questions to keep in mind
failure to distinguish between euthana- when a researcher reports that nurses en-
sia and assisted suicide in the ques- gage in euthanasia without the consent of
tionnaire. In either case, however, the patients and families (Mawdsley, 1997).
intent to cause death was explicit . . .
( iii ) Implications and Conclusions
The results are subject to nonre-
The issues surrounding euthanasia
spondent bias . . . as nurses most upset
and assisted suicide are complex. In
by the survey might be those least likely
some cases the practice can appear to
to engage in such practices and least
be a genuine response to human suffer-
likely to respond. The high response
ing. Permitting health professionals to
rate suggests that such a bias could not
carry out these activities may seem ap-
have altered the ndings much. Finally,
propriate when the decision clearly fos-
the results of the study are based on
ters the patients autonomy. From this
self-reports. Some respondents may
perspective distinctions made between
have underreported or overreported ac-
euthanasia and the withholding or with-
tivities and some may have misunder-
drawing of life-sustaining treatment ap-
stood the questions. These limitations
pear articial and hard to sustain. In
call for caution in interpreting the point
most cases, the aims and consequences
estimates provided by the study.
of the actions are the same. On the other
Critique. The discussion is well done. Al- hand should euthanasia be sanctioned,
ternative explanations are offered for the it might become too easy an option.
research ndings. The researcher is thor- Maintaining legal and professional pro-
ough in the discussion, mentioning im- hibitions against euthanasia or assisted
portant limitations such as the question- suicide may limit such tragedies better
able representativeness of the sample, than any procedural safeguards . . . The
potential sample bias, and the lack of clar- results of this study should prompt
ity in denition of terms. However, the re- nurses, physicians, and other health
searcher fails to discuss the limitations care professionals to examine their
created by the lack of establishing the va- practices more openly and collabora-
lidity or reliability of the data collection tively, with the aim of understanding
instrument. Furthermore, the researcher and reducing disagreement over goals
failed to ask some key questions such as: and plans . . . Regardless of the policy
implications of this study it is clear that
Under what conditions were large
nurses in this study practice, often with
amounts of opiates administered by
little support, in extraordinarily difcult
nurses?
situations. In these complex environ-
Why was there no differentiation be-
ments, professional, moral, personal,
tween the administration of opioids
and religious values frequently collide.
during euthanasia and the withdrawal
of life support? Critique. Conclusions are well explored
Why did the researcher not acknowl- and implications of the study ndings for
edge the common practice of with- practice and future research are clearly
drawing mechanical ventilation in the stated. The study challenges nurses to
presence of large amounts of opioids ponder why a small number of critical
and sedatives and t these actions into care nurses reported hastening death
the denitions of euthanasia and as- without the consent of patients, families,
sisted suicide? Doing so would have or physicians. It also raises the question
eliminated confusion for the respon- of whether nurses can clearly differenti-
dents. ate between euthanasia and withdrawing
of life support. The study highlights the for Dying Patients (Rittman et al., 1997).
necessity for clear terminology regarding Excerpts from the original study are pro-
end-of-life decisions so that when one vided in italics to give the reader the
member of the health-care team proposes essence of each component of the re-
withdrawing life support, everyone is search study. This is followed by the cri-
clear regarding the meaning of the term. tique of that aspect of the report. The re-
(Mawdsley, 1997, p. 11). search report may be read in its entirety
by consulting the reference in the recom-
mended readings for the original source.
d. Stage 4: Overall Quality of Study
The questions provided in Box 20.4 pro-
vide a framework to analyze the report
( i ) Major Strengths
using the four stages of the critiquing
The major strengths of this study include
process.
its ability to challenge readers to face the
ethical dilemmas encountered everyday
in clinical practice, the clear identica- a. Stage 1: Purpose, Statement of
tion of the limitations of the study by the Research Phenomenon, and
researcher, and the indications for the Congruence with Design and
direction of future research. Methodology
( ii ) Major Limitations ( i ) Purpose or Statement of
The major limitations of the study include
Research Phenomenon
the lack of reporting on the validity and Little is known about how nurses ex-
reliability of the study instrument, the
perience caring for dying patients. Yet,
lack of clarity regarding terms, and the
entering the patients world often in-
questionable representativeness of the
volves dealing with death and dying and
sample to the population of practicing
is a major challenge to oncology nurses.
critical care nurses in the United States.
The purpose of this study is to explore
( iii ) Suggestions for Improvement the experiences of nurses engaged in re-
Much of the confusion for study partici- lationships with patients who were dy-
pants in completing the instrument could ing in order to describe skills and shared
have been eliminated if the researcher practices of oncology nurses. Three
had acknowledged the common practice questions guided the study: (a) What
of withdrawing ventilation in the pres- skills do nurses use in providing care to
ence of large amounts of opioids and dying patients and their families? (b)
sedatives and had incorporated these ac- What do nurses experience while caring
tions into the denitions of euthanasia for dying patients? (c) What meanings
and assisted suicide. In conclusion, the sustain nurses while working with dying
study addresses an important area of patients?
practice that is underinvestigated in the Critique. The purpose of the study and
nursing literature and brings to the fore- the research phenomenon are clearly
front an important dilemma for nurses stated. The researcher refers to the lack of
and other health-care professionals. knowledge about nurses experience of
caring as a rationale for conducting a qual-
2. Example of a Critique of a itative investigation in this area. A single
Qualitative Study broad question was used to address the
study purpose. Nurses were asked to de-
The following is a critique of the report scribe an experience of caring for a dying
Phenomenological Study of Nurses Caring patient. In addition, three subquestions
guided the study. No reference is made to for delaying the review of the literature un-
the philosophical stance on which the til this time is to limit the introduction of
phenomenological approach is based. Be- bias into the study process. The re-
cause the philosophical framework di- searcher is aiming for a pure description
rects the questions that are asked and in- of the phenomenon. In this study, the re-
uences the collection and interpretation searcher appropriately conducted a liter-
of data in qualitative studies, it would be ature review after data analysis, to place
helpful for the researcher to outline the the study ndings in the context of what is
philosophical orientation of the approach. already known about the topic of nurses
caring for dying patients. The researcher
( ii ) Literature Review does a thorough job of relating current
No section titled Literature Review is ndings to earlier work on the topic.
found in this report; rather, a section
called Background is presented by the ( iii ) Method: Design
researcher. The following excerpt is from
The hermeneutic method was used to un-
the Background section.
cover meaning embedded in nurses sto-
These ndings expand earlier work on ries or narratives about caring for dying
how nurses care for dying patients in on- patients. The purpose of a hermeneutic
cology nursing. Previous ndings have analysis is to achieve an understanding
been reported on the meaning of oncol- of the phenomenon being studied.
ogy nursing practice, description of crit-
Critique. The research design is clearly ar-
ical behavior in caring for dying pa-
ticulated and appropriate to the research
tients, and coping strategies used when
purpose. Hermeneutic phenomenology is
caring for dying patients and their fami-
a special kind of phenomenology designed
lies. Saunders and Valente described
to reveal concealed meanings in phenom-
bereavement tasks of nurses after a pa-
ena. It bridges the gap between the famil-
tient dies. Findings from our study indi-
iar and unfamiliar in our worlds.
cate that many bereavement tasks de-
scribed by Saunders and Valente occur
( iv) Method: Sample
in the daily practice of oncology nurses
and are used from admission through- Data consisted of six narratives written
out the dying process to achieve a good by experienced oncology nurses work-
death . . . This study also expands cur- ing on an oncology unit who are con-
rent understanding of reciprocity in re- sidered by their peers and head nurse
lationships between nurses and pa- to have a high degree of expertise. All
tients. Findings revealed that nurses of the nurses had at least ve years of
and patients experienced a mutually oncology nursing experience. They also
benecial process during their relation- reported that they had personally expe-
ships. Being engaged with patients who rienced the death of a family member
are dying provided an opportunity for or a close friend.
nurses to deal with their own mortality
Critique. It is implied that purposive sam-
and to develop a certain comfort with
pling was used to select nurses who had
death. Nurses often spoke of death not
experience caring for dying patients. The
as an enemy but as a friend.
researcher reports the number of nurses
Critique. With the phenomenological ap- in the study but does not describe how
proach, the review of the literature is not the six were selected from among other
usually conducted until data analysis is nurses on the oncology unit meeting the
complete (see Chapter 6). The rationale inclusion criteria.
(v) Method: Data Collection tions t with their lived experience of

oncology nursing practice. During dis-
Nurses were asked to describe an ex-
cussions and further analysis of all data,
perience of caring for a patient who
the themes cutting across all narratives
was dying. They were asked to select an
were identied.
experience that was important to them
because it taught them something about Critique. The use of a research team for
what it means to care for a dying pa- analyzing and interpreting data in a her-
tient. Stories included as much detail as meneutic study is a desirable way to pro-
possible about the nurses thoughts and ceed because different perspectives pro-
feelings in the situation. vide for a rich interpretation of the data. In
this case, the team has a good cross-sec-
Critique. The data collection method was
tion of nurses with varied and related ex-
clearly identied and congruent with the
periences. This is a valuable asset in the
study purpose, the research question,
analysis process. The researcher refers to
and the qualitative tradition. The three
the data analysis process used by Dikel-
guide questions used to facilitate the
mann et al. but does not give any detail
writing of the nurses narratives were
about what the process was except to say
previously identied. There is insuf-
that themes that cut across all narratives
cient detail, however, related to the
were identied by the research team after
length of the data collection period, the
discussion of their shared interpretations.
context and setting in which the narra-
What differences of opinion existed, and
tives were written, and the length of the
how were differences in interpretation
narratives. Also, there is no information
handled by the research team? Mention is
on whether the data were saturated (see
not made of decision rules for analyzing
Chapter 6). A more detailed discussion of
the data, nor is reference made to satura-
data collection procedures is required to
tion of data. It would be helpful to have the
enable replication of the study.
researcher provide a more in-depth dis-
(vi ) Method: Data Analysis cussion of the process for identifying
themes. For example, what steps were
A research team was used for data analy-
taken to organize and categorize the data?
sis purposes. The team included the
How were the data collapsed into four
principal investigator, two experienced
main themes? The researcher failed to
oncology staff nurses, the head nurse of
summarize the literature at this point,
the oncology unit, and a nurse adminis-
which would have facilitated the readers
trator responsible for supervising the on-
understanding of how pattern categories
cology service but not directly involved
and themes were recognized and ac-
with the direct care. The method of data
cepted. This is usually when the literature
analysis was adapted from the process
review is conducted in a phenomenologi-
described by Dikelmann et al. (1989).
cal study. However, it can be implied from
The team rst read each narrative as a
the background section of the report that
whole to gain an understanding of the
the literature was used effectively to in-
text. Members then wrote descriptions of
form the analysis.
the meanings evoked by the data. The
team shared interpretations and began
identifying possible themes in the data. b. Stage 2: Conduct of the Research
The principal investigator wrote in-depth
interpretations of each narrative. These ( i ) Human Rights
interpretations were validated by the Critique. There is no mention of protec-
participants to nd out if the interpretation of human subjects by the researcher.
It is not known what the procedure for in- ipants. This would have made it easier to
formed consent was or if anonymity was judge the usefulness of the ndings.
guaranteed. It can be inferred that the
writing of the narratives by the nurses c. Stage 3: Research Outcomes
who told their stories was considered con-
sent to participate, but we do not know the ( i ) Findings
extent to which the consent was free, in-
The shared experiences of nurses car-
formed, and noncoercive. It would be
ing for dying patients was represented
helpful to know if approval was received
in the data that clustered around four
from an ethics review board.
themes: knowing the patient, preserv-
ing hope, easing the struggle, and pro-
( ii ) Procedure
viding for privacy.
Critique. Three criteria are used by qual-
itative researchers to assess the truth Critique. The presentation of results is con-
value of their ndingscredibility, trust- sistent with a phenomenological study.
worthiness, and usefulness. Although the The ndings are presented in the context of
researcher does not explicitly address the group that was studied: oncology
the procedures for assessing this, it can nurses who cared for dying patients. When
be inferred from the description of the reading this report, readers are able to rec-
data analysis. The use of the research ognize what the experience is like for
team provides an opportunity for shared nurses caring for dying patients. Rich de-
responsibility for the interpretations scriptions of the themes enable readers to
made by the members. The team ap- vicariously experience the meanings asso-
proach helps to keep the principal inves- ciated with oncology nursing practice. The
tigator honest about the method, mean- four themes present a meaningful and com-
ings, and interpretations of the data. The prehensive picture of the experience of car-
interpretations of the team members ing for dying patients. The ndings include
were shared with the participants to de- a balance between the researchers inter-
termine if the interpretations t with their pretation of the meanings and the partici-
lived experience of oncology nursing pants quotations from their narratives.
practice. This use of member checks is a The ndings appear true to the data.
very valuable way to check the accuracy
( ii ) Discussion, Implications, and
and credibility of the ndings and conclu-
Conclusions
sions. No mention is made of researcher
bias. It would be helpful for the re- The four themes contribute to knowledge
searcher to identify the prejudices, bi- development about how nurses enter
ases, and past experiences of the re- into and experience caring for dying pa-
search team that shaped the approach to tients. The themes expand our under-
the study. Also, the relationship between standing of the nurses experience in
the research team and the participants oncology nursing practice. Nurses who
should be claried. This information is have expertise in caring for dying
useful in qualitative investigations be- patients establish different levels of in-
cause the researcherparticipant rela- volvement in different situations. Rang-
tionship inuences the sharing of the ing from intense and meaningful close-
experience in phenomenological investi- ness to engagement with less intensity.
gations. Although the researcher briey The ability to develop different levels of
described the participants, little attention intensity in relationships helped nurses
was given to providing in-depth descrip- manage the emotional demands in their
tions of the research setting or the partic- practice. Whatever the level of involve-
ment, nurses maintained a standard of and perspectives related to the re-

care that included knowing the patient, search topic
preserving hope, easing the struggle, and
( ii ) Major Limitations
providing privacy. These shared prac-
The major limitations of the study include
tices and meanings sustain nurses as
the lack of detail regarding the data col-
they encounter death daily.
lection and analysis procedures. In fair-
Critique. The researcher concludes that ness to the researcher, it may be that
based on this study, nurses who have omissions were evident in these areas be-
cared for dying patients develop shared cause of the stylistic considerations im-
practices and meanings that sustain them posed by the journal in which the original
as they face the death of patients on a daily study appeared. Cancer Nursing is a highly
basis. The conclusions follow logically respected journal that has a broad publi-
from the study ndings. The researcher cation mandate. It includes theoretical
notes that the four themes expand our un- and clinical articles as well as research
derstanding of the nurses experience in manuscripts and targets a broad profes-
caring for dying patients. It would be help- sional audience that includes researchers
ful if the researcher reported more speci- as well as nurses engaged in clinical prac-
cally the context within which the ndings tice across a variety of settings. This ne-
of this study could be used. No mention is cessitates a certain style and content re-
made of new research questions emerging quirement that differs somewhat from
from this study or implications for theory journals that focus exclusively on re-
building, education, or research. It ap- search reports.
pears that this study has signicant impli-
( iii ) Suggestions for Improvement
cations for theory development in nursing
Providing additional detail on the steps
related to caring for dying patients, but the
taken to collapse the data into the four
researcher does not mention this.
themes could enhance the study. Discus-
sion of the data analysis process would
have facilitated readers understanding of
d. Stage 4: Overall Quality of Study how the themes emerged. Additionally,
the researcher may wish to consider tri-
( i ) Major Strengths
angulation of method for the purpose of
The major strengths of this study are
corroborating evidence to illuminate the
threefold:
themes. This could be accomplished by in-
1. Its contribution to enhanced under- terviewing the participants who provided
standing of how nurses experience written narratives about their experiences
caring for dying patients and observing them in practice situations
2. The vivid and meaningful picture of the with dying patients and their families.
experience presented in the descrip- Overall, the study makes an important
tion of the four themes that resulted contribution to knowledge development
from the interpretations of the nurses in nursing. It enhances our understand-
narratives by the research team. ing of the nurses experience of caring for
3. The use of a research team for her- dying clients and describes the hallmarks
meneutic analysis composed of mem- of a high standard of practice in caring for
bers with rich and varied backgrounds dying patients.
E X E R C I S E S
1. Discuss the value of critical appraisal 3. Select a qualitative research article
skills to a practicing nurse who wishes from a nursing research journal such
to engage in research-based practice. as the Western Journal of Nursing Re-
search, Image, Qualitative Health Re-
search, or a comparable periodical and
critically appraise the article using the
2. Should different guidelines be used to guidelines presented in this chapter
critique quantitative and qualitative for qualitative critiques.
research reports? Justify your answer.
4. Select a quantitative research article

from a nursing research journal such
as Nursing Research, Image, or Re-
search in Nursing and Health and criti-
cally appraise the article using the
guidelines presented in this chapter
for quantitative critiques.
Asch, D.A. (1996). The role of critical nurses in Canadian Journal of Nursing Research, 25(4),
euthanasia and assisted suicide. The New 4757. This concise article discusses seven
England Journal of Medicine. This study guidelines for critiquing qualitative studies.
was based on a mailed survey to critical Leininger, M. (1990). Ethnomethods: The
care nurses and is a good example of how philosophic and epistemic bases to expli-
to report a quantitative study. cate transcultural nursing knowledge. Jour-
Beck, C.T. (1994). Reliability and validity is- nal of Transcultural Nursing, 1, 4051. This
sues in phenomenological research. West- article is a good source of criteria for qual-
ern Journal of Nursing Research, 16(3), itative work.
254267. This article compares and con- Lincoln, Y.S. (1995). Emerging criteria for
trasts the views of three phenomenologists quality in qualitative and interpretive re-
on the issue of reliability and validity (rigor search. Qualitative Inquiry, 1, 275289. A
in qualitative inquiry). contemporary approach to critiquing qual-
Creswell, J.W. (1998). Qualitative Inquiry and itative research projects is provided in this
Research Design: Choosing Among Five Tra- article.
ditions. Thousand Oaks, CA: Sage. This Mawdsley, C. (1997). Study says critical care
comprehensive text addresses issues re- nurses quick to aid in death of terminally
lated to standards for critiquing qualitative ill: A research critique. Canadian Associa-
reports. tion of Critical Care Nurses (CACCN), 8(2),
Downs, F. (1984). Elements of a research cri- 811. This article critiques the study by
tique. In Downs, F., and Newman, M. (Eds): Asch (1996). It is a good example of a cri-
A Sourcebook of Nursing Research (2nd ed.). tique of a quantitative study.
Philadelphia: F.A. Davis. This is a classic Rittman, M., Paige, P., Rivera, J., et al. (1997).
reference for anyone conducting a re- Phenomenological study of nurses caring
search critique on quantitative studies. for dying patients. Cancer Nursing, 20(2),
Forchuk, C., and Roberts, J. (1993). How to cri- 115119. An example of a qualitative study
tique qualitative research articles. The is critiqued in this chapter.
Chapter 21
Research Utilization in
Nursing
CHAPTER OUTLINE
A. Understanding Research Utilization C. Strategies to Facilitate Research-Based

1. What Is Research Utilization? Practice
2. Why Is Research Utilization Important 1. The Process of Planned Change
in Nursing? 2. Other Strategies
3. What Are the Steps in the Research D. Example of Research Utilization
Utilization Process?
1. Background to the Project
B. Barriers to Research Utilization 2. Research Utilization Model
1. Characteristics of the Nurse 3. Planned Change
2. Characteristics of the Setting 4. Evaluation of the Change
3. Characteristics of the Research E. The Future of Research Utilization
4. Characteristics of the Innovation
KEY TERMS
Change agent Evidence-based practice Researchpractice gap

Conceptual utilization Planned change Research utilization
Instrumental utilization
632
RESEARCH UTILIZATION IN NURSING 633
I
t is not enough for nurses to conduct ventions, new procedures, clinical proto-
high-quality, relevant research; the re- cols, and guidelines.
search must also be used in practice to A number of research utilization mod-
improve client care and benet society. els are available to guide you in promoting
The purpose of this chapter is to discuss quality care. Examples include the West-
the nature and value of research utiliza- ern Interstate Commission for Higher Edu-
tion in nursing. Steps in the research uti- cation (WICHE), Conduct and Utilization of
lization process are outlined and popular Research in Nursing (CURN) model, Nurs-
utilization models and projects are briey ing Child Assessment Satellite Training
described. Barriers to and strategies for (NCAST) project, Stetler Model, Dracup-
using research ndings in practice are dis- Breu Model, Iowa Model of Research in
cussed. An example of a research utiliza- Practice, and Horne Model. The models
tion project that is guided by the theory of have many features in common. Box 21.1 il-
planned change is presented. The chapter lustrates the major components of each
concludes with a discussion of what the model. Because of space limitations, only
future holds for research utilization in the Iowa Model is featured in the research
nursing. utilization example provided at the end of
the chapter.
A. UNDERSTANDING 2. Why Is Research Utilization

RESEARCH UTILIZATION Important in Nursing?
The ultimate value of research utilization

Research utilization is a complex process
is twofold: (1) to facilitate an innovative
that includes critically analyzing the liter-
change that will lead to improved client
ature, selecting appropriate interventions,
outcomes and (2) to validate existing nurs-
implementing them, and evaluating the
ing procedures and interventions. In to-
outcome (Goode et al., 1996).
days environment of scal restraint and
professional accountability, it is critical
1. What Is Research Utilization? that nurses are able to demonstrate that
the services they provide are relevant,
In the eld of nursing, research utiliza- cost effective, based on evidence, and lead
tion may be dened simply as the use of to improved client outcomes. The use of
research ndings in practice to improve research to direct client care is one way to
care. It occurs at two levels: instrumental engage in evidence-based practice. Such
and conceptual. Conceptual utilization practice challenges nurses to critically ex-
refers to the use of ndings to enhance amine traditional practices, procedures,
ones understanding of a problem or issue and nursing rituals and question those
in nursing. Through conceptual utiliza- that are not substantiated by research or
tion, you use the knowledge gained from other evidence. The use of empirical evi-
research to cognitively restructure the dence to inform nursing practice ensures
way you think about a situation, problem, effective use of scarce nursing resources.
or phenomenon in nursing. It enables you Research utilization is important at all
to see different alternatives and possibili- levels of nursing. It is crucial to the nurse
ties in nursing situations. Instrumental at the bedside as well as to nurses at the
utilization is the direct, explicit applica- broader levels in the health-care organi-
tion of knowledge gained from research to zation and the profession at large. Box
change practice. It includes (but is not 21.2 highlights the value of research uti-
limited to) the adoption of nursing inter- lization to each level.
BOX 21.1 Research Utilization Models
The Iowa Model focuses on change in practice mine the most effective mode. The following
based on research findings applied by the skilled components composed the dissemination of new
practitioner and emphasizes an organizational knowledge in this model:
focus with administrative support as an essential
component. The six main concepts of the model Recruitment of appropriate learners
are: Translation of research into clinical termi-
nology
Identification of knowledge or problem- Dissemination via various satellite modes
focused triggers Evaluation of outcomes
Primary goal is change in practice
Uses a process of planned change The Stetler Model (1994) is a prescriptive
Administrative support is essential model designed to guide individual nurses in the
Application of findings at the practitioner level implementation of research findings. The com-
Evaluation of change by those who imple- ponents of this model include:
mented it
Preparation to specify the reason for the re-
The Western Interstate Commission for search review
Higher Education (WICHE) was the first Validation is the critical appraisal of research
federally funded nursing research utilization Decision making related to application or
project in the United States. The five compo- nonapplication of results
nents of this model include: Translation involves identifying the practice
implications from the research
Definition of nursing care problem Evaluation of outcomes and dissemination of
Retrieval of relevant research findings
Critical review of the research
Development of research-based plan of care The Dracup-Breu Model focuses on using re-
Evaluation of the effects of change search to solve specific problems in the clinical
setting. The six steps of this model include:
The Conduct and Utilization of Research
in Nursing (CURN) was a federally funded Identify the problem
project carried out by the Michigan Nurses As- Select appropriate research
sociation with the purpose of developing re- Establish project objectives
search-based protocols for clinical practice. The Analyze the setting and devise the plan
components of the project include: Implement the plan
Evaluate outcomes
Identification of research studies and estab-
lishment of a research base The Horne Model focuses on research uti-
Transformation of findings into research- lization at the organizational level rather than at
based protocols the individual level of the practitioner. The four
Transformation of protocols into specific components of the model are:
nursing interventions
Clinical trials in the practice setting Organizational commitment identified in
goal statements and in the commitment of
Evaluation of the research-based practice
resources to research utilization
The Nursing Child Assessment Satellite Change agents whose primary responsibility
Training (NCAST) focused on teaching nurses is to create change in practice that is re-
new health assessment techniques for children. It search based
was directed at individual nurses. It used Rogers The process of planned change is used to
diffusion of innovation theory. Nurses across the prepare the organization
United States were taught specific assessment The outcomes of research-based practice in-
techniques using several communication tech- clude new policies, procedures, and proto-
niques. An evaluation process was used to deter- cols based on research evidence.
SOURCE: Compiled using data from Goode, C., Butcher, L., Cipperley, J., et al. (1996). Research Utilization: A Study
Guide (2nd ed). Ida Grove, IA: Horn Video Productions.
BOX 21.2 Value of Research eld, there remains a signicant gap be-
Utilization tween what we know in nursing as a result
of research and the actual application of
Value of research utilization this knowledge in practice. This is referred
to as the researchpractice gap. Recent
Promotes critical thinking and reflective
literature suggests there is a 10- to 15-year
practice
Enhances professional self-concept gap between the discovery of potential in-
Ensures provision of safe and effective novations and the implementation of
care these innovations in practice (Bostrom
Practice is based on current, scientifi- and Wise, 1994). All nurses share respon-
cally sound knowledge sibility to reduce this gap by appropriately
Self-confidence of the nurse is enhanced implementing empirical ndings in prac-
tice. This chapter can help you to develop
Value to the researcher
the basic skills to do so.
Validates the efforts of the researcher
Motivates scholars to continue to dis-
cover new knowledge 3. What Are the Steps in the
Reinforces professional accountability Research Utilization Process?
Helps uncover new clinical problems for
investigation The research utilization process is differ-
ent from but complimentary to the re-
Value to the health-care agency
search process (see Chapter 1). Although
Cost-effective nursing care the research process is essential to enable
High-quality care the use of ndings in practice, by itself it is
Improved client outcomes not enough. It would be naive to think that
Retention and recruitment tool if you conduct a scientically sound study
Professionally satisfied and stimulated and disseminate the results in a usable
nursing staff
and understandable format to those who
Value to the profession can use the knowledge, then your ndings
will be implemented. Unfortunately, the
Enhanced autonomy of practice
task of narrowing the researchpractice
Positive professional image
gap is complicated by a number of politi-
Strengthen professional status
cal, organizational, social, and personal
Expand the field of nursings scientific
knowledge base factors, some of which are beyond the
control of nurse researchers or practition-
ers. Let us consider the steps that nurses
can take to narrow the researchpractice
gap through research utilization.
Research utilization is also important to
clients. You will recall from Chapter 1 that Step 1. The rst task in narrowing the
nursing researchs goal is the improve- researchpractice gap is to select a rel-
ment of client care. With this goal in mind, evant problem area that requires em-
research utilization in nursing takes on a pirical evidence to bring about a posi-
different level of signicance than in other tive change in practice. Consensus on
disciplines in which the purpose of re- the problem area should be reached
search may be the discovery of knowledge among those responsible for the re-
itself for its own value. In nursing, research search utilization. Open discussion at
utilization should result in the improve- team conferences or unit meetings is
ment of practice and enhanced client care. one way to identify topics and reach a
Despite the many advantages that are shared vision about the problem to be
articulated by a variety of experts in the addressed. There are two types of trig-
gers that help nurses to identify poten- your practice setting. Is there a signi-
tial practice problems that can be im- cant research base to enable adoption
proved through research: problem-fo- of the innovation in practice? You
cused triggers and knowledge-focused should look for a preponderance of ev-
triggers (Oates, 1997). idence that supports the use of the
Problem-focused triggers are usually ndings in your practice setting. The
evident to nurses in the practice setting. research base should be substantial in
They represent clinical problems such both quantity and quality to support
as the increased incidence of urinary implementation of the ndings. Look
tract infections in settings where nurses for integrative reviews of the litera-
have been replaced by less skilled work- ture, meta-analyses of research stud-
ers or insomnia in clients receiving in- ies dealing with your clinical problem,
termittent intravenous therapy. and replication studies that yield simi-
Knowledge-focused triggers emerge lar ndings. It would be unwise to sug-
from the use of journal clubs in practice gest a change in practice based on only
when research articles are discussed one study. A number of authors have
by nursing members at regular club suggested guidelines for evaluating im-
meetings, the attendance of nurses at plementation of research ndings. Be-
professional and academic conferences fore making a decision to adopt an in-
where scientic papers are presented, novation in practice, the following
the reading of research literature, and criteria should be considered by those
the reection on standards and guide- responsible for research utilization:
lines for care. A knowledge-focused Utility to nursing practice: Does
trigger may be a report in a nursing nursing have control over the inter-
journal about the use of a new injection vention, procedure, or problem?
technique for administering insulin to Applicability to practice: Is the
juvenile diabetics. study sample, setting, context, and
Step 2. Review the literature to determine clinical problem applicable to your
what is known about the problem and situation?
related topic areas (see Chapter 3). Af- Replication: Have the ndings been
ter the literature sources have been replicated by others? If so, under
identied and retrieved, they should be what conditions?
carefully critiqued using the guidelines Scientic merit: Is the methodology
provided in Chapter 20 to determine if rigorous, yielding valid and reliable
the research is sufcient in quantity results and scientically sound con-
and quality to merit implementation of clusions?
the ndings in practice. A helpful way to Client safety: What risks and benets
synthesize the literature and summa- are available to the client if the inno-
rize the research ndings is to create a vation is adopted or not adopted?
grid that concisely displays the nd- Feasibility: Are there adequate re-
ings. You may need to review the for- sources ( i.e., time, money, people,
mat for setting up such a grid (see Table expertise) available to implement
3.1). The research utilization team may the ndings?
wish to divide the literature sources Step 4. After a decision is taken by the uti-
among members to critique the studies lization team to implement the ndings,
and then come together to discuss their a plan must be developed in writing to
ndings and synthesize the results. communicate the research-based inter-
Step 3. Determine if the ndings from the vention or protocol. This will ensure
literature are appropriate to apply in consistency in approach and substanti-
ate the research base for the innova- the research utilization process is the
tion. Such written research-based pro- nal step. The results of the evaluation
tocols and procedures should be in- of the innovation implementation and
cluded in the policy and procedure the evaluation of the process compo-
manuals of agencies and institutions nent of research-based practice need
(LoBionda-Wood and Haber, 1998). to be communicated to others in the
Step 5. Implementation of the planned in- nursing community. By publishing the
novation follows. This involves care- results of the research utilization pro-
fully detailing and executing strategies ject in professional journals, in clini-
for change. Usually new attitudes and cally based journals, in scholarly pa-
new behaviors must be established and pers, and at scientic and nursing
old attitudes and behaviors discour- practice conferences, the nursing com-
aged. All staff involved with the innova- munity accumulates evidence that can
tion must be part of the change process. be used to support or challenge origi-
Strategies for implementing the change nal research and direct future research
are varied and may include involvement utilization projects.
of stakeholders in the change process;
education of all staff on the nature, mer-
its and feasibility of the innovation; and
feedback on the change process and the B. BARRIERS TO RESEARCH
effects of the innovation. A number of UTILIZATION
models are available to guide the imple-
mentation of innovation based on re- In the 1980s, nursing, medical, and scien-
search evidence (Lewin, 1951; Rogers, tic journals and conferences were the
1995; and Healthcare Quest, 1997). Fig- only available sources of reliable re-
ure 21.1 illustrates the steps for carrying search data of relevance to practice. Now,
out a planned change in practice. The however, a preponderance of readily ac-
change process is elaborated on later in cessible sources for research data are
the chapter. available to nurses. Such sources include:
Step 6. Evaluation of the success of the
innovation is the nal step in the re- 1. The Cochrane Database of Systematic
search utilization process. Have the in- Reviews, which is a subset of the
tended outcomes been achieved as a Cochrane Library. The Cochrane Col-
result of the introduction of the inno- laboration movement was founded in
vation? What changes, if any, are re- 1993 by individuals from nine countries
quired to continue with the innova- and is a major force in the British Com-
tion? If the innovation did not produce monwealth countries. Although it fo-
the desired outcomes, the utilization cuses primarily on medicine, it has re-
team must assess what went wrong cently inuenced the eld of nursing to
and make suggestions for future prac- engage in evidence-based practice (Es-
tice. In addition to outcomes of the uti- tabrooks, 1999).
lization project, the process compo- 2. The Agency for Health Care Policy and
nent of research-based practice should Research (AHCPR) in the United States,
also be evaluated. This involves infor- which is a major depository of re-
mation on how the innovation is being search evidence in nursing
implemented, barriers to implementa- 3. The new Canadian Institute for Health
tion, and factors that enhance the im- Research, which is shaping the agenda
plementation process. for health research in Canada
Step 7. Dissemination of the ndings of 4. The Annual Review of Nursing Research,
nursing has had difculty enticing its mem-

Clearly articulate the desired outcomes of the bers to respond to research evidence by
change process, especially the benefits to the changing their practice. Many sacred
client
cows persist in nursing and go unques-
tioned even in the face of new evidence
Gain consensus on the desired change by all that has potential to improve client out-
those involved in achieving it comes. Why is this the case? A discussion
of factors that act as barriers to research
utilization may help in understanding the
Develop strategies for implementing change
and mechanisms for monitoring the impact of
gap that exists between the discovery of
change new knowledge and its implementation in
practice. Barriers may be classied ac-
cording to the following headings: charac-
Pilot test the change and analyze the pilot results
teristics of the nurse, characteristics of the
setting, characteristics of the research,
and characteristics of the innovation.
Draw conclusions about the effectiveness of the 1. Characteristics of the Nurse

change. Has it achieved the desired outcomes?
Why? Why not?
Knowledge, attitudes, and beliefs are three
factors to consider in assessing character-
istics of the nurse that may inuence re-
Continue to implement the change. Make search utilization. Nurses require two
adjustments as required to maintain the kinds of knowledge to utilize research in
change practice. Nurses must be knowledgeable
about the specic research studies that re-
late to their practice setting; this means
Monitor the effects of the change until the they must read widely in the research lit-
change becomes solidified as the established erature. A study by Estabrooks (1999) sug-
practice gests that there is a huge gap between pub-
lishing for scientic credit and publishing
Figure 21.1 How to plan and carry out a change for consumption by clinicians. She notes
in practice.
that in a random sample of 1500 staff
nurses, 38.7 percent read professional
which focuses on integrative research nursing journals rather than traditional
reviews of nursing studies scientic journals, in which most research
5. An array of nursing research journals studies are reported. Hence, many nurses
that publish the latest in research nd- are not exposed to research ndings in the
ings such as Nursing Research, the literature they read. Nurses must also be
Western Journal of Nursing Research, knowledgeable about the critiquing pro-
Applied Nursing Research, and Clinical cess so they can critically evaluate ndings
Nursing Research, to name a few. and be responsible consumers of re-
search. Although most baccalaureate pro-
Despite these sources of research- grams introduce nurses to skills in reading,
based knowledge, the reality of nursing interpreting, and evaluating research stud-
practice is that much of what we do as ies, there remains a strong need to build on
nurses is not based on scientic research. this knowledge through continuing educa-
Similar to other professions, the eld of tion sessions, conferences, and in-service
sessions. Many nurses who are prepared (1998) uncovered a number of beliefs
at the diploma or associate degree level do held by staff nurses that create unneces-
not have the necessary skills or knowledge sary obstacles and limit research efforts.
to critically evaluate research studies. A number of erroneous beliefs of staff
In addition to knowledge, nurses must nurses about research were identied.
develop positive attitudes and belief sys- These fallacies and other erroneous be-
tems toward research utilization. A liefs are listed in Table 21.1. The articula-
healthy attitude toward research is position of negative beliefs is a starting point
tively correlated with research utiliza- in changing attitudes and in developing
tion. In her work with nurses conducting respect for use of research among clinical
research in clinical settings, Morrison nurses.
Table 21.1 Fallacies About Research Held by Staff Nurses

Fallacies Facts
The best design is experimental or quasi- Nurses have a choice of many designs. The
experimental. current trend is to use designs that capture the
context even if there is less control.
A researcher must control all variables. In addition to controlling variables in the study,
they can be controlled statistically. In some
designs, control is not the goal.
The best measurement is physiological. False. The best measurement is holistic ( i.e., a
combination of bio, psycho, social).
The literature is searched for the exact study Search for conceptual similarities, even if the
using the same sample and equipment. studies were conducted on a different sample
and a slightly different topic.
Clinical expertise is a prerequisite. Research expertise and commitment are impor-
tant qualications.
Research is conducted only by academics who Academics and practitioners work together as
know little about practice. members of a research team.
Research is easy and can be conducted quickly Everyone should not conduct research, it
by anyone. requires time, and is not easy.
Ask for help when problems arise. Seek help during the planning stages so that
problems can be prevented and practice
inuenced positively.
Power and prestige accompany research funding. Real prestige comes from publication of results.
Hire someone to collect and analyze data. You learn best by completing your own data
collection and analysis.
It is not important to inform others. Professional ethics requires informing others
and seeking appropriate permission.
It is not necessary to clarify ownership of work. These issues should be openly discussed prior
to conducting the study.
If a study is not funded, it has no value. The merit of a study is judged by the quality of
its design and execution.
Oral presentations are better than poster It depends on the audience, both have their
presentations. place.
Source: Adapted from Morrison (1998). Erroneous beliefs about research held by staff nurses. The Journal of Continuing
Education in Nursing, 29(5), 196203.
Lack of time is a barrier often cited by quality circles or research dissemina-

nurses. Many nurses do not have time tion groups may improve the channels
while on duty to read and critically evalu- of communication.
ate research. Given the recent cuts to 3. Freedom from organizational constraints,
health care and the downsizing of the nurs- which implies that subordinates will as-
ing workforce, more and more demands sume responsibility for accountable
are being placed on nurses time. Pringle practice and managers will encourage
(1999) notes that it is unrealistic to expect and reward innovative practice. Often
staff nurses who do not feel valued, who health-care organizations value tradi-
are overworked and underpaid, and who tion and discourage change and change
feel their opinions do not count because agents from engaging in innovative
they are rarely asked, to engage in re- practice. They fail to reward nurses for
search utilization on their own time. Nurse change actions and prefer to maintain
researchers, administrators, and educa- the status quo, which they believe to be
tors must support the work life issues that in keeping with their vision of the insti-
staff nurses have so that their circum- tution.
stances can be improved. Only by working 4. Supportive leadership requires a com-
together can we expect staff nurses to en- mitment to change and innovation.
gage with researchers to identify clinical Leaders are either change agents them-
problems, test interventions, and apply selves, opinion leaders, or supporters
new knowledge in practice for better client of others on their team who demon-
outcomes (Pringle, 1999). strate skill in directing change and inno-
vation. Traditionally, leaders in health
care have often operated with an au-
2. Characteristics of the Setting thoritative management style, which is
opposed to change.
Health-care organizations have factors
5. Trust must be present to risk the fail-
operating that both promote and oppose
ures that come at times with change
innovation. For a research culture to ex-
and innovation. If the environment is
ist in an organization, ve characteristics
one in which collegiality and mutual
must exist (Crookes and Davies, 1998):
trust exist, it is likely the staff will work
1. An ethos of openness to new ideas and a to achieve common goals. However, if
willingness to question the status quo trust does not exist, then politics and
with an eye toward continual quality impersonal interests will predominate
provement. Health-care organizations and set the stage for failure of innova-
and those in positions of power often tion.
have a vested interest in maintaining
the status quo. They fail to initiate
change and to reect on new and innov- 3. Characteristics of the Research
ative answers to problems.
2. Interpersonal and information linkages The best research studies are of little
are essential for new research ideas to value if they are incomprehensible to
be communicated openly and receive those who apply knowledge in practice or
support. In bureaucratic structures, if they are published in sources that are
communication ow may be slowed foreign to typical staff nurses. Nurse clin-
or unidirectional from the top down. icians must be able to read and under-
This impedes the cascading of infor- stand research results if they are to apply
mation and limits upward movement the ndings in practice. This places the
of feedback or ideas. The formation of burden on the researcher to communi-
cate results clearly and comprehensively norms and practices, then innovation is
and to overtly identify the relevance of difcult. In such situations, it may be
their ndings for nursing practice. Re- better to settle for small changes and
searchers need to publish widely in jour- work gradually toward the bigger pic-
nals that are user friendly to nurses who ture.
are in direct care positions. 3. The complexity of the innovation is in-
Nurses need to focus on the type of re- versely related to its successful adop-
search questions asked and ensure that tion.
they are studying problems of impor- 4. The trialability or pilot testing of an in-
tance to nursing practice. This is best ac- novation is less threatening than to-
complished by working together with tally endorsing a major change in prac-
staff nurses to identify clinical problems tice. If it is not possible to introduce
that require a solution. Staff nurses can the innovation in stages, it will be more
become part of the research team. This difcult to convince others to take the
helps to improve the research skills of risk of total implementation.
the staff nurse as well as enhance the rel- 5. Observability refers to how visible the
evance of clinical research programs. benets and limitations of the innova-
Replication of research studies is war- tion are to participants and observers.
ranted. In the past, a lack of replicated re- In other words, seeing is believing, so
search acted as a barrier to the utilization the more obvious the drawbacks, the
of ndings. Nurses did not appear to more difcult the implementation. Con-
value the merit of replication studies. To- versely, the more obvious the advan-
day we realize that studies need to be tages, the easier the implementation.
replicated in different settings and with
Based on the attributes of the nurse,
different populations so that a sound
the setting, the research, and the innova-
body of evidence may become increas-
tion, a number of challenges are continu-
ingly available to staff nurses who are in-
ously presented to nurses who are hoping
terested in evidence-based practice.
to implement an innovation in practice.
Let us now consider some strategies one
4. Characteristics of the can use to overcome such barriers and
Innovation successfully implement research ndings
in practice.
The attributes of a particular innovation
can be a barrier to successful implemen-
tation. Rogers and Shoemaker (1971)
C. STRATEGIES TO FACILITATE
identied ve characteristics that pre-
RESEARCH-BASED PRACTICE
dominate in successful attempts at inno-
vation. If any of these ve are missing, the
The implementation of research ndings is
implementation may be impeded:
a daunting challenge for individual nurses.
1. The innovation must offer a relative ad- The presence of a change agent is required
vantage over the status quo. If the in- on the unit. Vaughan and Pillmoor, cited in
novation is not perceived to improve Crookes and Davies (1998), dene a
the current situation, then implemen- change agent as a person or group of peo-
tation is unlikely. ple who can take an idea for change from
2. Compatibility of the innovation with the the drawing board, carry it through all
current practice enhances the likeli- the stages of its implementation, and eval-
hood of successful adoption. If the in- uate its success or failure. Change agents
novation is quite different from existing must be able to respond to the barriers in
their midst that mitigate against research ever, the restraining forces outweigh
utilization. They must be willing to take the driving forces, the chances of suc-
risks and accept setbacks when new ideas cessfully implementing the innovation
prove inappropriate to a given setting. It is are reduced considerably. Factors to
advisable for change agents to be clear be considered at this stage include the
about their purpose and follow some sort characteristics of the setting, of the in-
of plan to initiate and manage change. A novation, and of those responsible for
number of models are available to provide implementing the innovation; the inter-
structure and direction to the change est level for change; the depth of moti-
agent. Let us briey consider the concept vation for change; and current satisfac-
of planned change as a strategy for re- tion with the status quo. After the need
search utilization. for change has been recognized and ac-
cepted by the group, the change agent
moves to the next level of the change
1. The Process of Planned Change theory.
2. The moving phase follows the unfreez-
Planned change is an intentional process
ing stage. In this phase, the change
that involves a change agent who works to
agent identies, plans, and implements
bring about a demonstrable improvement
the new innovation, making certain
in the status quo. Models of planned
that the driving forces outweigh the re-
change are helpful to those who wish to im-
straining forces.
plement research-based innovations in
3. The refreezing phase is the nal stage.
practice. They provide a framework within
It involves consolidation and evalua-
which to assess factors that may inuence
tion of the innovation in the environ-
adoption of the change; they provide
ment. If refreezing does not occur, the
structure for the actual implementation
potential to return to the old prac-
stage; they highlight the need to restabilize
tice is high. The new practice must be
the change environment after the innova-
continually reinforced or the practice
tion implementation and to evaluate the
will become intermittent and eventu-
success of the planned change attempt
ally be replaced by the old practices.
(Crookes and Davies, 1998).
One of the simplest and most popular
Other theorists such as Rogers (1962,
models of planned change is the classic
1995) built on the work of Lewin by em-
model proposed by Kurt Lewin (1951),
phasizing the background of the potential
the father of change theory. Most of the
users, characteristics of the innovation,
contemporary change theorists, such as
and environment in which the change is to
Havelock and Rogers, have built on the
be implemented as major factors to be
work of Lewin. His theory consists of
considered in the planned change process
three stages to consolidate an innovation
(Crookes and Davies, 1998). These addi-
in practice. These are:
tional factors contributed to Rogers ve-
1. The unfreezing phase in which the stage theory of diffusion of innovation.
change agent carries out an analysis of The stages are:
the current situation to determine both
the driving and resisting factors to the 1. Knowledge or the awareness of an in-
proposed innovation. This is consid- novation for use in practice
ered the diagnostic phase of the model. 2. The persuasion stage, during which
If the driving forces outweigh the re- nurses form an opinion about the in-
straining forces, the innovation has a novation
good chance of being adopted. If, how- 3. The decision stage, in which a decision
is reached to adopt or reject the inno- care. The strategies range along a contin-
vation uum from informal sharing of research in-
4. The implementation stage, which information with colleagues to formal mod-
volves using the innovation to change els of planned change. Some suggested
practice activities for each category of nursing role
5. The conrmation stage, which either re- include:
inforces the decision to adopt the inno-
vation or rejects the change in practice
Practicing nurses should deliberately
expose themselves to research ndings
Rogers model provided the basis for so they develop a research-mindedness
implementing the rst major research uti- to their practice. This can be accom-
lization processes in the United States plished in a variety of ways, such as crit-
the Western Interstate Commission for ically reading research literature re-
Higher Education (WICHE) and the Con- lated to their practice areas, joining
duct and Utilization of Research in Nurs- journal clubs at work that focus on re-
ing (CURN) projects. The former was ini- search articles with ndings related to
tiated in the mid-1970s and was the rst practice, attending professional and sci-
major effort to support nurses in the eval- entic conferences where research pa-
uation of research aimed at solving prac- pers are presented, and supporting the
tice problems. The latter project was con- efforts of researchers in their settings.
ducted from 1975 to 1980 and had as its Educators can role-model utilization of
goal the increased use of research nd- research in their teaching of nursing
ings in nursing practice by disseminating content and process. This can occur at
quality ndings and supporting their im- the undergraduate and graduate levels
plementation. Rogers diffusion theory as well as in continuing education pro-
also provided the theoretical basis for grams. Educators should engage their
the Nursing Child Assessment Satellite students in reective practice and crit-
Training project (NCAST, 1976 to 1985). ical appraisal of the research literature
This project focused on nurses working through critiquing assignments and re-
in the home environment and eventually search reviews of the literature.
was expanded to include education of Researchers should conduct rigorous
nurses in the hospital setting about research and clearly identify the impli-
motherchildnurse interaction and as- cations of their research for practice.
sessment of infant states and behavior in Researchers should work collabora-
hospital. The project reported a high uti- tively with practitioners to identify rel-
lization rate by nurses who reported re- evant practice problems and include
ceiving the education regarding the re- practitioners on their research teams.
search-based protocols. In brief, Rogerss Researchers need to present their
model has been used effectively to assist work at local, regional, national, and
nurses as change agents in using re- international conferences and publish
search-based innovations in practice. in appropriate research and clinical
journals, as well as popular literature
so that their work is exposed to those
2. Other Strategies who are in the best positions to imple-
ment the ndingsthe practitioners
The utilization of research ndings in
of nursing. Replication of previous
practice is a shared responsibility. Nurses
studies should be a goal of researchers
at all levels, including practice, education,
who are interested in mounting a body
administration, and research, have unique
of evidence supportive of research-
roles to play in promoting research-based
based practice. These strategies will
go a long way in promoting use of re- dren is a serious problem that negatively
search ndings to improve client care. impacts client outcomes.
Scholars should incorporate research A large childrens hospital in Philadel-
ndings into their textbooks and schol- phia used a needleless valve device to
arly publications when appropriate. They maintain the intravascular catheter. The
also need to create theories to help ex- device provided staff nurses with a route
plain research ndings and work with re- to administer medication and draw blood
searchers to test theory through re- samples. A potential problem of this sys-
search. tem was the risk of infection caused by the
Administrators should set institutional increasing frequency with which needle-
goals that include the use of research less systems had stagnant blood around
ndings to promote excellence in prac- the valved device. Hence, a clinical nurse
tice. Management can support the ef- specialist organized a multidisciplinary
forts of practitioners to work to accom- team to discuss the problem and decide
plish such goals by rewarding research how to provide better care to children on
utilization efforts in tangible ways such the unit with intravascular devices.
as extra days off, nancial stipends, em-
ployment honor rolls, and special din-
2. Research Utilization Model
ners and other celebrations to recog-
nize staff efforts. Financial and human The Iowa Model for Research-Based Prac-
resources need to be designated for re- tice (see Fig. 21.2) provided an algorithm
search utilization efforts. ( i.e., a set of steps) for the research uti-
lization team to follow in addressing the
problem. The model is initiated by identi-
D. EXAMPLE OF RESEARCH cation of a problem trigger or a knowl-
UTILIZATION edge focus trigger. In this situation, both
triggers were present. The multidiscipli-
The process of research utilization is a nary research utilization team identied
complex one requiring critical thought, the problem trigger as contamination of
planned action, and deliberate choices the needleless device. The intermittent
by nurses committed to excellence in nature in which intravascular catheters
practice. This chapter equips you with are entered contributed to the teams
some basic skills to help you overcome nding many valves that contained stag-
barriers to research utilization and pro- nant blood. When blood is stagnant for
ceed logically with a plan of action. This long periods of time, the potential for mi-
section presents an example of a re- croorganism growth exists; this is a seri-
search utilization project that was guided ous problem in a pediatric population
by the Iowa Model for Research-Based with compromised immune systems. The
Practice (Fig. 21.2) and the Theory of team believed that the valve should be re-
Planned Change (Havelock, 1972, 1995). moved until more data could be ob-
tained. The knowledge trigger emerged
from the Bloodborne Pathogen Standard
1. Background to the Project and the guidelines produced by the Cen-
ters for Disease Control and Prevention
The placement of an intravascular cath- to protect health-care workers from in-
eter is a common treatment for children jury and to provide protection to clients.
who require antibiotics, chemotherapy, Applying the Iowa Model, the next step
total parenteral nutrition, or uid ther- was to identify relevant research literature
apy. Infection of central lines in ill chil- on the topic. In the autumn of 1995, the clin-
ical nurse specialist and staff educator for with equipment vendors. The consensus
pediatrics and the pediatric intensive care was that there was sufcient evidence to
unit (PICU) completed a critical review of support conducting an evaluation of a new
the literature and conducted interviews product: a nonvalved needleless system. A
3-month trial period of the new product uation committee. The CNS presented a
was granted to the pediatric setting and the summary of the trial to pediatric nurses,
PICU. physicians, and the product evaluation
To prepare for the trial, communication committee and recommended that the
was established with the nurses through product be used in the pediatric depart-
the unit-based council, with physicians ment. As a result, the product was ac-
through individual meetings, and with ven- cepted for use with modications.
dors through the clinical nurse specialist
(CNS). The CNS conducted an in-service
with 100 percent of the staff on all shifts. 3. Planned Change
Throughout the trial period, vendors were
available and nurses considered to be ex- Many factors must be considered in initi-
pert users of the new system supported ating a planned change. Someone ( i.e., a
other staff members on all shifts. The re- change agent) must be mandated to help
search utilization team maintained a visible plan and manage change, and a problem
presence during the trial, providing infor- must be identied. In this situation, the
mation as needed to staff members. An change agent was the CNS and the client
evaluation survey was conducted midway system was the nursing staff and together
through the trial and again in the summer they developed a well-thought-out plan
of 1996. The results from both surveys to guide the change process and imple-
were positive. Results were shared with the ment a new procedure for maintenance of
nursing directors of pediatrics and PICU, intravascular catheters. Havelocks The-
the physician group, the purchasing de- ory of Planned Change guided the project
partment, and the hospitals product eval- (Box 21.3).

4. Evaluation of the Change 1. Sort through published results and de-

cide which ndings are credible and
The evaluation of the change was con- which are not.
ducted during and at the end of the trial 2. See patterns, inconsistencies, and con-
to determine appropriateness of the tradictions in ndings from several
change. It is necessary to continue to studies on the same topic.
monitor outcomes after successful im- 3. Decide whether a summary of evi-
plementation. The Iowa Model divides dence compiled by someone else is
outcomes into three categories: patient credible (Brown, 1999, p. 188).
and family, staff, and fiscal. Outcomes
are monitored through quarterly quality In the past, the eld of nursing missed
improvement tools, infection control opportunities to shape practice through
data, monthly product cost, and satis- nursing research. Today, however, nurses
faction of the staff and patients (and pa- are better educated and better prepared
tients parents). Staff must continue to than ever before to engage critically in ev-
provide feedback to the CNS on the idence-based practice and to value it as an
change. In this example, as the change important part of professionalism.
was consolidated in the system, the The future of nursing research utiliza-
change agent ( i.e., the CNS) was gradu- tion will require commitment from individ-
ally removed, leaving the client system ual nurses, hospital and community-based
( i.e., the staff ) to monitor and maintain nursing units, the health-care delivery sys-
the change. Individuals within the client tem, and the education system. We must
system that could maintain the change all do our share to ensure that the many
must be identified. When the client sys- ne examples of nursing and interdiscipli-
tem had embraced the change and acted nary health research are used by practi-
as the change agent, the change process tioners to shape and improve nursing
was terminated. practice and health care for our clients. If
research-based practice is to become part
of the culture of nursing, then the following
should occur:
E. THE FUTURE OF RESEARCH
UTILIZATION
Practitioners must seek out continuing
education opportunities that are re-
What does the future hold for research search-based.
utilization? Given todays increasing em- Organizations must provide nurses with
phasis on quality, evidence-based prac- the computer and library resources to
tice, accountability, and scal responsi- do literature searches and the time for
bility in health care, it is likely that practitioners to read research and im-
research utilization will be a major focus plement ndings in practice.
of all health professionals in the future. Job descriptions of nurses should re-
Having reached the last chapter in this ect the necessity of using research
book, you have obviously made the com- ndings in practice.
mitment to learn the skills required to be Organizations should hold practition-
a responsible consumer of research and ers accountable for using research nd-
now have sufcient knowledge to engage ings appropriately in their practice.
in research-based practice at a beginning Nursing education must incorporate
level. The utilization of research is a com- research ndings into course content.
plex process requiring an informed con- Research skills should be presented to
sumer who has sufcient knowledge to students as fundamental skills essen-
be able to: tial to understanding the nursing liter-
ature and to being a responsible con- utilization will not always be easy. Many
sumer of research ndings. nurses may question why it is important
Textbooks should be selected for their to do research. Some will say that re-
inclusion of research-based knowledge search dollars are better spent in direct
and citations of empirical studies. care at patients bedsides. This caution-
The discipline must communicate ary note is included not to discourage
clearly to its members and students the you but rather to prepare you for con-
importance of being informed con- frontation, skepticism, and doubt you
sumers of researched-based knowledge may face, and the fact that you will raise
(Brown, 1999). more questions than you answer. Im-
provements in practice will only occur
If these activities are consistently when you can demonstrate that new
adopted, the eld of nursing will be suc- knowledge leads to better health out-
cessful in communicating the high value comes. Research is necessary to create
it places on research-based practice. this new knowledge. Be persistent and
The future will see nurses engaged in have faith in your ability to create posi-
networking extensively across a global tive change through research. Even the
village to bring the latest research nd- most die hard cynics must agree with
ings to their nursing practice areas. change in the face of powerful evidence.
Through the use of the Internet and jour- As you conclude reading this text, we
nals such as the Online Journal of Knowl- hope it is just the beginning of your re-
edge Synthesis for Nursing, nurses will search career in nursing. We encourage
have access to evidence they can use to you to re-read sections of this text as you
shape future practice. Clinical settings engage in research-based practice. As
will support nurses in promoting com- you progress to more advanced levels of
prehensive research activity. Indeed, the nursing practice, you will come to realize
future for research utilization looks that you can make a difference through
bright in nursing. In this era of innovation research. You now have the attitude,
and health-care reform, everyone must knowledge, and skills to do so. What re-
share in the responsibility to see practice mains is the desire, motivation, and com-
improved through appropriate applica- mitment to make it happen. Remember
tion of research ndings. that the ultimate value of research is the
A word of caution, howeverresearch extent to which it is used in practice.
E X E R C I S E S
1. Select a problem from the practice set- 2. Using a theory of planned change,
ting that is of concern to the eld of identify the steps one would engage in
nursing. Do a review of the literature to to implement a research utilization
determine the quantity and quality of project to deal with the problem you
research available on the problem identied in exercise 1. How would you
topic. Summarize the critique of the lit- evaluate the success or failure of the
erature. Make a determination as to plan? Why is consolidation of the inno-
whether or not there is sufcient evi- vation in practice important to a re-
dence to proceed with a research uti- search utilization project?
lization project. Justify your decision.
3. You are attending a clinical conference Authoritative management style

on your nursing unit and a comment is
made that the ongoing nursing re-
search utilization project is taking too
much staff time and shouldnt be part Negative staff attitude toward re-
of staff nurses jobs. How would you research
spond to such a comment? Whose re-
sponsibility is research utilization?
Provide a rationale for your answer.
Limited research critiquing skills
4. As a practicing nurse, how would you

increase research use in nursing? Limited replication studies on the
topic
5. What strategies would you suggest for

dealing with the following barriers to 6. Select a nursing procedure from the
research utilization? procedure manual at your clinical
Lack of time agency or your nursing skills text. Do a
literature search to determine what, if
any, research evidence is available to
support the steps in the procedure.
Bostrom, J., and Wise, L. (1994). Closing the Nilson, K., Nordstrom, G., Krusebrant, A., and
gap between research and practice. Jour- Bjorvell, H. (2000). Perceptions of research
nal of Nursing Administration, 24(5), 2227. utilization: Comparisons between health
Comprehensive descriptions tell how a care professionals, nursing students, and a
group of nurses improved the quality of reference group of nurse clinicians. Journal
nursing care by facilitating the transfer of of Advanced Nursing, 31(1), 99109. A good
research knowledge to current practice. discussion of barriers to and facilitators of
Crookes, P., and Davies, S. (1998). Research nurses use of research ndings in practice
into Practice. London: Bailliere Tindall. This from various perspectives.
is an excellent guide to critically evaluat- Oates, K. (1997). Models of planned change
ing, synthesizing, and utilizing research- and research utilization applied to product
based literature in professional practice. evaluation. Clinical Nurse Specialist, 11(6),
Goode, C., Butcher, L., Cipperley, J., et al. 270273. A good example of both the re-
(1996). Research Utilization: A Study Guide. search utilization process and the change
Ida Grove, IA: Horne Video Productions. process applied to a clinical practice prob-
This helpful guide takes readers step by lem.
step through the process and activities in- Parahoo, K. (2000). Barriers to and facilitators
volved in research utilization. of research utilization among nurses in
National Health Services (1998). Changing Northern Ireland. Journal of Advanced Nurs-
Clinical Practice. London: Department of ing, 31(1), 8998. A survey of nurses per-
Health. A brief review of the process in- ceptions of barriers toand facilitators
volved in changing clinical practice. ofresearch utilization.

Part 8
APPENDICES
A
ppendix A contains an introduction to analyzing data, creat-
ing new variables, and working with output using SPSS for
Windows (version 10.0). Appendix B contains a sample
questionnaire. This questionnaire will serve as an illustra-
tion of formatting and also be used in the sample data set and exer-
cises contained in the instructors manual for this text.
651

Appendix A
Using SPSS for Windows*
APPENDIX OUTLINE
A. Methods of Issuing SPSS Commands 10. IF: Creating New Variables

1. Point-and-Click Commands 11. LIST: Listing Variables
2. Syntax Commands 12. MEANS: Comparing Means
3. Working with Output Viewer 13. PARTIAL CORRELATION: Testing for
an Intervening Variable
B. Basic SPSS Procedures
14. RECODE: Collapsing and Switching
1. ANOVA: Analysis of Variance Categories
2. COMPUTE: Creating New Variables 15. REGRESSION: Doing Multiple
3. CORRELATIONS: Correlational Analysis Regression Analysis
4. CROSSTABS: Cross-Tabular Analysis 16. RELIABILITY: Assessing Internal
5. DESCRIPTIVES: Computing the Mean Consistency
and Standard Deviation 17. SPEARMAN: Correlation for Ordinal
6. DISCRIMINANT: Doing Discriminant Data
Function Analysis 18. t TEST: Comparing Means
7. FACTOR: Doing Factor Analysis 19. Z SCORE: Using Z Scores to Create an
8. FREQUENCIES: How Many in Each Index
Category?
C. SPSS Manuals
9. GRAPH: Graphing Results
I
f you are just starting to work with SPSS, In addition to the instructions on how to
it would be a good idea to review the ma- begin an SPSS session included in Chapter
terials in Chapter 16 because it includes 16, you may wish to review Chapter 18,
an introduction to the Statistical Package which discusses four key multivariate pro-
for the Social Sciences (SPSS) and reviews cedures. Also, a number of manuals avail-
the procedures for logging onto the com- able from SPSS present many additional
puter and accessing an SPSS system le. features and options for the various pro-
This Appendix provides guidelines for the cedures. Some of these manuals are listed
basic procedures most likely to be used by at the end of Appendix A; you may wish to
the nursing researcher. Sample output is consult them as you develop your facility
annotated to assist you in using SPSS. in working with SPSS.
*SPSS is a trademark of SPSS, Inc. of Chicago, Illinois, for its proprietary computer software.
653
654 APPENDICES
A. METHODS OF ISSUING SPSS FACTOR: Analyze/Data Reduction/ Fac-

COMMANDS tor
SPSS procedures may be run using either
FREQUENCIES: Analyze/Descriptive
Statistics/Frequencies
a point-and-click method or by entering
syntax commands. Many users nd that
GRAPHS: Graphs/Scatter/Dene
both techniques are useful. Beginning
IF: File/New/Syntax (enter IF commands
on screen)
users typically nd the point-and-click
method easiest. We primarily focus on
LIST: File/New/Syntax (enter LIST com-
mands on screen)
this method here.
After each procedure is completed, the
MEANS: Analyze/Compare Means/means
results are sent to the Output Viewer that
PARTIAL CORRELATION: Analyze/ Cor-
relate/Partial
is automatically displayed on your moni-
tor; you can save, edit, and print your out-
RECODE: Tansform/Recode/Into Differ-
ent Variables
put from the Output Viewer. You can also
directly copy the output to a word proces-
REGRESSION: Analyze/Regression/Lin-
ear (MethodBackward)
sor le from the Output Viewer. We will
begin by examining the fundamentals of
RELIABILITY: Analyze/Scale/Reliabil-
ity (ModelAlpha)
submitting commands to SPSS and then
discuss how to work with output followed
SPEARMAN CORRELATION: Analyze/
Correlate/Bivariate/Spearman
by a presentation of the basic procedures
that have been covered in this text.
t-TEST GROUPS: Analyze/Compare
Means/Independent-Samples t Test
t-TEST PAIRS: Analyze/Compare Means/
1. Point-and-Click Commands Paired-Samples t Test
The toolbar across the top of the SPSS Z SCORE: Analyze/Descriptive Statis-
Data Editor screen provides the starting tics/Descriptives/(Save Standardized
point for issuing analysis commands to values as variables)
SPSS. One can draw samples, select cases, After a procedure has been selected,
sort cases, add or delete variables, and an SPSS procedure screen appears and
initiate a vast range of statistical proce- the user simply uses the mouse to select
dures by clicking on the appropriate tool- variables and move them into the appro-
bar item. The main statistical procedures priate analysis boxes shown on the Win-
are listed below. To access each of them, dow. A few general tips in using the
the user simply clicks an item on the tool- method are listed below:
bar and follows the sequence to identify
the desired procedure. The ones we focus To illustrate an SPSS procedure win-
on here are as follows: dow, the crosstabs one is shown in
Figure A1.
ANOVA: Analyze/General Linear Mo- To select one variable, simply click
del/Univariate on it.
COMPUTE: Transform/Compute To select multiple variables from a list,
CORRELATE: Analyze/Correlate/Bivari- hold down the Control button and
ate click on the variables.
CROSSTABS: Analyze/Descriptive Sta- To select several variables in a se-
tistics/Crosstabs quence, click on the rst one in the se-
DESCRIPTIVES: Analyze/Descriptive quence, move the mouse to the last
Statistics/Descriptives one, hold the Shift key down, and click
DISCRIMINANT: Analyze/Classify/Dis- on that item; all the variables between
criminant the two clicked on will then be high-
USING SPSS FOR WINDOWS 655
3
1
6 4 5
Notes:
1. The variables in the file are listed in this pane.
2. Move the dependent variable(s) to the Row(s) analysis pane
by highlighting the variable(s) and then clicking it over by
by clicking on the button.
3. Move the independent variable(s) to the Column(s) pane by
highlighting the variable(s) and then click on the to move it
over.
4. Click on Statistics to select the Statistics you want (probably
Chi-Square)
5. Click on Cells to indicate how you want percentages run
(choose column)
6. Click on Display clustered bar charts if you wish to have a
bar chart of the data
Figure A1 Crosstab Screen
lighted and may then be moved to an tistics boxes to see the additional pro-
analysis box by clicking the key. cedures that are available within each
To move quickly through a variables procedure.
list, click and hold down the scroll bar
button (to the right of the variables 2. Syntax Commands
list) and slowly drag down the scroll
bar button. To enter syntax commands, the user
Click on the various Options and Sta- brings up a Syntax Editor screen, types
656 APPENDICES
commands into the screen, and then sub- in your data, you can x them and then
mits the commands to SPSS for process- quickly re-run the analysis by simply re-
ing. To bring up a Syntax Editor screen, submitting the commands contained in
click on File/New/Syntax on the toolbar. the syntax le.
An SPSS Syntax box appears at the bottom If you wish to list several variables for
of the screen; if you click on it, the Syntax the rst 25 cases, it may be done using
Editor screen appears. You then type in the LIST command in a syntax le. This
SPSS commands into the screen and, procedure is available only as a syntax
when ready to submit, point and click on command and is useful when conrming
Run on the windows toolbar. See Figure that various RECODE, IF, or RECODE com-
A2 for a copy of the Syntax window. mands are working properly (see LIST
It is a good idea to retain a copy of the procedure).
syntax commands used when running an
analysis. When you exit from SPSS, you 3. Working with Output Viewer
will be asked if you wish to keep a copy of
the syntax le. If you respond yes, then The results of each analysis are sent to the
you will be prompted for a name for the Output Viewer. You can edit, print, and
le. These les use a le extension of .sps. save output from the Viewer. Figure A3
Another use of syntax les is that when shows output as it appears in the Output
using the point-and-click method, you Viewer for an analysis of suicide thoughts
can also click on Paste before you click by gender. The Output Viewer provides
on OK when running procedures; this two panes: (1) the outline pane, on the left,
pastes into a Syntax Editor le the syntax lists items in the output and you can click
commands for the procedure you have on these to move, save, or delete them;
just invoked. These syntax les are worth and (2) the contents pane, on the right
retaining because if you discover errors side, shows the output from the analysis.
Notes:
1. To get Syntax Editor Screen click: File/New/Syntax.
2. Enter SPSS procedure commands in screen.
3. When ready to submit, click on Run on the toolbar.
Figure A2 Syntax Editor Screen
Notes:
1. The left pane indicates elements; click on an element to have it highlighted.
2. The pane on the right displays the results of the analysis; to edit the results
(changing width of columns etc. click anywhere on output and then double
click on it; now you can change widths of columns etc.).
3. To add your table title click on title and then double click; type in new title.
4. To copy to another file; on left pane click on element(s) you wish to copy, and
then on toolbar click Edit/Copy Objects ; switch to file you wish to copy the
material into and do a Copy/Paste.
Figure A3 Output Viewer Screen
Some tips for using the Output Viewer To print one or more sections, click on
are: the items on the left pane and then
print the results (you can preview
To go directly to a section of the out- what will be printed by clicking on the
put, simply click on the name of the Preview icon).
section on the outline pane (the screen You can move objects around by click-
on the left side). ing and dragging to new locations on
To delete a section of output, simply your output.
click on the item on the left screen and You have various templates available
press the Delete button. You will gen- to choose the format of your output; to
erally want to get rid of the case pro- see options available, double Click on
cessing summary. the table (this activates the Pivot
658 APPENDICES
Table), click on Format, and click on B. BASIC SPSS PROCEDURES

Tablelooks
There are two basic procedures for exam-
A number of optional formats will be
ining individual variables. When you want
presented and you can choose from
a count of the number of cases that fall
among them.
into each category of a nominal or ordinal
You can highlight a section (or sec- variable, use the FREQUENCIES procedure
tions) on the left pane, Edit/Copy the described later. If you have a ratio level
material into the clipboard, and then variable and you wish to calculate the
bring up your word processing pro- mean and standard deviation, use the DE-
gram and Edit/Paste the material into SCRIPTIVES procedure described later.
the new le. When analyzing the relationship be-
If you wish to attach your name as a tween variables, there are many proce-
header for your output, go to File/Page dures that may be used. Beginning re-
Setup/Options when you are in the searchers are advised, however, to apply
Output Viewer and type in what you the 3M approach (model, measurement,
want in the header panel. Click on method) to determine which procedure is
Make Default and the header will ap- appropriate given the relationship being
pear on all your subsequent outputs. examined and the level of measurement
On lengthy outputs, you may nd that involved in the measures (see Chapter 16).
not all the output shows. Double click The following sections present the ba-
inside the box containing the output, sic procedures of SPSS. Each procedure is
move the cursor to the bottom of the illustrated with annotations on the sample
output, and drag down the bottom mar- output that should help the researcher un-
gin. When you do this, the remainder of derstand and interpret the results of the
the output will show on the screen. analysis. The procedures are listed alpha-
If you nd that there are unwanted page betically to make them easier to nd.
breaks in the output, double click in-
side the output (a thick border should
be showing). The little square box 1. ANOVA: Analysis of Variance
marks page breaks; you can delete
these using the Delete or Backspace Experimental researchers use analysis of
key. Recheck your preview to ensure variance (ANOVA) when their design has
that the unwanted page breaks are more than one independent variable or
gone before you print. has more than two treatment levels. (t
When you get errors using the Syntax tests are typically used in studies with
procedure, it is a good idea to read the fewer than 30 cases in each of the two
error carefully and then highlight and groups and when there is one treatment
delete the error message before you level and one independent variable.) The
resubmit the job. Note that you can inclusion of additional complexity is rela-
run point-and-click procedures with- tively simple and provides an opportunity
out returning to the Data Editor be- to explore the interaction of treatment
cause the analysis toolbar is at the top variables (when effects change for differ-
of the Output Viewer screen. ent combinations of treatment variables)
When you exit SPSS, you will be asked if in inuencing the dependent variable.
you wish to save your output viewer Although regression analysis is appro-
le. This le will have a .spo le exten- priate in many situations, there are sev-
sion. eral situations in which analysis of vari-
ance (ANOVA or MANOVA [multivariate variables for the various factors and any
analysis of variance]) would be more ap- covariates by clicking on them and then
propriate than regression analysis. For clicking them to their appropriate screens
example, if you had predictor variables by using the arrow keys; you will be
that have qualitative differences (e.g., prompted to identify the lowest and high-
three different drugs); or if rather than est values of treatment values such as 0, 2;
having a ratio variable measuring the in this case, the implied values include 0,
amount of time spent in the hospital, we 1, and 2. ANOVA assumes that there are no
have a variable that indicates whether blank categories, so you will have to RE-
the person was hospitalized or treated as CODE values if there are no values in one
an outpatient, we again would probably of the categories implied by the range in-
opt for analysis of variance. In short, if an dicated. Click OK when you are ready to
independent variable is made up of val- process the data.
ues that differ in kind rather than in quan-
tity, we would opt for analysis of variance. b. Sample Output: ANOVA
A second situation that is better han-
dled with analysis of variance is one in Note that the variations that are compared
which the relationship between the in- in an analysis of variance are composed of
dependent and the dependent variable two elements: random error and possible
changes over the continuum. Perhaps treatment effects. The test that is done
the relationship is nonlinear; once again, compares the ratio in the following way:
analysis of variance should be consid-
ered as an alternative to regression. Random error possible treatment effects
F
Typically, analysis of variance tech- Random error
niques should used when:
The interpretation of the results of an
An experimental design is being used. ANOVA test is split between a concern for:
The dependent variable (or variables)
The main effect of treatment variable A
is measured at the ratio level.
(Box A1 indicates that COUNTRY has a
One or more of the treatment variables
statistically signicant main effect on
are measured at the ratio level and oth-
egalitarianism).
ers at the nominal or ordinal levels.
You have multiple dependent measures
The main effect of treatment variable B
(Box A1 indicates that v50 [gender]
that you wish to examine simultane-
does not have a statistically signicant
ously (i.e., in MANOVA).
main effect on egalitarianism).
Ratio level independent variables may
An interaction of the treatment vari-
be identied as covariates.
ables (A B) (the interaction between
In nonexperimental designs, you wish
COUNTRY and GENDER is not statisti-
to examine whether there are signi-
cally signicant).
cant interactions among independent
variables.
An error term (within groups). This is
the residual source of variation.
a. Point-and-Click Method
2. COMPUTE: Creating New
Analyze Variables
General Linear Model
COMPUTE allows researchers to create
Univariate or Simple Factorial
new variables by performing mathemati-
Move the dependent variable and then the cal operations on variables.
660 APPENDICES
BOX A1 ANALYSIS OF VARIANCE: Sample Output
Between-Subjects Factors
Value Label N
NAME OF COUNTRY 0 CANADA 65
1 NEW ZEALAND 57
2 AUSTRALIA 60
GENDER 1 MALE 59
2 FEMALE 123
Tests of Between-Subjects Effects

Dependent Variable: EGALITARIANISM
9
9
Source Type III Sum of Squares df Mean Square F Sig.

Corrected Model 1575.426a 5 315.085 4.531 .001
Intercept 139602.183 1 139602.183 2007.419 .000
COUNTRY 1194.586 2 597.293 8.589 .000
V50 33.219 1 33.219 .478 .490
COUNTRY * V50 125.990 2 62.995 .906 .406
Error 12239.591 176 69.543
Total 186677.000 182
Corrected Total 13815.016 181
a. R Squared .114 (Adjusted R Squared .089)
Notes 3. Test of significance results: needs to be 0.05

1. Degrees of freedom column. or less to be statistically significant.
2. F test of significance value.
a. Point-and-Click Method see a window that allows you to place a

variable label on the new variable); then
Transform
move the cursor to the Numeric Expres-
Compute
sion window and type in the computa-
As illustrated in Figure A4, enter the name tional formula you wish to use. Enter OK,
of the new variable you wish to create in and the new variable will be created.
the Target Variable window (you will now Figure A4 shows the screen illustrating
2
1
3
Notes:
1. Name new variable here
2. Type in numerical expression here
3. Click OK when done
Figure A4 Compute Screen
the creation of a new variable, famsize, SUM(var list) sums values in vari-
representing the addition of values in two able list
variables (i.e., number of brothers plus MEAN(var list) mean of values
number of sisters plus 1 added to the to-
tal). Computations within parentheses
are performed rst and are used to con- 3. CORRELATIONS:
trol the order of the mathematical opera- Correlational Analysis
tions. Make certain that there are an equal
When variables are measured at the ratio
number of open and closed parentheses.
level, then various correlational tech-
SPSS uses the following symbols to indi-
niques are appropriate. As discussed in
cate basic functions:
Chapter 11, a correlation, or r, measures
Addition the strength of an association between
Subtraction two variables. The values can range from
* Multiplication 1.00 to 1.00. Typically, they are re-
/ Division ported to two decimal places, as in 0.56.
** Exponentiation Explained variance is the square of the
correlation coefcient or r 2.
Many other functions are available to
the researcher in SPSS; three commonly
used ones include: a. Point-and-Click Method
RND(var) rounds to whole number Analyze

662 APPENDICES
Correlate CORRELATION section to see how that

Bivariate type of correlation is run.
The correlation screen will appear. Click

on and move variables to the Variable 4. CROSSTABS: Crosstabular
window. For ratio variables, you should Analysis
use Pearson correlations; however, if you
are using ordinal variables, you should CROSSTABS is a procedure used to ex-
use Spearman correlations; click on the amine the association between a nominal
appropriate box. You may also specify or ordinal dependent variable and a nom-
one- or two-tailed tests; click OK when inal, ordinal, or ratio independent vari-
everything is set appropriately. able. To have sufcient cases in each of
Box A2 includes a summary table of the table cells when there are several cat-
correlations. The CORRELATIONS proce- egories in the independent variable, they
dure may also be used along with the PAR- would be recoded ( i.e., a new variable
TIAL CORRELATION procedure to test for would be created regrouping the cate-
intervening variables and for sources of gories into two or three meaningful cate-
spuriousness variables. See the PARTIAL gories) before the analysis is run.
BOX A2 CORRELATIONS: Sample Output
Correlations
- -
More -
Money
- --
Abortion
Government Should be Homo- -- Womans
Increase -- Personal
Given to Sexuality
Foreign Aid Welfare --
Okay Choice
9 - -
GOVERNMENT Pearson Correlation 1.000 .323** .029 .023 -
INCREASE Sig. (1-tailed) . .000 .347 .379 --
FOREIGN AID N 184 183 182 183
9
MORE MONEY Pearson Correlation .323** 1.000 .272** .234**

SHOULD BE GIVEN Sig. (1-tailed) .000 . .000 .001
TO WELFARE N 183 184 183 184
HOMOSEXUALITY Pearson Correlation .029 .272** 1.000 .435**

OKAY Sig. (1-tailed) .347 .000 . .000
N 182 183 183 183
ABORTION WOMANS Pearson Correlation .023 .234** .435** 1.000

PERSONAL CHOICE Sig. (1-tailed) .379 .001 .000 .
N 183 184 183 184
**. Correlation is signicant at the 0.01 level (1-tailed).

9--------
Notes 2. One-tailed test requested (** indicates sig-

1. Pearson correlation is the r value. nificance at the 0.01 level).
3. N number of cases.

a. Point-and-Click Method necessary to minimize the number of cat-

egories in the independent and in the con-
Analyze trol variables. Generally, there should be
Summarize no more than two or three categories
Crosstabs within these variables. There are two ma-
The CROSSTABS screen will now appear. jor reasons for this limitation: (1) the num-
ber of cases in each cell will become too
Place the dependent variable (or vari- small if there are many categories in either
ables) in the Row(s) window. the independent or the control variable,
Place the independent variable (or and (2) the interpretation of the table is
variables) in the Column(s) window. very difcult if simplicity is not main-
Click on Statistics, on Chi-Square, and tained.
then on Continue. Chapter 17 discusses interpretations
Click on Cells and click on Column Per- of three-variable cross-tabular tables. Re-
centages. Then click Continue. member that to apply Jacksons rule of
Click OK. thirds, it is necessary to compare the orig-
inal difference in the percentages be-
b. Sample Output and Notes tween the categories of the independent
variable and re-run the relationship, con-
Box A3 presents the results from a simple trolling for the intervening (or source of
CROSSTABS analysis. Note where the chi- spuriousness) variable. Only if the origi-
square value is printed and the probability nal difference is reduced by more than
is identied. Usually CROSSTABS results one-third in each category of the control
are reformatted when prepared for pre- variable do we have any evidence consis-
sentation in a paper. See Table A1 for sug- tent with the proposed model.
gestions on how to reformat a CROSSTABS
table for presentation in a paper.
Table A1 presents the relationship be- 5. DESCRIPTIVES: Computing
tween gender and suicidal thoughts. Be- the Mean and Standard
cause we are attempting to assess the Deviation
impact of the independent variable on
the dependent one, we are interested in The DESCRIPTIVES procedure computes
the percent of respondents who have means and standard deviations and is ap-
thoughts of suicide. Thus, we might de- propriately used for ratio level variables.
scribe the ndings as indicating that The output provides the mean, standard
while 23.3 percent of the male respon- deviation, and the minimum and maximum
dents report having had thoughts of sui- values for the variable (or variables) iden-
cide, this was true for 39.5 percent of the tied on the variable list. In addition, the
female respondents. In short, we com- number of valid (nonmissing) cases is re-
pare percentages in each column. Usually ported (Box A5 provides a sample output.)
it is sufcient to use one row ( in this case,
just the row for those reporting having
had thoughts of suicide). a. Point-and-Click Method
For analyses involving three or more
variables simultaneously, the procedures
Analyze
are the same, with the control variable
identied in the Layer 1 of 1 window (Box
Descriptives
A4 provides a sample output). The DESCRIPTIVES screen will now ap-
When control variables are used, it is pear. Identify the variables for which you
664 APPENDICES
BOX A3 CROSSTABS: Sample Output

----
---
THOUGHTS ABOUT SUICIDE * GENDER Crosstabulation -- ---
GENDER
9-
MALE FEMALE Total
-----
THOUGHTS ABOUT YES Count 14 49 9---63---
9- SUICIDE % within GENDER 23.3% 9-- 39.5% 34.2%
-- ----
----
NO Count 46 75 121 ---
--
% within GENDER 76.7% 60.5% 65.8%
Total Count 60 124 184
% within GENDER 100.0% 100.0% 100.0%
Chi-Square Tests ---------------------------------

( Asymp. Sig. Exact Sig. Exact Sig.
Value df (2-sided) (2-sided) (1-sided)
Pearson Chi-Square 4.703b 1 .030
Continuity Correctiona 4.012 9- ----------------------------

1 --- .045
Likelihood Ratio 4.879 1 .027

- --
------
Fishers Exact Test .032 9----.021
Linear-by-Linear Association 4.677 1 .031
N of Valid Cases 184
a. Computed only for a 22 table

b. 0 cells (.0%) have expected count less than 5. The minimum expected count is 20.54.
Notes square value; if not a 2 2 table, use the

1. The dependent variable is suicidal thoughts. Pearson chi-square value.
2. The independent variable is gender. 7. Use this value to test null hypothesis (no re-
3. Number of females who have thought about lationship between suicide thoughts and gen-
suicide. der).
4. Compare these percentages when discussing 8. Because the value is 0.05, the decision is
results. to reject the null hypothesis; there is a sta-
5. df degrees of freedom. tistically significant relationship between sui-
6. With 2 2 tables, use this value as the chi- cidal thoughts and gender
want DESCRIPTIVES by moving desired sired, click Save Standardized. SPSS will
variables to the Variable window. Click create a standardized version of the vari-
OK when ready. able and will add a Z to the beginning of
If Optional Statistics are desired, click the variable name (e.g., v24 will become
on Options. If a standardized (Z score) zv24). The new variable is immediately
version of the variable (or variables) is de- available for use.
Table A1 Suicide Thoughts by Gender

GENDER
Male Female Total
Thoughts About Suicide N % N % N %
Yes 14 23.3 49 39.5 63 34.2
No 46 76.7 75 60.5 121 65.8
Total 60 100.0 124 100.0 184 100.0
Chi-square 4.012 (corrected for continuity); degrees of freedom 1; Sig. .032 (2 tailed test)
BOX A4 CROSSTABS: Sample Output Showing a Control Variable
THOUGHTS ABOUT SUICIDE * GENDER * NAME OF COUNTRY Crosstabulation

GENDER
9----- NAME OF
-
COUNTRY MALE FEMALE Total
CANADA THOUGHTS ABOUT YES Count 8 14 22
SUICIDE % within GENDER 26.7% 40.0% 33.8%
NO Count 22 21 43
% within GENDER 73.3% 60.0% 66.2%
% within GENDER 22.2% 40.0% 34.5%
NEW THOUGHTS ABOUT YES Count 4 16 20

ZEALAND SUICIDE % within GENDER 22.2% 40.0% 34.5%
NO Count 14 24 38
% within GENDER 77.8% 60.0% 65.5%
% within GENDER 100.0% 100.0% 100.0%
AUSTRALIA THOUGHTS ABOUT YES Count 2 19 21

SUICIDE % within GENDER 16.7% 38.8% 34.4%
NO Count 10 30 40
% within GENDER 83.3% 61.2% 65.6%
% within GENDER 100.0% 100.0% 100.0%
Notes and females in each country. Have the dif-

1. Categories of the control variable (country). ferences (compared with those shown in
2. To test if country of origin is spuriously in- Box A3) increased, stayed the same, de-
fluencing relationship between gender and creased or disappeared, or are they mixed?
suicidal thoughts, compare the percentage See Chapter 17 for details about interpret-
difference in suicide thoughts between males ing this output
666 APPENDICES
BOX A5 DESCRIPTIVES: Sample Output, Including Z Score Values

--- ----
----
----
N Minimum Maximum Mean 9-- Std. Deviation
EGALITARIANISM 182 10.00 54.00 30.8187 8.7365 9--------

- - -
-
Zscore: ---
EGALITARIANISM 182 2.38296 2.65339 2.9e-16 9 1.0000000 9
----
----
Valid N (listwise) 182 ----
--
Notes 3. The 2.9E-16 indicates the mean of the
1. Shows the mean value for an egalitarianism scores when converted to Z scores; the
index score. value is 0.0000000000000002.9, which is
2. The standard deviation is shown. very close to zero!
b. Sample Output: DESCRIPTIVES gression analysis, standardized weight-

ings are provided that permit the re-
Box A5 shows the output you can expect searcher to assess the relative impact of
from a DESCRIPTIVES analysis. See Chap- independent variables on the dependent
ter 19 for suggestions about how to com- one. Also, similar to regression analysis, it
press the reporting of means and stan- is possible to use dummy independent
dard deviations for a number of variables. variables. See examples of how to con-
struct these in the Regression Analysis
section of this Appendix. For more details
6. DISCRIMINANT: Doing
on discriminant analysis, see Chapter 18.
Discriminant Function Analysis
Discriminant function analysis is a proce- a. Point-and-Click Method

dure for examining the relationship be-
tween a nominal dependent variable and Analyze
ratio independent variables. The basic Classify
idea behind discriminant function analy- Discriminant
sis is to correctly classify into which cate-
The Discriminant screen will now appear.
gory of the dependent variable each case
will fall. It can also be used to predict Click on dependent variable and click
which category a case will fall into when to move to Grouping Variable win-
measures of the dependent variable are dow.
missing. The statistics Lambda was dis- Click on Dene Range to specify mini-
cussed in Chapter 11. Readers should re- mum and maximum values for the de-
call that this statistic measures the pro- pendent variable (hit the Tab key after
portionate reduction in error that is each entry).
achieved when a variable is added to an Click on Continue.
analysis. Lambda is used in discriminant Click on Independent variables and
function analysis to measure the success click to move them to the Indepen-
in correctly classifying cases. Similar to re- dent variables window.
Click on Analyze and then click on Square the proportions and add them
Means. together and then multiply by 100
Click on Continue. (0.362 0.642) 100 53.9 %.
Click on Classify and then click on Sum- Proportion of errors in random assign-
mary Table. ment (1 0.54 0.46).
Click on Continue. Proportion of errors using discrimi-
Click OK. nant analysis based classications (1
Identify variables for which you want 0.61 0.39).
procedure run; click OK.
The percent error reduction using the
model compared with random assign-
b. Sample Output and Notes ment may be calculated by comparing
the proportion of errors in model com-
Box A6 presents sample output from a dis- pared with random assignment divided
criminant analysis of suicidal thoughts by the proportion of errors using the
(v12) among respondents in an interna- model, multiplied by 100: [(0.46 0.39)
tional survey. Selected for independent 0.46] 100 15.2 percent.
variables were FAMILY SES (socioeconomic The results indicate that we were able
status), v50 (gender, a dummy variable), to improve our classication of people
v58 (size of home community), and EGAL into the thought-of-suicide category ver-
(an index measuring egalitarianism). sus the have-not-thought-of-suicide cate-
The output shown in Box A6 shows how gory by 15.2 percent by using the informa-
many cases were used in the analysis and tion on the four independent variables.
the number of respondents who reported But how important were each of the inde-
that they had thought about committing pendent variables?
suicide (n 60) and the number who re- There are two basic ways of estimating
ported that they had not had such the relative importance of variables in cor-
thoughts (n 108). The mean scores for rectly estimating into which category of
each independent variable are then re- the dependent variable that a case will fall:
ported for each category of suicidal
thoughts. Note that those with suicide 1. Use the DISCRIMINANT procedure, in-
thoughts had higher mean scores on egal- cluding all variables on the rst run.
itarianism, family SES, gender, and size of Then continue to repeat the analysis
home community. but exclude a different independent
The last presentation on the output variable each time until you have an
shows the percent of cases correctly clas- analysis excluding each independent
sied. It is reported that 61.3 percent of the variable. The contribution of each in-
cases were correctly classied using the dependent variable may then be
four independent variables. But how might judged by comparing the percentage
we have fared if we had just randomly as- of correct assignment using the vari-
signed cases to the suicidal thought or able with when it is not used, as in:
nonsuicidal thought categories? To com-
pute the number of correct classications Contribution (% correct with
variable included) (% correct
we would have made using random as-
without variable included)
signment, the following steps may be
taken:
The variable that produces the greatest
Compute the proportion of cases that decrease in correct assignments when
fall into each category ( in this case, 60 omitted is the most important factor,
168 0.36; 108 168 0.64). the one with the second greatest drop
668 APPENDICES
BOX A6 DISCRIMINANT ANALYSIS: Sample Output

--- ---
- ----
Group Statistics
- - - ----
--- Valid N (listwise)
-- ----
THOUGHTS
ABOUT SUICIDE Mean
9--
Std. Deviation Unweighted Weighted
YES EGALITARIANISM 32.6833 8.9962 60 60.000
FAMILY SES 68.4190 21.4355 60 60.000
SIZE OF HOME COMMUNITY 5.2500 2.8560 60 60.000
GENDER (Dummy) .7667 .4265 60 60.000
NO EGALITARIANISM 29.0926 8.5002 108 108.000

FAMILY SES 63.3498 22.6049 108 108.000
GENDER (Dummy) .6204 .4876 108 108.000
Total EGALITARIANISM 30.3750 8.8241 168 168.000

FAMILY SES 65.1602 22.2634 168 168.000
GENDER (Dummy) .6726 .4707 168 168.000
Analysis 1: Summary of Canonical

Discriminant Functions
Eigenvalues
Function Eigenvalue % of Variance Cumulative % Canonical Correlation
1 .091a 100.0 100.0 .289
aFirst 1 canonical discriminant functions were used in the analysis.
Wilks Lambda
Test of Function(s) Wilks Lambda Chi-square df Sig.
1 .916 14.307 4 .006
Standardized Canonical Discriminant Function Coefcients

Function
1
EGALITARIANISM .743 9-
---
FAMILY SES .528 9-------
--
.146 9- - - - ---
SIZE OF HOME COMMUNITY
- -
- ---
GENDER (Dummy) .591 9

BOX A6 DISCRIMINANT ANALYSIS: Sample Output (Continued )
Structure Matrix
Function
1
EGALITARIANISM .660
GENDER (Dummy) .500
FAMILY SES .365
SIZE OF HOME COMMUNITY .142
Pooled within-grouped correlations between discriminating variables and standardized canonical

discriminant functions.
Variables ordered by absolute size of correlation within function.
Functions at Group Centroids

Function
THOUGHTS ABOUT SUICIDE 1
YES .403
NO .225
Unstandardized canonical discriminant functions evaluated at group means.
Classication Statistics
Prior Probalitities for Groups

Cases Used in Analysis
THOUGHTS ABOUT SUICIDE Prior Unweighted Weighted
YES .500 60 60.000
NO .500 108 108.000
Total 1.000 168 168.000
Classification Resultsa
Predicted Group
THOUGHTS ABOUT Membership
SUICIDE YES NO Total
Original Count YES 37 23 60

NO 42 66 9-----108
---- -----
% YES 61.7 38.3 100.0
NO 38.9 61.1 100.0
a61.3% of original grouped cases correctly classied.
Notes sured by the standardized values presented

1. The mean values for each independent vari- here. Hence the egalitarianism score of .743
able is presented for both categories of the is the most important item.
dependent variable 3. The 103 cases on this diagonal were cor-
2. The relative importance of the independent rectly classified; the other diagonal contains
variable in helping to classify the date into 65 incorrectly classified cases.
the suicide thoughts categories may be mea-
670 APPENDICES
is the second most important variable, ability of the difference occurring on a

and so forth. chance basis (Sig.). In our illustration, the
2. Another method to assess the relative signicance level is 0.006, which means
importance of factors may be derived that because of chance factors in sampling,
by examining the standardized discrim- that much of a difference could only be ob-
inant function coefcients. These coef- served six times in a thousand samples.
cients are similar to the beta weights in The number of functions that the DIS-
regression analysis in that they are CRIMINANT procedure will calculate is
standardized (so that differences in one fewer than the number of categories
the range and variability of values in a in the dependent variable. Thus, as is the
variable does not inuence the coef- case in the example we are using, when
cient; if it helps, think of these as Z there are two categories in the depen-
scores in which each variable has a dent variable, one function will be com-
mean of 0 and a standard deviation of puted; if we had three categories, two
1). The larger the coefcient, the more functions would be calculated.
importance the variable has in pre- As in regression analysis, if there is a
dicting into which group the individual high correlation among the independent
will fall. The relative importance of variables (e.g., above 0.70), then it is a
each variable is directly proportional good idea to either drop one of the highly
to its coefcient and may be estimated correlated variables or multiply them to-
in a manner similar to that suggested gether to provide a composite measure of
for the beta weights in regression the two variables.
analysis (see Regression section).
7. FACTOR: Doing Factor
Box A6 should be examined. First, it Analysis
should be noted that the total amount of
the variation explained by the combina- Factor analysis is a procedure used to ex-
tion of variables can be determined by plore communalities among a number of
squaring the canonical correlation. In our variables (see Chapter 18 for a comparison
example, the canonical correlation is to regression analysis). The typical uses in
0.289, which means that together the four nursing include instrument development,
independent variables are accounting for instrument validation, and data reduction.
8.4 percent of the variance in suicidal Simply stated, factor analysis is a tool used
thoughts (0.2892 0.0835 100 8.4%). to identify groups of variables that cluster
The eigenvalue (0.091) represents the together within a data set. For example, if
ratio of the between-groups variance to the a researcher had 20 Likert items measur-
within-groups variance. The higher this ing nurses job satisfaction, factor analysis
value, the greater the ability of the inde- might be used to see if the various indica-
pendent variables to discriminate between tors fall into three or four groupings. One
categories of the dependent variable. factor might be job autonomy, another
The Lambda value varies from 0 to 1.0 might be feeling a part of the team, and
and is a direct measure of the discriminat- so forth. To do a factor analysis in SPSS, ac-
ing power of the model being testedthat cess the procedure as follows:
is, the higher the value, the higher the dis-
criminating ability of the model to distin- a. Point-and-Click Method
guish between categories of the dependent
variable. In this case, the Lambda is 0.916. Analyze
A chi-square value is reported along with Data Reduction
the degrees of freedom (df ) and the prob- Factor
The Factor Analysis screen will now ap- cluded in the analysis and appears to
pear. be the most commonly used approach.
Some of the terms that need to be un-
Box A7 illustrates some of the output
derstood in reading factor analysis out-
that results from a factor analysis. In this
put include the following key terms:
case, two signicant factors were ex-
tracted from the communalities among
An eigenvalue is the sum of the squared
the eight variables purporting to measure
factor loadings for any one factor. Thus,
liberalism. By examining these factors, we
the eigenvalue is a measure of how
note that factor 1 seems to include those
much of the total variance is explained
items related to public policy issues, so
by a given factor. The higher the eigen-
we labeled the factor Public Policy Liber-
value relative to other ones, the greater
alism. The second factor included two
the variance explained by that factor.
items with heavy loadings (abortion is a
A correlation matrix is said to be fac-
womans personal choice and homosex-
torable if there are reasonably robust
uality is okay). This factor we identied
correlations among the items. Ordi-
as Individual Rights Liberalism.
narily, one would expect the interitem
It is interesting to compare the results
correlations to average above 0.25 to
with a RELIABILITY analysis of the same
consider the variables to be worthy of
items. The results are the same and the
factor analysis. If it is any lower than
researcher would end up creating two in-
that, and you may not nd that any fac-
dexes, one for the public policy liberal-
tors would be extracted.
ism and one for the individual rights lib-
A marker variable is one that pro-
eralism.
duces the highest correlation with the
underlying factor that is identied in
the analysis. The marker variable is 8. FREQUENCIES: How Many
useful to the researcher in attempting Cases in Each Category?
to label the underlying factor.
Nominal and ordinal variables are exam-
ined using the FREQUENCIES procedure.
In determining the second factor, a ro-
This procedure provides a count of the
tation is required, either (1) an orthogonal
number of cases falling into each cate-
(varimax is the most common version of
gory. This procedure is also used to
orthogonal rotations) forces a 90 degree
check the distribution of a variable be-
rotation, producing an entirely indepen-
fore recoding it. FREQUENCIES can also
dent solution or (2) an oblique rotation
produce bar charts and histograms. Click
that selects the second best solution but
on Options to select whether you want
does not insist on the 90 degree rotation.
bar charts, pie charts, histograms, or
There are two main ways in which fac-
none (the default). Various statistics are
tors are extracted:
available on request. By double clicking
on the output, various changes may be
1. The principal components method,
made to the various charts (Box A8 pro-
which includes an analysis of all the
vides a sample output.)
variance in the observed variables, not
just the shared variance among the
variables. a. Point-and-Click Method
2. The principal factors method (known
in SPSS as the Principal Axis Factoring Analyze
method). This approach relies on the Descriptive Statistics
shared variances of the variables in- Frequencies
672 APPENDICES
The Frequencies screen will now appear. Variables window by clicking on the
Click on the variables for which you want button on the menu. When you have the
frequencies run and move them to the variables you want listed, click OK. The
BOX A7 Factor Analysis of Liberalism Measures: Sample Output
Total Variance Explained

Extraction Sums of Rotation Sums of
Initial Eigenvalues Squared Loadings Squared Loadings
% of Cumulative % of Cumulative % of Cumu-
Factor Total Variance % Total Variance % Total Variance lative %
9--- 1 2.471 35.295 35.295 1.842 26.320 26.320 1.622 23.165 23.165

2 1.405 20.075 55.370 .839 11.993 38.313 1.060 15.148 38.313
3 .801 11.449 66.819
4 .650 9.283 76.102
5 .642 9.167 85.269
6 .549 7.841 93.109
7 .482 6.891 100.000
Extraction Method: Principal Axis Factoring.
Factor Matrixa
Factor
Factor 1: Public Policy Liberalism 1 2
GOVERNMENT INCREASE FOREIGN AID .546 .360
INCREASE OF VISIBLE MINORITIES IN JOBS .535 .179
VISIBLE MINORITY NOT GIVEN EQ EDUCATION .550 4.76E-02
IMMIGRATION POLICIES TOO STRICT .513 .243

9
--
ABORTION WOMANS PERSONAL CHOICE .310 --
--.610
--
HOMOSEXUALITY OKAY .438
-
.490 - --
--
--
MORE MONEY SHOULD BE GIVEN TO WELFARE .636 9----6.161E-02 -
-------- -
---
a2 factors extracted. 16 iterations required.

BOX A7 Factor Analysis of Liberalism Measures: Sample Output

(Continued )
Rotated Factor Matrixa

Factor
Factor 2: Individualism Rights Liberalism
1 2
GOVERNMENT INCREASE FOREIGN AID .651 26.20E-02
INCREASE OF VISIBLE MINORITIES IN JOBS .556 9.243E-02
VISIBLE MINORITY NOT GIVEN EQ EDUCATION .508 .216
IMMIGRATION POLICIES TOO STRICT .567 2.612E-02
ABORTION WOMANS PERSONAL CHOICE 1.20E-02 .684

--
9----
HOMOSEXUALITY OKAY .157 .639
MORE MONEY SHOULD BE GIVEN TO WELFARE .533 .353

Rotation Method: Varimax with Kaiser Normalization.
aRotation converged in 3 iterations.
Notes 2. The loadings on factor 1 indicate that all but

1. Two significant factors are identified in the two of the items have factor loadings above 0.5.
analysis of the eight variables. 3. On factor 2, two items predominate: abor-
tion should be a womans personal choice
and homosexuality is okay.
output will be sent to the Output Viewer items draw attention to key points on the
and displayed on your monitor. output.
b. Sample Output: FREQUENCIES 9. GRAPH: Graphing Results

It is often preferable to present nominal There are a variety of graphing procedures
or ordinal data in summary tables. Table available in SPSS, including bar charts, line
19.3 illustrates such a table for nominal charts, pie charts, scatterplots, and his-
or ordinal data, which is based on infor- tograms, among others (see Figure A9).
mation generated using the FREQUEN- The SCATTER procedure enables the re-
CIES procedure. In summary tables, it is searcher to produce a scatterplot of the re-
useful to report both the numbers and lationship between two variables. The re-
the percentages. searcher may also designate various labels
Box A8 shows a sample output from a to be placed on the output as well as spec-
FREQUENCIES procedure. The circled ify the scale for each dimension of the plot.
674 APPENDICES
.
.
.
Box A9. propriate for the type of graph you are

creating. Then click OK.
Scatterplot: Mother's Occupation
& Family Socioeconomic Status
b. Sample Output and Notes
The sample output shows the results of a

scatterplot of the relationship between
two variables, with the dependent vari-
able on the vertical axis. Various options
are available to permit researchers to
control the intervals printed along the
two dimensions and to print various sta-
tistics. By double clicking on the output
in the Output Viewer, a nurse researcher
Bar Chart: Father's Education can alter various settings on the charts,
including such things as the size of the
units used for scaling, colors, the format
of bars, and so forth. Consult an SPSS
manual for other graphics options.
10. IF: Creating New Variables
A third way new variables can be created

is through IF statements. The simple form
of the command (using Syntax editor) is
as follows:
Pie Chart: Father's Education COMPUTE newvar 0.

IF (v15 LT 25) newvar 1. (See footnote
to Box A10 for a list of abbreviations
used.)
Here the new variable is initially set to
0. Then whenever v15 has a value less
than 25, NEWVAR is set to 1.
More complex uses of the IF statement
are illustrated in the following example.
Suppose we are studying poverty in a
community and we have survey data on
300 people over the age of 18 years. We
wish to create a variable identifying dif-
a. Point-and-Click Method ferent types of poverty, taking into ac-
count the variables known as age and in-
Graphs come. The following represents one way
Bar/Line/Area/Pie/High-Low/Scatter/ of creating the new variable:
Histogram
COMPUTE poortype 0
Dene
IF (age LT 25 AND income LT 20000)
Identify the variable (or variables) as ap- poortype 1
676 APPENDICES
BOX A10 Creating an Occupational Prestige Index Using IF

Statements
The following syntax commands were used to or house husband, the score is based on the
create a family socioeconomic index using: (1) employed persons occupational prestige. The
the higher of the mothers and fathers occupa- original data used 99 as a missing value and 98
tional prestige score plus (2) one third of the for housewife or house husband. The com-
lower spouses score. In cases of a housewife ments on the right side explain the commands.
Syntax Commands Comments
des var v63 v64. V63 fathers occupation
missing values v63 v64(). V64 mothers occupation
des var v63 v64. removes missing values; checks
if (v64 ge v63 and v64 lt 98) ses v64. values before and after removal
if (v63 ge v64 and v63 lt 98) ses v63. uses mothers score if higher
if (v64 ge 98 and v63 lt 98) ses v63. than fathers, etc.
if (v63 ge 98 and v64 lt 98) ses 64. if one spouse not in work force
list var id v63 v64 ses/cases 25. other spouses score is used;
des var v63 v64 ses. ses score computed
compute ses1 ses. to add 1/3 of lower persons
var labels ses1 SES .33 of employed spouse. score the following is used:
if (v64 ge v63 and v64 lt 98) ses1 (ses1 (v63 * .33)). adds 1/3 of fathers score
if (v63 gt v64 and v63 lt 98) ses1 (ses1 (v64 * .33)). adds 1/3 of mothers score
missing values v63,v64(98,99). puts missing values back on
des var v63 v64 ses ses1. computes new means lists 25
list var v63 v64 ses ses1/cases 25. cases to confirm data
SPSS Abbreviations
gt greater than eq equal to

ge greater than or equal to ne not equal to
lt less than and both conditions must be met
le less than or equal to or either condition must be met
IF (age GE 25 AND age LT 65 AND in- the young and old and those with more
come LT 25000) poortype 2 than $25,000 for the middle-aged respon-
IF (age GE 65 AND income LT 20000) dents) and three categories of the poor
poortype 3 (young respondents with incomes under
VARIABLE LABELS poortype Type of $20,000; middle-aged respondents with in-
Poor. comes under $25,000; and senior respon-
VALUE LABELS poortype 0 all ages, dents with incomes under $20,000).
nonpoor. Another illustration of the IF statement
1 under 25 under $20 K is the creation of an SES index that identi-
2 2564, under $25 K es for each respondent the higher of the
3 over 64, under $20 K mothers or fathers occupational prestige
rating. On the assumption that information
The above set combines age and income has been collected on both the mothers
levels to create a new variable called and fathers occupations and that the pres-
POORTYPE, which identies the nonpoor tige scores for the occupations have been
(those with incomes more than $20,000 for entered into the computer, our task is to
develop a new variable that will use the been properly executed. LIST is available
higher of the mothers or fathers occupa- through a syntax command, as follows:
tional prestige rating to reect the SES
level of the family unit. The challenge here
File
is to have SPSS compute the index. Box A10
New
shows how the new variable could be cre-
Syntax
ated using SPSS. The blank syntax screen should now ap-
Examine Box A10 carefully. Because pear on your monitor. Use the following
housewife (coded as 98) and those who command format to initate a listing of
did not answer the question (coded as cases for each specied variable:
99) were both identied as missing val-
list var id v1 v1r/cases = 30
ues, it was necessary to remove the miss-
ing values identiers before the index On the toolbar click on Run and click on ei-
could be created. Through the use of IF ther All or Selection (if you have high-
statements, SPSS was then instructed to lighted the LIST command line). SPSS will
use the higher of the mothers or fathers then list the values for the specied vari-
occupational prestige score to determine ables (id, v1, v1r) for the rst 30 cases in
the value for the new variable ( i.e., SES). the le. The researcher can then check to
In this index, we have combined several see that the changes made in v1r are cor-
elements: rect. If the cases specication is omitted,
all the cases will be listed for the variables
Selected the higher of the mothers and
identied on the variable list.
fathers occupational prestige score
If either is not in the labor force, the
score for the one who is is used to in-
dicate the occupational prestige 12. MEANS: Comparing Means
If both are in the labor force, the higher
The MEANS procedure is used to com-
occupational prestige score is used
pare the mean values of a dependent vari-
and then one third of the other per-
able across categories on an independent
sons occupational prestige is added in
variable. The test of signicance associ-
to derive the Family SES score
ated with the test is a one-way analysis of
It is crucial in creating more complex variance.
variables to list a number of cases to be
certain that the values are being properly
assigned to the new variable. After the IF a. Point-and-Click Method
statements have been completed, it is nec-
essary to have SPSS do some procedure
Analyze
before the missing values for v63 and v64
Compare Means
(the two occupational prestige scores) are
Means
reassigned. The syntax commands are il- Identify the dependent variables for which
lustrated in boldface in Box A10. you want MEANS procedure run. Click
them over to the Dependent List window
by pressing the button. Identify the in-
11. LIST: Listing Variables dependent variable list and click those
over to the Independent List window.
After any computation creating or modi- Check the Options window to select any
fying a variable is completed it is a good statistics you want; click them over to the
idea to list some of the cases out so that Cell Statistics. When done, click Continue
you can conrm that the changes have and then OK.
678 APPENDICES
b. Sample Output and Notes each category of the independent vari-

able. Table A2 presents the kind of data
Box A11 presents the results resulting that would be appropriate for this kind of
from the use of the MEANS procedure. analysis. Note that the dependent vari-
Note that the average values for all re- able (income) is ratio level one and the in-
spondents on the egalitarianism index dependent variable is nominal (gender).
are presented for each country. The categories of the table should be
When you have a dependent variable carefully labeled. The means and stan-
measured at the ratio level and either a dard deviations of the dependent vari-
nominal or ordinal independent variable, able along with the number of cases are
then it is appropriate to compute the normally reported in columns across the
mean values of the dependent variable for table (Table A2 provides an example).
BOX A11 MEANS ANALYSIS: Sample Output
Report
EGALITARIANISM
COUNTRY Mean N Std. Deviation
Canada, New Zealand 32.4000 125 8.4204
Australia 27.3509 57 8.4779
Total 30.8187 182 8.7365
ANOVA Table
Sum of Mean
Squares df Square F Sig.
-
EGALITARIANISM*
- - ----
Between Groups (Combined) 998.034 1 998.034 14.016 9 .000 9---
Canada, New Zealand, Australia 12816.982 180 71.205

Within Groups
Total 13815.016 181
Measures of Association
Eta Eta Squared
EGALITARIANISM*
Canada, New Zealand, Australia .269 .072
Notes than 0.05, we reject the null hypothesis; there

1. The F statistic here. is a statistically significant difference between
2. The significance value here; because it is less the countries in their level of egalitarianism.
Table A2 Mean Income by Gender

Gender Mean Income, dollars Standard Deviation Cases, n
Male 37,052 14,707 142
Female 34,706 11,693 37
Combined mean 36,567 13,474 179
If appropriate, test of signicance values entered here.
A summary table may be used to re- 13. PARTIAL CORRELATION:

port the relationship between one depen- Testing for an Intervening
dent variable and a series of independent Variable
variables. A sample of such a table is pre-
sented in Figure 19.6. The computations for testing for interven-
In cases in which there are many cate- ing or sources of spuriousness models
gories in the independent variable, they when you have ratio level variables involve
will have to be regrouped into two or three a combination of CORRELATIONS and
before the analysis is run (click on the PARTIAL CORRELATIONS. Box A13 shows
Transform/Recode/Into Different Variable how such models may be tested. Jacksons
procedure). rule of thirds may be applied (see Chapter
In interpreting the outcome of an 17). And for each, the question is what hap-
analysis, compare the mean values for pens to the original relationship when con-
each category. In Table A2, for example, trol variables are applied. Box A13 indi-
the average incomes of the men are com- cates that the correlation between v63 and
pared with those of the women. v64 is 0.45; when community size is con-
The MEANS procedure may be used trolled, the correlation between v63 and
to test for intervening variables or for v64 drops to 0.42. Because the partial cor-
sources of spuriousness models (Box relation remains within one third of the
A12). value of the original relationship, it would
Note that the test to see if the relation- lead us to argue that the control variable is
ship between egalitarianism and country not a source of spuriousness.
is spurious, caused by the number of each Testing for a source of spuriousness
gender who happened to be in the sample, using CORRELATIONS and PARTIAL COR-
we rst compare the original mean differ- RELATION is identical to testing for an in-
ence between the countries (32.4000 tervening variable. And if the results were
27.3509 5.0491). Second, we now see if to come out as shown in Box A13, we
the difference holds up when we control would reject the source of spuriousness
for the gender of the respondents: Box A12 model. Remember, to nd support for the
indicates that among the male respon- source of spuriousness model, the partial
dents, the difference is 3.1849 (31.7143 correlation would have to be less than
28.5294 3.1849). For the female respon- 0.30, which would indicate that the origi-
dents, the difference in egalitarian scores nal relationship had been reduced by
is 5.897 (32.7470 26.8500 5.897). Using more than one third (0.45 0.33 0.30).
Jacksons rule of thirds, we note that
among the men, the difference decreased, a. Point-and-Click Method
but among the female respondents, the
difference remained the same. The result Analyze
is mixed and therefore does not support Correlate
the spuriousness model. Partial
680 APPENDICES
BOX A12 MEANS ANALYSIS: Sample Output with a Control Variable
Report
EGALITARIANISM
Canada, New GENDER Mean N Std. Deviation
Canada, New Zealand MALE 31.7143 42 9.8161
FEMALE 32.7470 83 7.6602
Total 32.4000 125 8.4204
Australia MALE 28.5294 17 10.8519

FEMALE 26.8500 40 7.3504
Total 27.3509 57 8.4779
Total MALE 30.7966 59 10.1349

FEMALE 30.8293 123 8.0253
Total 30.8187 182 8.7365
Notes Decreased or disappeared: 3.37

1. Is gender spuriously causing a false relation- Mixed: different category for each cell of
ship between country and egalitarianism? control
2. If gender is spuriously causing the relation- 6. To find evidence consistent with a source of
ship, then when we control for gender, the spuriousness (or an intervening variable), the
difference between the countries should dis- results of the test must indicate that the dif-
appear within both of the gender categories. ference falls into the decreased or disappeared
3. Original difference between countries was category in both categories of the control vari-
5.0491 (32.4 27.3509). able. In the present case, the difference be-
4. Compare these values to get the difference tween both the males (in the different coun-
scores. tries) and the females (in the different
5. Applying Jacksons rule of thirds: 5.0491 3 countries) would have to be less than 3.37.
1.68 7. Observed differences: males: 3.1849 (de-
Increased: 5.0491 1.68 6.7291 and creased or disappeared category); females:
over 5.897 (stayed the same category).
Stayed the same: 5.0491 1.68 3.37 to 8. Conclusion: the relationship is not spuri-
6.7290 ously caused by gender differences.
Identify variables for which you want Par- in particular, have many occasions to use
tial Correlations run; click OK. this procedure. For example, you may
have eight categories reecting commu-
14. RECODE: Collapsing and nity size and you may wish to regroup the
Switching Categories cases into three categories. The best way
to proceed with an ordinal variable is to
The RECODE procedure is used to tem- run a FREQUENCIES on the variable, not-
porarily change a variable during analy- ing on the Cumulative Frequencies col-
sis or to create a new variable. The pro- umn where cut-points of 33 percent and
cedure is primarily used to regroup the 66 percent are located. To do this se-
values in a variable. Survey researchers, quentially, click on the toolbar.
BOX A13 PARTIAL CORRELATION: Sample Output
PARTIAL CORRELATION COEFFICIENTS Notes

Controlling for.. V58 9---------

V63 V64
9--- 1. V63 fathers occupational prestige; V64
V63 1.0000 .4192 mothers occupational prestige; V58 com-
(0) ( 122) 9- munity size.
P . P .0009 2. The top number is the partial correlation co-
V64 .4192 1.0000 -- efficient.
(122) ( 0) 3. The middle number refers to the number of
P .000 P . cases in the analysis.
(Coefficient/(D.F.)/2-tailed Significance) 4. The last number is the test of significance
. is printed if a coefficient cannot be com- (probability).
puted 5. The control variable name (V58) is identified.
CORRELATIONS: Zero-Order Correlation: Sample Output

Fathers Occupation Mothers Occupation
FATHERS OCCUPATION Pearson Correlation 1.000 .446** 9------
Sig. (1-tailed) .000
N 179 131
MOTHERS OCCUPATION Pearson Correlation .446** 1.000

Sig. (1-tailed) .000
N 131 133
**Correlation is signicant at the 0.01 level (1-tailed).
Notes TIAL CORRELATION table), first-order par-

1. Testing to see if size of community is spuri- tial is 0.42.
ously influencing relation between mothers 4. Conclusion: relationship is within one third
and fathers occupational prestige. of its original magnitude; therefore, commu-
2. Original zero-order correlation (shown nity size is not spuriously accounting for re-
above) is 0.45. lationship between mothers and fathers oc-
3. With control for community size (see PAR- cupational prestige.
a. Point-and-Click Method Hint: If you are frequently reverse

scoring Likert items, consider doing it in
Transform the Syntax Editor (to get there, click on
Recode File/New/Syntax) to create a template le
Into Different Variables containing the following:
A second important use for the RECODE RECODE v33 (9 1)(8 2)(7 3)(6 4)
procedure involves reverse-scoring vari- (5 5)(4 6)(3 7)(2 8)(1 9) INTO V33R.
ables. Suppose we have three nine-point VARIABLE LABELS V33R Reversed V33.
Likert-type items we wish to reverse score
(i.e., change the 9s into 1s, 8s into 2s, Save the le with a name such as
etc.) In most cases, it is safest to create a ip.sps. In the future when you wish to
new variable containing the new codes. ip a nine-point scale, just access ip.sps,
682 APPENDICES
change the variable names to the ones the BACKWARD solution, note that all the
you want, and run the job. independent variables are entered into the
Note that the new variable names sim- equation at step 1, then the variable that is
ply have an R appended to the original not statistically signicant is removed and
name. It is a good idea to maintain the the equations are recomputed and shown
original name in the new name so that the at step 2. If we had more nonsignicant
researcher will know that it is a recoded variables, the process would have contin-
version of the original variable. Variable ued until only statistically signicant vari-
labels may be attached as indicated. ables remain. Note that the R 2 (explained
variance) value, the b coefcients (labeled
B), and beta weights are all reported on the
15. REGRESSION: Doing Multiple output (see Model Summary) as well as the
Regression Analysis test of signicant (labeled Sig.). The
model numbers refer to the steps. In this
The REGRESSION procedure does various
case, SPSS went through three steps to ar-
forms of multiple regression analysis. Re-
rive at the nal solution. Only the fathers
gression analysis was one of the proce-
educational level remained as a signicant
dures featured in Chapter 18. Recall that
predictor of Family SES.
this type of analysis allows the researcher
Table A3 presents an example of regres-
to determine an equation to describe the
sion results table. Note that along with the
relationship between a dependent vari-
b coefcients and the beta weights, and the
able and multiple independent variables
variance explained (R2) by the models
as well as to indicate the amount of varia-
tested. In cases where a number of statisti-
tion explained by each independent vari-
cally signicant predictors survive to the
able.
last model, one can use the following for-
mula to compute the percent contribution
a. Point-and-Click Method of each of the independent variables (Jack-
son, 1999):
Analyze
% Variance explained by each variable
Regression
1 R2
Linear

s
100
Identify the dependent variable, clicking it

where 1 is the beta weight for the vari-
into the Dependent window. Next move
able; R2 is the r 2 (explained variance) for
the independent variables over to the In-
the equation; and s is the sum of the
dependent(s) window. On the Method win-
beta weights ( ignoring and signs).
dow, switch to Backward; on the Statistics
Table A4 shows how these calculations
window, click on Estimates, Model t, and
could be presented. Note that this is a
Descriptives. Then click on Continue and
rough estimate because the betas repre-
then OK.
sent the particular variables in the equa-
tion and would change if other variables
b. Sample Output and Notes were included.
Pure researchers have particular in-
Box A14 presents a regression analysis ex- terest in the beta weights because they
amining the relationship between Family provide a basis for directly comparing
SES and three independent variables: the impact of the different independent
mothers educational background, fathers variables on the dependent one. Applied
educational background, and size of home researchers are usually more concerned
community. Given that we have requested with b coefcients, especially those that
BOX A14 REGRESSION ANALYSIS: Sample Output
Mean Std. Deviation N
9---
FAMILY SES 65.4439 22.2776 165
MOTHERS EDUCATIONAL BACKGROUND 4.39 1.17 165
FATHERS EDUCATIONAL BACKGROUND 4.59 1.30 165
SIZE OF HOME COMMUNITY 5.13 2.72 165
Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .390a .152 .136 20.7083
2 .389b .152 .141 20.6460
3 .385c .148 .143 20.6242

9-------------------------------------
aPredictors: (Constant), SIZE OF HOME COMMUNITY, MOTHERS EDUCATIONAL BACKGROUND, FATHERS
EDUCATIONAL BACKGROUND
bPredictors: (Constant), SIZE OF HOME COMMUNITY, FATHERS EDUCATIONAL BACKGROUND
cPredictors: (Constant), FATHERS EDUCATIONAL BACKGROUND
Notes able is dropped; only one statistically significant

1. Dependent variable: family socioeconomic variable remains at the end of this analysis.
status (SES). Independent variables: mothers 3. Final equation: Family SES 35.1 6.6 (fa-
education, fathers education, and size of thers education).
home community. All variables were entered 4. Beta weights show the relative influence of
initially, then dropped one at a time, until only variables.
statistically significant variables remain. 5. R2 is the total explained variance by the for-
2. SPSS recomputes the model each time a vari- mula predicting family SES.
Coefficientsa
Unstandardized Standardized
--- --
Coefficients Coefficients
--- ----

9-- Model B Std. Error Beta 9- t Sig.
1 (Constant) 33.459 7.230 4.628 .000

MOTHERS EDUCATIONAL BACKGROUND .313 1.899 0.16 .165 .869
FATHERS EDUCATIONAL BACKGROUND 6.739 1.711 .392 3.939 .000
SIZE OF HOME COMMUNITY .476 .597 .058 .798 .426
2 (Constant) 32.944 6.504 5.065 .000

SIZE OF HOME COMMUNITY .481 .594 .059 .810 .419
3 (Constant) 35.118 5.918 5.934 .000

9----
aDependent Variable: Family SES

684 APPENDICES
Table A3 Simplified Table Summarizing Results of Statistically Significant Variables in Predicting

Variations in Prestige Rating, Fictional Data
Multiple Regression Analysis
Dependent variable prestige rating
Multiple R 0.91493
R2 0.083709
df Sum of Squares Mean Square

Regression 2 3036.73385 1518.36692
Residual 12 590.99949 49.24996
F 30.82981
Signif F 0.0000
Variables in the Equation

Variable B SE B Beta T Sig T Percent Contribution
ED 3.65645 1.45103 0.54434 2.520 0.0269 47.8
Income 4.573087E-04 2.42024E-04 0.40817 1.890 0.0832 35.9
(Constant) 8.96285 11.81029 0.759 0.4626
Total variance explained 83.7; variance unexplained 16.3.
Variables in the Equation

Variable Beta In Partial Min Toler T Sig T
EX -0.06440 -0.11431 0.18567 -0.382 0.7100
can be changed through policy alter- socioeconomic variables compared with

ations. They provide the basis for under- the variety of experience variables).
standing how much change in the depen- A second method for estimating the
dent variable may be produced for each relative importance of variables is to cal-
change in the independent variable. Vari- culate a part correlation coefcient. This
ables can be grouped for presentation to coefcient is calculated for each inde-
illuminate the degree to which different pendent variable and represents the dif-
variable types inuence the dependent ference in the R 2 when the variable is in-
variable (e.g., how much of the variation cluded versus when it is not included.
in the dependent variable is caused by The variables may then be ordered in
Table A4 Calculating the Percentage Contribution of Each Independent Variable

Variable Beta Calculation Percent Contribution
ED 0.54434 ([0.54434 * 0.83709]/0.95251) * 100 47.8
Income 0.40817 ([0.40817 * 0.83709)/0.95251) * 100 35.9
Total variance explained 83.7; variance unexplained 16.3.

terms of the unique contribution of each Recode

independent variable. To get Part Corre- Into Different Variables . . .
lations in regression analysis, click on
Statistics and then click on the PART and Using this procedure, create one fewer
PARTIAL CORRELATIONS window. new variable than categories in the origi-
In presenting the results of a regres- nal religious afliation variable. Each of the
sion analysis, most researchers present new variables will be entered into the re-
the nal table after the nonsignicant gression analysis as an independent vari-
variables had been dropped. But you will able. In the example of a three-category re-
want to include the following: ligious afliation variable (Protestant,
Catholic, and Jewish), one would make one
Note the names of the variables that
variable PROT with a coding of 1 for
remained.
Protestants and everyone else coded 0. A
Note the names of the variables that
second variable, CATH, would be coded 1
were dropped.
for the Catholics and 0 for everyone else.
Point out total variation explained (R2
The third religious category, Jewish, is
value).
taken into account in the residual category
Note the percentage contribution of
of the two previous categories. One has to
each of the statistically signicant vari-
be cautious in interpreting the results
ables.
when dummy variables are used. These
Discuss the implications of the ndings.
variables are at some disadvantage in ex-
In addition, the formula describing the
plaining variance because they have only
relationship may be presented. In the
two values, 0 and 1. In short, we may some-
case presented here, the formula
what underestimate the importance of the
would be as follows: family SES 35.1
variable religion in such an analysis.
6.6 (fathers education)
Note that the formula uses b coef-
cients and the a value is the constant
(the general form of the equation is Y d. Some Cautions
a b1X1 b2 X2)
Caution 1. Ensure variables are theoret-
ically independent from one another.
What this means is that one cannot use
c. Dummy Variable Analysis aspects of the dependent variable as
independent variables. Include only
There may be times when a researcher
meaningful potential causes of the de-
wishes to use a nominal variable within a
pendent variable, not alternate mea-
regression analysis. Using dummy vari-
sures of it.
ables can do this. Suppose we had three
Caution 2. Watch out for highly corre-
religious categories, Protestant, Catholic,
lated independent variables. The
and Jewish. To submit religion as a vari-
weightings that will be attached to the
able, we would create two new variables
variables will be unstable if the inde-
(one less than the number of categories)
pendent variables are strongly corre-
and enter each of these into the regres-
lated with one another. The program
sion analysis. These variables are coded
prints out a warning if there is a prob-
as 1 for presence and 0 for absence.
lem in this area.
These are best created using the RECODE
Caution 3. Interpret weightings with
procedure, as in:
care. Understand that the weightings
( i ) Point and Click are for the particular combination of
Transform independent variables for a particular
686 APPENDICES
sample and that these weightings may case, however, the researcher has to en-
not apply reliably to other samples. sure that appropriate items are included in
Caution 4. Monitor the number of cases each subindex.
carefully. By default, SPSS deletes a Chapter 13 presents a discussion of
case if it has a missing value in any of item analysis and introduces the princi-
the variables in the equation. Thus, if ples of index construction. The RELIABIL-
there are several variables in an analy- ITY procedure evaluates which items
sis, there is a danger of losing many should go into an index. RELIABILITY ex-
cases. And as the number of cases amines the components of a proposed ad-
drops closer to the number of vari- ditive index providing a variety of diag-
ables, the R2 increases dramatically. To nostics for each item. The procedure does
determine the number of cases used in not actually compute the index; rather, it
the analysis, add 1 to the total degrees of provides an assessment of each item. The
freedom reported in the table. If a large actual index would be constructed with a
number of cases have been dropped, COMPUTE command. Of particular note
try running the analysis using PAIR- for beginning nurse researchers is that RE-
WISE treatment of missing cases. If LIABILITY computes item means and stan-
there are still many missing, try MEANS dard deviations, interitem correlations, a
(this will substitute the mean of the series of item-total comparisons, and
variable for any missing cases). Cronbachs alpha for index reliability.
The RELIABILITY procedure in SPSS
An additional possibility, after an ini-
uses the internal consistency approach
tial narrowing down of the variables has
to reliability. The method that is used to
been done, is to resubmit the analysis,
calculate the standardized alpha involves
naming only the signicant variables plus
the number of items going into the index
perhaps two or three that were dropped
and the average interitem correlation
in the last few steps. This preserves the
among the items, as in:
cases that were dropped because of miss-
ing values in variables that are not in the Alpha
nal equation.
n (average inter-item correlation)

1 (average inter-item correlation (n 1))
16. RELIABILITY: Assessing
Internal Consistency Where n is the number of items proposed
for the index being tested and the aver-
Indexes combine two or more indicators to age interitem correlation is the average of
reect complex variables such as SES, qual- the correlations between the items: this
ity of life, health status, or an attitude to- value is printed when the SUMMARY op-
ward a particular nursing issue. Frequently, tions MEANS is selected
a researcher constructs subindexes that The standardized alpha for the index
may be treated alone or combined with evaluated in Box A15 would be as follows:
other subindexes to form a composite mea-
sure. For example, a researcher measuring 5(0.3507)
Alpha
attitudes toward abortion may construct a 1 [0.3507 (5 1)]
subindex for soft reasons (e.g., economi- Alpha 1.7535 2.4028
cally inconvenient, preference for having a Alpha 0.7298
baby later) and hard reasons (e.g., preg-
nancy as a result of rape, severely handi- Cronbachs alpha varies with the average
capped). These subindexes might also be interitem correlation, taking into account
combined to form an overall index. In each the number of items that make up the in-
BOX A15 RELIABILITY ANALYSIS: Sample Output
RELIABILITY ANALYSISSCALE (ALPHA)

Mean Std Dev Cases
}
1. V29 4.4341 2.4435 182.0
2. VR30 7.2143 2.2773 182.0
3. VR31 6.7253 2.4430 182.0 9-----
4. VR32 7.6209 1.9845 182.0
5. VR33 5.4341 2.8291 182.0
}
Correlation Matrix
V29 VR30 VR31 VR32 VR33
V29 1.0000
9---
VR30 .1331 1.0000
VR31 .3135 .3185 1.0000
VR32 .1629 .5046 .4422 1.0000
VR33 .3043 .3740 .5569 .3975 1.0000
No. of Cases 182.0
No. of
Statistics for Mean Variance Std Dev Variables
Scale 31.4286 69.4507 8.3337 5
Item Means Mean Minimum Maximum Range Max/Min Variance
6.2857 4.4341 7.6209 3.1868 1.7187 1.7480
Inter-item
Correlations
9--------- Mean Minimum Maximum Range Max/Min Variance
.3507 .1331 .5569 .4238 4.1837 .0173
-
Item-total Statistics
----- ------
- ------
Scale Scale Corrected 9----
Mean Variance Item- Squared Alpha
if Item if Item Total Multiple if Item 9----
Deleted Deleted Correlation Correlation Deleted
V29 26.9945 52.4696 .3110 .1227 .7464
VR30 24.2143 49.8599 .4479 .2908 .6947
VR31 24.7033 44.2098 .5932 .3880 .6367
VR32 23.8077 50.6203 .5274 .3508 .6714
VR33 25.9945 40.2596 .5901 .3773 .6358
Reliability Coefficients 5 items
Alpha .7268 Standardized item alpha .7298
Notes 4. The Corrected Item-Total Correlation is a

1. Lists the means, standard deviations, and measure of the correlation between the item
number of cases for the variables. and the total score minus the item score.
2. Shows the correlation matrix (check to see 5. Provides the standardized alphas if the item
that all are positive). were deleted (suggests we should drop item
3. Note especially the mean interitem correla- V29 because that would result in a higher
tion (should be above 0.25). alpha).
688 APPENDICES
dex. If there is an increase in either the in- be above 0.25 (in the case we are ex-
teritem correlation or the number of amining, the value is 0.3507). Note here,
items, alpha increases. For example, with however, that if you were using items
two items and a 0.4 mean correlation, the based on Likert items with fewer than
alpha value would be 0.572; with eight nine points, you might wish to accept a
items and a 0.4 mean interitem correla- slightly lower mean interitem correla-
tion, the alpha would be 0.842. With 0.6 in- tions, perhaps as low as 0.25.
teritem correlations, the alpha for two Examine the item-total statistics. This
versus eight items would be 0.750 and part of the analysis examines the indi-
0.924, respectively. Table 13.4 summa- vidual items, comparing each with the
rizes the relationship between items and total score for the index being evalu-
interitem correlations (see Chapter 13 for ated. The corrected item-total correla-
more details). tion presents the correlation between
In examining Box A15, note that we are the item and the total index score with
examining ve nine-point Likert-type lib- the effects of the individual item re-
eral gender attitude items, known as v29, moved. Generally, you would expect
v30, v31, v32, and v33. Note that v30 to each of these items to be above 0.25.
v33 are inverse measures (the higher the At the bottom of the table, two alphas
score, the less liberal the attitude). Let us are reported. Earlier the calculation
go through the steps to evaluate these for the standardized item alpha was
items. demonstrated. We attempt to get as
high an alpha as possible when con-
Step 1. Reverse score any items that are structing an index. The alpha reported
negative ( in this case, v30 to v33) us- in Box A15 is 0.7268.
ing the RECODE procedure. The alpha if item deleted column re-
Step 2. Point-and-Click Method ports on what the alpha would be if the
Analyze item were deleted. In Box A15, we note
Scale that if we dropped v29, the alpha would
Reliability be 0.7464; for all other cases, the alpha
Identify variables for which you want would be lowered if the item were
procedure run (V29, V30R, V31R, V32R, deleted. Our decision will be to drop
and V33R). Click on Statistics and then v29 when the nal index is calculated
click to place a checkmark on all items using the COMPUTE procedure.
in the Descriptives pane. In the Sum- Step 4. If two or more items are to be re-
maries pane, check Means and Correla- moved from the nal index, it is neces-
tions. In the Interitem pane, check Cor- sary to resubmit another RELIABILITY
relations. In the ANOVA pane, check job. If one or no items need to be elim-
None. After this is done, click on Con- inated, then one can go immediately to
tinue. At the Reliability Analysis win- step 5.
dow, check OK. Step 5. After deciding which items are to
Step 3. Examine the results shown in Box be included in the index, a new job
A15. should be run, adding together the
As a rule of thumb, you have an ac- items using a COMPUTE command.
ceptable index if: (Either point-and-click or syntax com-
The correlation matrix indicates that mands may be used.) The Syntax com-
the correlations between the items are mand would look like:
positive ( in this case, all are positive
because they vary from 0.13 to 0.56). COMPUTE sexatt v30r v31r v32r v33r.
The mean interitem correlations should VARIABLE LABELS sexatt Sex Role Attitude.
The system le should be updated, sav- a. Point-and-Click Method

ing the new index scores with a File/
Save command. There are many other Analyze
options available within the RELIABIL- Compare Means
ITY procedure. It is possible to compute Independent Samples t-Test
split-half coefcients and various coef- or
cients proposed by Louis Guttman. Con- Paired Samples t-Test
sult the SPSS Users Guide for additional The following steps need to be taken to
models and options. complete the analysis:
17. SPEARMAN: Correlation for For the between-subjects design (using

Ordinal Data the Independent Samples t-Test), move
the test variable to the Test Variables
Spearman correlations are used when or- window. Next place the Grouping Vari-
dinal level measurement has been at- ables in its window. You will need to
tained. They may be reported and inter- then enter the values for the Grouping
preted in the same manner as Pearson (usually 1 and 2, or 0 and 1). Click OK.
correlations. Similar to Pearson correla- For the within-subjects design (using
tions, they vary from 1.00 to 1.00 and the Paired Samples T-Test), move the
measure the strength of an association. paired variables to the window and
click OK.
a. Point-and-Click Method
b. Sample Output and Notes
Analyze
Correlations Box A16 shows sample output from two t-
Bivariate TEST analyses. The rst shows the com-
Spearman mand format and results from a between-
In the Bivariate Correlations box, click on subject design; the second shows how
Spearman. Identify the variables for to set up a job to do a within-subjects
which you want the procedure run. Then analysis.
Click OK.
19. Z SCORE: Using Z Scores to

18. t-TEST: Comparing Means Create an Index
The t-TEST procedure tests whether the Z scores are used to standardize variables
difference in the means between two so that each variable will have a mean of 0
groups is statistically signicant. For a and a standard deviation of 1. Suppose you
between-subjects design, use the Indepen- wished to create an index that adds the
dent Samples t-Test; for a within-subjects mothers educational category to her oc-
design, use the Paired Samples t-Test. cupational prestige, equally weighting the
(Recall that a between-subjects test in- two elements. One way to proceed (deal-
volves the comparison of two groups of ing with the different range of variables in-
subjects. Typically one group is exposed volved) would be to standardize the
to a treatment and the other group is not scores (Z scores) and then add together
exposed to it. A within-subjects test is the standardized scores. Z scores versions
used when the same subject is exposed to of variables are available in the DESCRIP-
the different treatments; these are also TIVES procedure by simply naming the rel-
known as repeated measures designs.) evant variables and clicking on the box
690 APPENDICES
BOX A16 T-TEST: Sample Outputs
(A) Within-Subjects Sample Output

----
Paired Samples Statistics - - ---- - --------
------
------Std.
Mean 9- N Deviation Std. Error Mean
Pair time 1 speed 13.260 10 .324 .102
1 time 2 speed 13.160 10 .267 8.459E-02

Paired Samples Correlations -- ---
-- ---
N Correlation - --- Sig.
Pair time 1 speed & time 2 10 .941 9 --- .000
Paired Samples Test

Paired Differences
95% Confidence
Interval of the

Std. Std. Error Difference
- - ---
Mean Deviation Mean Lower Upper t ---
df Sig. (2-tailed)
-
---
Pair time 1 speed
9 - - -
- - - - -----
time 2 speed 1.000E-01 .115 3.651E-02 1.74OE-02 .183
-
2.739 9 9 .023 9-----
(B) Between-Subjects Sample Output
Group Statistics
Type of Tire N Mean Std. Deviation Std. Error Mean
TIME Standard 10 18.390 2.168 .685
Experimental 10 18.060 9----2.447 .774
----- -----------
--------
Independent Samples Test
Levenes
Test for
Equality
of
- ------
-
Variances t-test for Equality of Means
- -- ----
-
---- 95%
-
- - - -- -
- - - - ------
Confidence
-
- ---- Mean
-Sig. - - ------Interval of the
---Std.
- - - - - Error Difference
F Sig.
-
t 9 df 9 ----- Difference Difference Lower Upper
(2-tailed)
TIME Equal variances .524 .478 .319 18 .753 .330 1.034 1.842 2.502
assumed
Equal variances .319 17.741 .753 1.844 2.504
not assumed
9--.330 1.034
------------------
----
BOX A16 t-TEST: Sample Outputs (Continued )
Notes (Within-Subjects t-TEST) Notes (Between-Subjects t-TEST)

1. The mean times at time 1 and time 2 shown 1. The mean lap times for the standard and
here. experimental tires shown here.
2. The correlation of the measures at the two 2. Computed t value.
times is 0.941. 3. Degrees of freedom.
3. The t-test value. 4. Two-tailed test of significance.
4. The degrees of freedom. 5. Conclusion: accept the null hypothesis; there
5. The two-tailed test of significance. is no statistically significant difference between
6. Conclusion: reject the null hypothesis; there the lap times on the different tires.
is a statistically significant difference in the
speeds at the two times.
called Save Standardized Values as Vari- The SPSS package contains a broad range
ables. A Z will be prexed to the name of of options, statistics, le management, and
the variable. Next you would simply do a analysis procedures. The chances are that
COMPUTE, adding together the new stan- most analyses you wish to do are possible
dardized variables. The COMPUTE could within SPSS. After you have mastered the
be done as follows: basic procedures, you will be ready to be-
gin exploring SPSS in greater detail.
COMPUTE newses zv62.1 zv64.
VARIABLE LABELS newses SES Score:

Education plus Prestige. C. SPSS MANUALS
Incidentally, if you wished to weight There are many manuals currently avail-
the elements so that two thirds of the - able for SPSS for Windows version 10.0.
nal scores were based on occupational The major ones of particular relevance to
prestige and one third on educational this text are listed in the Recommended
level, the COMPUTE command would be Readings section. They are available in the
altered to: United States from SPSS Inc., Publication
Sales, 444 N. Michigan Ave., Chicago, Illi-
COMPUTE newses (zv62.1 * 0.33) nois 60611. In Canada, the distributor of
(zv64 * 0.67). SPSS publications is Prentice Hall Canada
Inc., 1870 Birchmount Road, Scarborough,
The previous section presented the ba- Ontario, M1P 2J7. In Canada, call 1-800-
sic commands for examining relationships. 567-3800.
SPSS. (1999). SPSS Base 10.0 Applications GLM repeated measures and loglinear and
Guide. Chicago: SPSS Inc. This basic man- other advanced techniques.
ual contains the basic procedures covered SPSS. (1999). SPSS 10.0 Syntax Reference Guide.
in Doing Nursing Research. Chicago: SPSS Inc. For those who prefer to
SPSS. (1999). SPSS Advanced Models 10.0. use syntax commands, this is the complete
Chicago: SPSS Inc. This advanced tech- reference guidesome 1412 pages long!
niques manual is of interest to those doing
Appendix B
A Student-designed
Questionnaire
The following questionnaire is part of a study being conducted as a Nursing 300 pro-
ject. To ensure the condentiality of your responses, please do not write your name
on the questionnaire. Thank you for your cooperation.
1. Gender: I am:
Male-1
Female-2
2. In what year were you born? 1 9 ___ ___
3. What program of study are you currently enrolled in?

Bachelor of Arts -01
Bachelor of Business Administration 02
Bachelor of Education 03
Bachelor of Engineering 04
Bachelor of Music 05
Bachelor of Science in Nursing -06
Bachelor of Science in Human Nutrition -07
Bachelor of Science in Human Kinetics08
Bachelor of Science 09
Bachelor of Arts in Information Systems 10
Other 11
If other, please specify______________________________
4. What year of study are you currently in?

Freshman 1
Sophomore -2
Junior -3
Senior 4
Other -5
If other, please specify__________________
692
A STUDENT-DESIGNED QUESTIONNAIRE 693
5. Approximately, what was the population of your hometown/or city before com-
ing to the university?
Rural area under 1000 -1
10004,999 2
50009999 3
10,00019,999 -4
20,00029,999 -5
30,00039,999 -6
40,00049,999 -7
50,000 and over -8
6. What is your religious afliation?

Roman Catholic 01 Jewish 07
Anglican 02 Buddhist 08
United 03 Muslim 09
Baptist -04 None 10
Presbyterian 05 Other -11
Other Protestant 06 Please specify_________
7. Currently, how often do you attend church services?

More than once a week-1
Once every week -2
23 times a month -3
Once a month 4
710 times a year -5
26 times a year 6
Once a year -7
Never -8
8. What is (or was) the main occupation of your father? (e.g., social worker, ele-
mentary school teacher)?
Job position________________________________________________________
Brief job description_________________________________________________
9. What is (or was) the main occupation of your mother? (e.g., high school teacher,
homemaker, lawyer)?
Job position_________________________________________________________
Brief job description__________________________________________________
10. Ann is 50 years old and is a cancer patient. She is in the last stage of the disease.
It has been decided by her physician that no treatment options will serve any ben-
et to treat her cancer or reduce discomfort. She has left the hospital of her own
free will and wishes to die. The physician prescribes medication for Ann, know-
ing she intends to overdose on it. In regards to the physicians actions, I . . .
Strongly Disagree 1 2 3 4 5 6 7 8 9 Strongly Agree
694 APPENDICES
11. Ben is 30 years old and has been paralyzed from the neck down as a result of an ac-
cident some years ago. He will be bedridden for the rest of his life. He no longer has
control over urine or bowel movements and needs to be fed and cared for. He nds
a passive life in an institution unbearable. Ben wishes to die. The physician pre-
scribes medication for Ben and sets it up so that he will be able to take it in over-
dose amounts himself when he chooses. In regards to the physicians actions, I . . .
12. Margaret is a 90-year-old woman who lives by herself. She has weak vision and
arthritis and can no longer do the things she enjoys. Margaret spends most of her
day alone except for 1-hour visits from a home-care nurse. She is lonely and
wishes to die. The physician prescribes medication for Margaret, knowing she
intends to overdose on it. In regards to the physicians actions, I . . .
13. Ann is 50 years old and is a cancer patient. She is in the last stage of the disease.
It has been decided by her physician that no treatment options will serve any
benet to treat her cancer or reduce discomfort. She has left the hospital of her
own free will and wishes to die. She has asked her physician to inject a lethal
drug into her to bring about her death and the physician has agreed to do it. In
regards to the physicians actions, I . . .
14. Ben is 30 years old and has been paralyzed from the neck down as a result of an
accident some years ago. He will be bedridden for the rest of his life. He no longer
has control over urine or bowel movements and needs to be fed and cared for. He
nds a passive life in an institution unbearable. Ben wishes to die. He asks his
physician to give him a lethal injection to end his life and the physician has
agreed to do it. In regards to the physicians actions, I . . .
15. Margaret is a 90-year-old woman who lives by herself. She has weak sight and
arthritis and can no longer do the things she enjoys. Margaret spends most of her
day alone except for 1-hour visits from a home care nurse. She is lonely and
wishes to die. She has asked her physician to inject a lethal drug into her to
bring about her death, and the physician has agreed to do it. In regards to the
physicians actions, I . . .
16. I would help a friend ght for the right to die if that was his or her choice.
17. I would refuse to provide a terminally ill person with a lethal dosage of a drug, if
he or she requests it.
18. I would rather die than live with a chronic, painful, nonterminal illness.
19. I would personally give a lethal injection to a terminally ill individual, if requested
by the person.
20. I would provide emotional support to a terminally ill friend who wished to end
his/her life.
21. People have the right to take their own lives if they are terminally ill.
22. People have the right to take their own lives if they have suffered a great nancial
loss.
23. Active euthanasia is dened as an action by a person with the intention of has-
tening the death of another person who is terminally ill or in such a condition that
causes suffering. Active euthanasia is wrong under all circumstances.
24. Physician-assisted suicide is an action taken by the physician that assists a

person in ending his or her own life. Physician-assisted suicide should be legal-
ized.
25. I am in favor of active euthanasia.

26. I am in favor of physician-assisted suicide.

27. I feel depressed.

28. I worry about the way I look.

29. I feel lonely.

696 APPENDICES
30. I worry about what others think.

31. I have a tendency to get involved in unhealthy relationships.

32. I usually avoid talking about my feelings.

33. I nd that other peoples moods inuence my own mood.

34. I occasionally question whether my life is worth living.

35. My religious beliefs are strong.

36. I have been at the bedside of one who was dying:

Yes1
No-2
37. Indicate whether or not anyone from the following categories is currently suffer-
ing from a long-time illness or has died following a long period of suffering. (Check
Yes or No for each item.)
Yes No
Close friend _____ _____
Acquaintance _____ _____
Immediate family member - _____ _____
Relative _____ _____
Neighbor _____ _____
Other - _____ _____
If other, please specify_________________________________________
38. If yes for any part of question 37, for the person you had the most contact with:
How much time did you spend with that person while he or she was sick?
Lived with 1
Visited every day-2
Visited 23 times a week 3
Visited once a week-4
Visited 23 times a month -5
39. How long was this person ill?

12 weeks -1
Less than a month -2
Less than 6 months 3
Less than 1 year 4
1 year or more 5
40. I would rather die than become dependent on my family to care for me.
41. I believe in a superior being with whom I have a relationship.

42. I would rather die than live in a nursing home.

43. If a family member had cancer and decided against having chemotherapy, I would
question his or her decision.
44. God would not allow anyone to have more pain than they could handle.
45. I love to get up each day to live life to the fullest.

46. I believe that a higher being is there for those in times of anguish and suffering.
47. I often overcommit myself.

48. I love to be the center of attention.

49. When children are young, a mothers place is in the home.

50. I worry about dating and relationship problems.

698 APPENDICES
51. Capital punishment is acceptable if the offender cannot be rehabilitated.

52. Capital punishment is never acceptable.

53. Abortion is okay in the case of unplanned pregnancy.

54. Abortion is okay in the case of rape.

55. My life is meaningless.

56. My religious beliefs greatly inuence decisions I make.

57. When I face difcult times in my life, I believe a greater force will meet me
halfway.
58. If I were diagnosed with a disabling illness, my biggest fear would be of becoming
dependent on others.
59. If a terminally ill person refuses to eat to hasten death, health-care workers should
implement alternative ways to provide nutrition.
60. I believe that I am greatly needed in this world.

61. Abortion should be entirely the personal choice of the woman involved.
62. I believe in life after death.

***Thank you ***

Glossary
Action research: The systematic collection Androcentricity: A presentation given ex-

and analysis of data for the purpose of taking clusively from a male perspective.
action and making change. It provides prac- Antecedent variable model: A causal model
titioners, organizations, or communities with that proposes a variable that causes variation
tools to solve their problems. in an independent variable which, in turn, in-
Accessible population: A population that is uences the dependent variable in the model.
feasible for a researcher to access. Thus, the antecedent variable is one that pre-
Advocacy research: Social research that ad- cedes both the main independent and depen-
vocates social changes or advances the per- dent variables.
sonal or collective agendas of its practitioners. Appeal to authority: Using the opinion of a
Aesthetics: The study of the nature of beauty person on a topic; the persons expertise falls
or art. In nursing, aesthetic knowledge allows outside the topic area.
nurses to perceive the beauty underlying each Applied research: Focuses on guring out
encounter with each client. how to bring about specic changes in society.
Alternative hypothesis: See Research hy- Audience: The specic people a research
pothesis. report is geared toward. Knowing who your
Alternative explanations: A primitive form audience is helps you determine what content
of theory testing. This is the case because the- to include, the level and complexity of ideas to
ory and explanation do the same thingto an- present, and the areas to be emphasized.
swer a particular why question. We propose Audit trail: Documenting all the raw data
a series of alternative explanations for a rela- generated as well as methods and sources of
tionship. After that, all we have to do is to mea- data generation and analysis decisions.
sure the appropriate variables and do the Axiomatic derivation: A method of logically
analysis. deriving new statements of relationship from
Alpha: The probability of making a type I er- a given set of assumptions and propositions.
ror. b Coefcient: Associated with regression
Analysis of variance (ANOVA): A procedure analysis; a nonstandardized measure of the in-
for deciding if a ratio level dependent variable uence of an independent variable on the de-
is signicantly associated with a nominal or pendent one.
ordinal independent variable. BACKWARD (stepwise) solution: In regres-
Analytic les: Files relating to a specic sion analysis, all the variables are included in
topic or relationship explored in a eld study. the equation; then the least important vari-
Anecdotal evidence: Evidence for a gener- able is dropped and the equation is recalcu-
alization that is based on one incident that lated. This procedure is repeated until only
serves to illustrate why or how some relation- signicant variables remain.
ship exists. Baseline data: The score a subject has be-
699
700 GLOSSARY
fore the introduction of the treatment vari- Chi-square test: A test of statistical signi-
able. cance associated with contingency table
Baseline measure: A measure taken after analysis, in which the dependent variable is a
stability has been achieved in the dependent nominal one.
variable at the beginning of a set of observa- Classic experimental design: A between-
tions. subjects design involves a control and an ex-
Baseline stability: Several measures are perimental group. Measures are taken from
taken; the treatment variable is not intro- members in both groups before treatment and
duced until the measures achieve stability. repeated after the treatment has been experi-
Benecence: The duty to promote or do good. enced.
Beta: The probability of making a type II error. Cleaning the data: The systematic search
Beta weight: Associated with regression analy- for errors in a data set so that they can be cor-
sis, the beta weight is a standardized measure rected before data analysis begins.
of the relative inuence of an independent vari- Clinically signicant: Findings that must
able (compared with other independent vari- have meaning for patient care in the presence
ables) on the dependent one. or absence of statistical signicance; this re-
Between-subjects design: Study design in quires the nurse to interpret the ndings in
which each group of experimental subjects is terms of their value to nursing.
exposed to either the control group treatment Closed-ended questions: Precoded, single-
or to the experimental group treatment. choice questionnaire items.
Bias: A preference or predisposition to favor Cluster sample: Sample with population di-
a particular conclusion. vided into groups (or clusters). The rst units
Biased sample: A sample that is not repre- or clusters to be sampled are the largest, fol-
sentative of the population it is intended to re- lowed by smaller and smaller units selected at
ect. random (see also Multistage area sample).
Bimodal distribution: A distribution with Coefcient of reliability: A measure of
two peaks. agreement between the coders on the catego-
Blocked design: Experiments in which sub- rization of the items being analyzed. The pro-
jects have been grouped together on a partic- portion of times there is agreement is re-
ular variable that needs to be controlled; sub- ected in the coefcient.
jects are then randomly assigned to treatment Comparative studies: Studies that typically
and control conditions. involve cross-cultural or historical analyses of
Bracketing: A cognitive process used by re- social behavior.
searchers to set aside ones biases and per- Componential analysis: Analysis that looks
sonal perspectives about a research topic. Its for all the contrasts among the cultural cate-
purpose is to make known what the researcher gories in the domains. It is the systematic
believes about the research topic so he or she search for the attributes (components of mean-
can then approach the topic with less bias. ing) associated with cultural categories.
Candidate variable model: A model that pro- Concept: A general idea referring to a char-
poses several independent variables as possi- acteristic of an individual, group, or nation.
ble causes of variation in a dependent variable. Conceptual framework: Does not contain a
Causal explanation: When an event or se- specic theory that explains the expected re-
quence of events is explained by making ref- lationship between variables but rather syn-
erence to preceding, inuencing events. thesizes relevant literature about the pro-
Causal model: A diagram showing the posed hypotheses.
causal connections between variables. Conceptual hypothesis: A statement of the
Change agent: A person or group of people relationship between two or more conceptual
who can take an idea for change from the variables.
drawing board, carry it through all the stages Conceptual level of research: Entails the
of its implementation, and then evaluate its denition of variables that are to be used.
success or failure. Change agents must be able Conceptual map: A diagram of the concepts
to respond to the barriers that mitigate against and relationships expressed in a theoretical
research utilization. framework. A conceptual map is often a more
GLOSSARY 701
efcient way to communicate what is known each piece of data is coded and compared
about a phenomenon than is a verbal descrip- with other pieces for similarities and differ-
tion. ences in the lives of those interviewed.
Conceptual utilization: The use of ndings Constructs: Concepts specied in such a
to enhance ones understanding of a problem way that they are observable in the real world.
or issue in nursing. Through conceptual uti- Making a concept potentially observable facil-
lization, a nurse uses the knowledge gained itates testing of the idea.
from research to cognitively restructure the Construct validity: A situation in which a
way one thinks about a situation, problem, or theoretically derived hypothesis turns out as
phenomenon in nursing. predicted.
Conceptual variable: An idea that has a di- Contemporary applied research: The goal is
mension that can vary. to produce knowledge that can be used for the
Concurrent validity: When two different tests specic purpose of generating positive change.
are given and the scores correlate highly with Unlike conventional research, which mainly
each other. contributes to knowledge development that
Conditional variable: A variable that ac- may not necessarily be used in practice, the
counts for a change in the relationship be- emphasis in contemporary applied research is
tween an independent variable and a depen- to use the research to inform the change.
dent variable when the general conditions Content analysis: A technique for making in-
change. ferences by objectively and systematically
Condentiality: Based on the ethical princi- coding information.
pal of respect for persons. Most surveys, in- Content validity: The extent to which the
terviews, experiments, and eld studies are measure reects the dimension (or dimen-
completed on the understanding that individ- sions) implied by the concept.
ual responses, or information that would per- Control by constancy: Within-subject de-
mit the identication of the individual, will signs in which the same subject experiences
never be released. different levels of the treatment; hence, the
Conict theory or perspective: Argues that subject acts as his or her own control.
society is fundamentally characterized by con- Control group design: An experimental de-
ict between interest groups; owners of the sign the researcher uses randomized assign-
means of production (e.g., factories, farms, ment to groups (or precision matching) to ad-
businesses), the bourgeoisie, seek to exploit just for known and unknown variations between
workers, the proletarians, whose labor is un- the two groups.
dervalued and underpaid. Controlled observations: Observations in
Conrmability: The objectivity of the data. which other confounding factors are mini-
Meanings emerging from the data have to be mized or taken into account.
tested for their plausibility, their sturdiness, Control variable: One that is taken into ac-
their conrmability, so that two indepen- count in exploring the relationship between an
dent researchers would agree about the independent and dependent variable. There
meanings emerging from the data. are three basic types of control variables:
Conrmatory factor analysis: Usually con- the intervening variable, the conditional vari-
ducted after an exploratory analysis has oc- able, and the source of spuriousness (or con-
curred or when the researcher has a hypoth- founding) variable.
esis about the existing factor structure that is Convenience sample: A nonprobability sam-
to be tested. It is usually based on theory. pling procedure that involves selection on the
Confounding variable One that may uninten- basis of ease or convenience.
tionally obscure or enhance a relationship. Core variable: A variable that focuses the
Consequentialist: This view of research theory and accounts for most of the variation
ethics stresses that ethical judgments about a in a pattern of behavior that is both relevant
research project should be made in terms of and problematic for the participants involved.
its consequences for the subject, for the acad- Correlation analysis: A procedure for mea-
emic discipline, and for society. suring how closely two ratio level variables
Constant comparative method: Used when co-vary together.
702 GLOSSARY
Correlation coefcient (r): A measure of the 1) receive treatment A (the nicotine patch) fol-
strength of association between two vari- lowed by treatment B (Zyban), and the other
ables; a correlation may vary from 1 to 1. half (group 2) receive treatment B (Zyban) fol-
Correlation matrix: A table showing the cor- lowed by treatment A (the nicotine patch).
relation coefcients between all designated Cross-sectional: Data collected at one point
variables. in time.
Counterbalancing: In experimental designs, Cross-tabulation tables: Present informa-
counterbalancing involves introducing, chang- tion so that the relationship between a nomi-
ing, maintaining, and then returning to the rst nal level dependent variable can be related to
level of the experimental treatment to control an independent variable.
for effects of learning on the subjects perfor- Cultural scene: An ethnographic term used
mance. to refer to a culture under study.
Covariates: Ratio level treatment variables. Cultural themes: Recurrent patterns in the
Credibility: The accuracy of the description data that are used to connect domains.
of the phenomenon under investigation. The Culture: A way of life belonging to a desig-
portrayal of the reality must be faithfully rep- nated group of people. It includes all the ways
resented and plausible to those who have ex- a group of people solve problems as reected
perienced that reality. in their language, dress, food, traditions, and
Criterion validity: The extent to which a customs. It is a pattern of living that guides the
measure is able to predict accurately. groups thoughts, actions, and sentiments.
Critical appraisal: Involves judging whether Curvilinear relationship: One in which the
or not a research study is described clearly plot between the variables goes in one direc-
and comprehensively enough to decide if the tion and then switches to another one.
ndings and implications are logical and be- Data collection: The process of gathering in-
lievable and should be considered seriously formation from identied participants to an-
in your practice. swer a research question.
Critical approach: Views human behavior Data distribution: A listing of all the values
as fundamentally characterized by different for any one variable.
groups attempting to enhance their interests Data entry: The process of transferring the
at the expense of less powerful groups. The information collected in a study to a comput-
fundamental goal of the critical approach is to ing device.
bring about a truly egalitarian societyone in Data massaging: The practice of playing
which there is an equality of opportunity and with the data until the analysis producing the
an equality of result. strongest association is identied and re-
Critical value: This value is determined by tained. In massaging the data, the bias will
looking up the appropriate cell on a statistical usually be to nd evidence supporting ex-
table. In order for a null hypothesis to be re- pected or preferred outcomes.
jected, the critical value must be equaled or Data reduction: The researcher may collect
exceeded by the computed test value. (When a large amount of data and then reduce the
a computer program is used, these values are data through factor analysis. A single com-
not necessary because the exact probabilities posite variable or several variables measuring
are provided on the computer output). different dimensions of a concept may be cre-
Cronbachs alpha: Used in assessing the inated out of a conglomerate of variables
ternal consistency of an index. This test is through factor analysis.
based on the strength of the intercorrelations Debrieng: Researchers explaining studies
of all the items in the instrument as well as the to subjects after the data have been collected,
number of items used. The alpha value ranges noting any deception and why it was neces-
from 0 to 1, with 1 indicating perfect internal sary and reassuring subjects that their partic-
consistency and 0 indicating no internal con- ipation was appreciated and helpful.
sistency. Declaration of Helsinki: This declaration
Crossover design: Assign subjects randomly differentiates between therapeutic research
to a specic sequencing of treatment condi- and nontherapeutic research. Therapeutic re-
tions. For example, half of the subjects (group search is research that provides participants
GLOSSARY 703
with an opportunity to benet from an experi- Design: Similar to a blueprint for a study, it
mental treatment. Nontherapeutic research is guides the investigator in planning and imple-
research that generates new knowledge that menting a study. It provides a detailed plan for
will bring future benets to society, but those data collection and analysis and is the critical
acting as participants most likely will not ben- element linking the theoretical framework and
et from their participation in the research. questions with the resultant data.
Deductive explanation: When the phenom- Dialectic critique: Makes explicit, internal
enon to be explained is a logically necessary contradictions in the data rather than compli-
consequence of the explanatory premises, as mentary explanations. By exposing the con-
in if A B and B C, then A C. tradictory nature of phenomena in the change
Deductive reasoning: A process by which a situation, the researchers and participants
nurse reaches a conclusion by moving from come to a clearer understanding of the change
the general to the specic. It is the approach process.
used to test predictions and validate existing Discrepancy: The difference between the ways
relationships through quantitative research things are in practice and the way they should
methods. be, or between what one knows and what one
Degrees of freedom: In cross-tabular tables, needs to know to eliminate a problem
the number of cells in which the expected fre- Discriminant function analysis: Used in situ-
quency would have to be computed before ations in which the dependent variable is mea-
the remaining cells could be determined by sured at the nominal or ordinal level and the re-
subtraction, given that the total expected searcher wishes to examine the impact of
must equal the total observed. several independent variables simultaneously.
Dehoaxing: The process of a researcher in- This procedure provides weightings maximiz-
forming subjects in a study about what was re- ing the likelihood of correctly predicting the
ally going on in an experiment, particularly in- category of the dependent variable each case
forming them of any deception that was used. will fall into.
Demand characteristic: A distortion intro- Domain analysis: Moving from observing a
duced when people respond in terms of how social situation (a set of behaviors carried out
they think they are expected to respond. by people in a social situation) to discovering
Deontological: A view of research ethics a cultural scene.
that proposes absolute moral structures that Double blind: In experimental designs, those
must never be violated. in which neither experimenters nor subjects
Dependability: Both the stability and the are aware of the experimental condition that is
trackability of changes in the data over time being applied to the subject.
and conditions. The issue of dependability in Double standard: In research, using differ-
qualitative investigations reects the reality ent means of measuring identical behaviors,
that situations constantly change and peo- attitudes, or situations for each gender, for ex-
ples realities differ. ample.
Dependent variable: A variable that is Dummy variable: One that is coded 1 for
viewed as being inuenced by other variables; presence and 0 for absence of a characteristic
it is the effect in a cause-and-effect relation- and represents one category of an indepen-
ship. dent variable. Dummy variables are used in
Descriptive research: Research that is about multiple regression and in discriminant analy-
what and how many of what. sis to enable the researcher to incorporate
Descriptive statistics: Statistics that include nominal variables.
various tools, conventions, and procedures Effect size: Concerned with the strength of
for describing variables or the relationship the relationship among study variables. It is a
between variables. Means, standard devia- measure of how false the null hypothesis is, or
tions, normal distributions, and Z scores are in other words, how strong the effect of the in-
used to describe individual variables; cross- dependent variable is on the dependent vari-
tabulations, means across categories, and able.
correlations are some of the procedures used Eigenvalue: The sum of the squared factor
to describe relationships between variables. loadings for any one factor in factor analysis.
704 GLOSSARY
Elite bias: A danger because informants are Evaluation case study: In-depth explorations
more likely to be drawn from the more articu- of phenomena, usually over an extended pe-
late, high-status elements in a society. Hence, riod of time, using diverse data collection pro-
unless care is taken, there is a tendency to cedures to collect detailed information about
overrepresent the views of the elite in ones an individual, group, organization, program, or
research. social phenomenon.
Emic: An insiders or natives point of view Evaluation research: A systematic appraisal
and how they derive meaning from their ex- using the methods of social research for the
periences. purpose of generating knowledge and under-
Empathetic explanation: One in which the standing that can be used for decision making.
experience of coming to see and coming to un- It is an applied form of research that provides
derstand are stressed. utilitarian answers to practical questions for
Empirical knowledge: Knowledge of the ex- decision makers.
perienced or empirical world. This knowledge Evidence-based practice: Nursing practice
is generated through scientic methods and is that is informed and modied in light of sys-
usually organized into laws and theories that tematic research that evaluates the effective-
help to describe, predict, and explain phe- ness of alternative interventions.
nomena. Expectancy: An anticipation of particular re-
Epistemology: The study of the nature of search results; this may lead to a distortion of
knowledge. It seeks to determine how we results in the direction of expectations.
know what we know. A number of authors Experiment: A scientic investigation that
have studied ways of knowing in nursing. tests cause-and-effect relationships while con-
Equivalence: The degree of agreement trolling for the inuence of other factors.
among two or more different observers using Experimental group: The group that is ex-
the same measurement tool or the agreement posed to the treatment intervention.
between two or more alternate forms of an in- Experimenter effect: A tendency to pro-
strument. duce ndings that are consistent with the ex-
Error checking: A process to locate and perimenters expectations.
correct errors in the data before data analysis. Expert sampling: A type of purposive sam-
Ethical knowing: Focuses on obligation on pling used with the Delphi technique.
what ought or should be done. This pattern of Explanatory research: Seeks to provide an-
knowing in nursing requires an understanding swers to why? questions.
of different philosophical positions regarding Exploratory factor analysis: Used when a
what is good, what is right, and what is desired. researcher does not a priori identify the factor
Ethics: The study of moral principles. structure of the variables in the data set. It is
Ethics review board (ERB): Committee that similar to stepwise regression analysis in that
is used to approve research involving humans. the variance of the rst factor identied is par-
The purpose of the review is to ensure that eth- tialed out before analysis of the second factor
ical principles are appropriately applied to re- proceeds.
search involving human subjects. Most ERBs External validity: The extent to which re-
reviewing biomedical research operate under sults may be extrapolated from a particular
the guidelines of the Declaration of Helsinki. study to other groups in general.
Ethnography: A qualitative research method Extinction: The cessation of a behavior that
that attempts to understand human behavior was once displayed.
in the cultural context in which it is embedded. F distribution: Used to test whether there is
Ethnomethodology: Typically involves a de- a signicant difference in the means of various
tailed examination of a single event or case; categories.
associated with the interpretive approach. Face validity: An evaluation of an indicator
Ethnonursing: The study and analysis of the that, on inspection, appears to reect the
local peoples viewpoints, beliefs, and prac- concept you wish to measure.
tices about nursing care. Factor analysis: A multivariate procedure
Etic: The view of the researcher or an out- used primarily to determine the structure of a
sider. set of variablesthat is, to show how variables
GLOSSARY 705
cluster together to form unidimensional con- Field notes: Attempt to capture the essence
structs. It determines the structure of a set of of the group being studied. Included are de-
variables by analyzing the intercorrelations scriptions of events and people as well as the
among them. It sorts the variables into cate- interpretations of the researcher and the par-
gories according to how closely related they ticipants.
are to other variables. Variables that are highly Field studies: Include investigations in which
correlated are grouped together into a factor. a researcher observes and records the behav-
Factorable: A correlation matrix is said to ior of individuals or groups in their natural set-
be factorable if there are reasonably robust tings.
correlations among the items. Ordinarily one File: The name for a data set, or the text of a
would expect the interitem correlations to av- report, entered into a computer.
erage above 0.25 to consider the variables to Focus group: Typically consists of 6 to 12 in-
be worthy of factor analysis; if it is any lower dividuals who are asked to discuss topics sug-
than that, you may not nd that any factors gested by a facilitator. The idea is for the re-
would be extracted. searcher to observe the interactions among
Factor extraction: The analyst seeks to de- focus group members detecting their atti-
termine the number of factors that need to be tudes, opinions, and solutions to problems
extracted in order to capture or explain the posed by the facilitator.
variation in the data set. To proceed, the ana- Folk wisdom: An important source of ideas
lyst then selects a method for how additional for the researcher are those passed down from
factors will be identied through a process of generation to generation; such ideas should
factor rotation. not, however, be accepted as true until tested
Factor loadings: The b weights (similar to the in a rigorous scientic fashion.
bs in regression analysis) in the factor analysis Formative evaluation: A form of inquiry
equation. Factor loadings may be combined to that focuses on how well a new service or on-
yield a factor score, which measures the extent going program or activity is meeting its objec-
that factor scores are similar, thus reecting tives. The thrust of formative evaluation re-
the underlying common factor (or factors). search is to identify what is and what is not
False dilemma: Occurs when a researcher working currently, so that remedial action can
argues that something is caused by either A or be taken to improve the situation at hand.
by B. Then, having provided some evidence Formal theory: Includes properties, propo-
that B is not responsible, the researcher sitions stating the relationships between
falsely concludes that A must be the cause. them, and the propositions forming a deduc-
Familism: A special case of gender insensi- tive system.
tivity that involves treating the family as the Four-variable causal models: A testable
unit of analysis when, in fact, it is individuals model containing four variables.
within the family unit that engage in a particu- Frequency distribution table: A systematic
lar activity or hold a certain attitude. Familism listing of all the values on a variable from the
is also a problem when we assume that a par- lowest to the highest with the number of times
ticular phenomenon has an equal impact on all (frequency) each value was observed.
members of the family when, in fact, it may af- Functional explanation: One in which the
fect different family members in different ways. presence of a particular phenomenon is ex-
Feminist theory: The basic premise is that plained in terms of the role it plays in main-
one cannot adequately understand human so- taining a particular system.
cieties without paying attention to the univer- Gamma: A measure of the strength of asso-
sal role of patriarchy, which refers to a domi- ciation between ordinal level variables.
nation of the social groups by men, who have Gatekeeper: An individual with special
greater power and privilege than women and group status who can lead the researcher to
children. other key informants.
Field experiment: When a researchers in- Gender insensitivity: A disregard of the dif-
tervention occurs in a natural setting rather ferential impacts of research conclusions or
than in a laboratory and is usually simple, lim- of social policy on men and women.
ited, and quickly completed. Generalizability: The extent that the nd-
706 GLOSSARY
ings may be extrapolated from the present portance of the study in terms of its potential
study to other groups in general. contributions to practice, education, theory,
Generalization: A statement that attempts and research, as appropriate.
to describe a pattern of behavior or of rela- Independent variable: A cause in a cause-
tionship that is, on average, correct. and-effect relationship. It is a variable that has
Grand theories (or conceptual models): been selected as a possible inuence on vari-
These embody the beliefs, traditions, goals, ations in a dependent variable.
and values of the discipline. They take into ac- In-depth interviews: Personal interviews in
count the phenomena of central concern to which probing is used to explore issues in de-
nursing: person, health, nursing, and envi- tail.
ronment. They broadly dene these concepts Index: A combined score based on two or
and link them together through relationship more indicators.
statements reective of the theorists view of Index system: Through a compilation of
the discipline. They explain universal relation- items, enables the researcher to create and
ships that describe what the discipline of nurs- manipulate concepts and emerging ideas.
ing is all about and remain broad, abstract rep- Individually delivered questionnaires: Ques-
resentations of reality. tionnaires handed to the respondents by a re-
Grounded theory: The idea that the conclu- searcher.
sions of a qualitative study should be grounded Inductive reasoning: Moves from the spe-
in the datathat is, based on direct and careful cific to the general; specific situations are ob-
observations of everyday life within the group. served and then combined into a larger,
Halo effect: A carry-over effect, in which a more general statement that can be tested
researchers rst rating may inuence the sec- through qualitative or quantitative research
ond and subsequent ratings. methods.
Hawthorne effect: Any variability in a de- Inferential statistics: Deal with making ex-
pendent variable that is not the direct result trapolations from a sample to the population
of variations in the treatment variable (see from which it was drawn; they also deal with
also Source of spuriousness variable and Con- tests of statistical signicance.
founding variable). Informant: Participant in an ethnographic
Health promotion research: The system- investigation.
atic investigation into the processes and sub- Informed consent: The right of a potential
stance of health promotion action. It encom- respondent to be informed as to the nature of
passes a broad range of studies, including the study, the kinds of issues will be explored,
much of the research in health policy, how the respondent was selected, and who is
lifestyle, and socio-epidemiology. sponsoring the research. When studies are
Hierarchical regression: The researcher done involving children, inrm patients, or in-
determines the order of entry, or hierarchy, of competent adults, the organization or individ-
the variables put into a regression equation. ual responsible for the prospective respon-
This technique is used in testing path models. dent should provide consent in writing.
History: In the context of experimental de- Informed opinion: An opinion based on evi-
sign, history refers to concurrent events that, dence that has been collected under con-
along with the experimental manipulation, trolled circumstances.
may be inuencing variation in the dependent Institutional review board: See Ethics re-
variable. view board.
Holistic ethnography: Denes culture as a Instrument decay: A deterioration in the
way of life and studies it as an integrated whole. measurement instrument over the course of
Hypotheses: Statements of predicted rela- measurements in a study.
tionship between two or more variables. Instrument development: A process involv-
Illegitimate appeal to authority: To argue ing application of specic rules to create a tool
that something is bad (e.g., euthanasia) by cit- to measure some research phenomenon. It is a
ing the opinion of a person who is not an ex- major use of factor analysis. Usually instru-
pert in the particular eld but who is well ment development proceeds with a large num-
known for accomplishments in another eld. ber of items used to represent the research
Implications: The author highlights the im- phenomenon later reduced to fewer items.
GLOSSARY 707
Instrument validation: Factor analysis is an one third or more, it should be interpreted as

important tool for validating new research in- an increase, or a strengthening, of the original
struments. In nursing research, it is used most relationship. If the difference remains within
frequently for this reason. Factor analysis of- one third of the original, it should be inter-
fers an empirical method of identifying the preted as an indication that the relationship
underlying factors that compose a construct. has remained the same. If the difference de-
Instrumental utilization: The direct, ex- creases by more than one third, it should be
plicit application of knowledge gained from re- interpreted as a decrease or disappearance of
search to change practice. It includes but is the relationship. Finally, if the relationship is
not limited to the adoption of nursing inter- markedly different when different control cat-
ventions, new procedures, clinical protocols egories are compared with one another (e.g.,
and guidelines, and so on. it disappears in one category but stays the
Internal consistency approach to reliability: same in the other), the result is mixed.
Involves comparing an individual items cor- Justice: The idea that all participants must
relation to the total index score. If an item is be treated fairly. The protection of partici-
consistent with the total score, it will corre- pants from incompetence and the right to re-
late with it. ceive research treatments are expectations of
Internal validity: A researcher has demon- the justice principle.
strated that the treatment does, in fact, pro- Key character les: Files established on key
duce the changes in the dependent variable. players in the organization or group being ob-
Interpretive perspective: Relies mainly on served. Here items that provide clues as to the
eld studies, with an emphasis on participant personality and manner of operation of each
observation studies ( joining a group and par- individual are drawn from the master le.
ticipating in it), in-depth interviews with peo- Key informants: Group members who are
ple, and on ethnomethodology (typically a de- most knowledgeable about the study phe-
tailed examination of a single event or case). nomenon.
Each of these studies typically involves a few Kuder-Richardson coefcient: A method
cases that are described in detail. A key ques- for testing internal consistency in an index.
tion for these researchers is, Does the expla- Lambda: A statistic measuring the propor-
nation offered make sense to the people tionate reduction in error that occurs in esti-
whose behavior is being explained? Commu- mating a dependent variable given knowledge
nication of the results of such studies usually of the independent variable.
emphasizes verbal descriptions rather than Leptokurtic distribution: Has little variabil-
numerical analyses. itya small standard deviation relative to the
Interquartile range: A measure of how far magnitude of the valuesand is sharply
apart the values are in the middle 50 percent peaked.
of a distribution. Likert-based index: An index or instrument
Intersubjectivity: Describes how subjective based on a combination of statements that
awareness and understanding can be reached use a numeric scale after each statement for
in a common world. response.
Intervening variable: A variable that links Likert items: Asking respondents to catego-
an independent variable to a dependent one. rize a statement by indicating whether they
An intervening variable represents an expla- strongly disagree, disagree, are undecided or
nation of how the independent variable inu- neutral, agree, or strongly agree with the
ences the dependent variable. statement.
Interview panels: Groups in which at least one Linear regression equation: A relationship
follow-up interview is conducted with them. between a number of independent variables
Interview schedules: Outline of the major and a dependent variable and which provides
questions that are to be raised. The inter- for the best linear (additive) weightings of the
viewer has autonomy in exploring questions independent variables and a constant calcu-
in detail. lated so as to maximize the prediction of the
Jacksons rule of thirds: In interpreting dependent variable.
three-variable relationships, if the original dif- Literature review: A critical step in focusing
ference between the categories increases by the research problem and statement of pur-
708 GLOSSARY
pose. It is a process of reviewing the current study of general and particular features of a
knowledge about the research problem, de- designated culture.
scribing the characteristics of previous stud- Mean: More formally known as the arith-
ies in the area, noting the similarities and dif- metic mean and less formally known as the av-
ferences in research results, evaluating the erage, it is computed by summing the values
strengths and limitations of previous studies, of a variable and dividing the result by the to-
and identifying gaps in knowledge relevant to tal number of cases.
the research problem. Measurement: The process of reecting ab-
Lived experiences: The everyday human stract concepts with empirical indicants or
experiences that are real to the individuals numbers.
who experience them. Measurement error: The extent to which in-
Longitudinal studies: Studies that involve dicants fail to reect the true underlying val-
data collection from the same group at differ- ues of variables.
ent points in time. Measures of central tendency: Use one
Looking glass self: The idea that our percep- number to typify a set of values.
tion of self is reected in how others see us and Measures of dispersion: The extent of the
that we come to see ourselves as others see us. variability, or dispersion, of values in a distri-
Macrovariables: Properties or characteris- bution.
tics of societies, as opposed to qualities of in- Mechanistic model: Model that asserts that
dividuals. a particular behavior is completely controlled
Magnitude estimation procedures: Proce- by one or more external actions or events.
dures that are useful when comparative judg- Median: The midpoint of a distribution.
ments are required. In this type of procedure, Meta-analysis: The statistical analysis of a
a respondent estimates the magnitude of a se- large collection of results from individual
ries of stimuli compared with a xed standard. studies for the purpose of integrating ndings.
Marker variable: A variable that produces Methodological triangulation: The applica-
the highest correlation with the underlying tion of diverse methods to generate and col-
factor that is identied in the factor analysis lect data about one phenomenon.
procedure. A marker variable is useful to re- Microvariables: Properties or characteris-
searchers attempting to label the underlying tics of individuals as opposed to properties of
factor. a society.
Master eld le: Made up of the complete Middle-range theories: Located on the the-
journal of eld notes. ory continuum midway between the most ab-
Master table: Records all collected informa- stract ideas (grand theory) and the most con-
tion so that required tables can be derived crete ideas (practice theory). They deal with
from it without going back to tally sheets. limited aspects of nursing phenomena and are
Matched pairs: The researcher predeter- more testable and generalizable than grand
mines signicant characteristics in the sam- theories. They contain well-dened concepts
ple that might inuence the dependent vari- and propositions.
able and matches the sample according to Mini ethnography: A small-scale ethnogra-
those characteristics. For example, if a re- phy focused on a narrow area of inquiry.
searcher is interested in measuring the im- Missing evidence: The idea that evidence
pact of a health promotion program on par- necessary for an argument to be true may be
ticipants, the researcher would try to match missing.
each program participant with another per- Missing value code: A value assigned to a
son with a similar predetermined characteris- variable for cases in which the information is
tics at the outset of the program. absent or not applicable.
Maturation: Any changes that occur in an in- Mode: The most frequently occurring re-
dividual subject over the course of an experi- sponse to a nominal variable.
ment which may, along with the experimental Mortality: Subjects selecting themselves
manipulation, inuence the outcome of the ex- out of a study.
periment. Multicolinearity: The extent of the correla-
Maxi ethnography: A large, comprehensive tion among the independent variables.
GLOSSARY 709
Multiple regression analysis: An equation referral (snowball) samples are examples of

to predict variations in a dependent variable nonprobability sampling techniques.
from two or more independent variables. Normal distribution: Approximates a bell-
Multistage area sample: When sampling in- shaped curve, in which there are few cases on
volves an attempt to reect a large unit such the extremes with most clustered at the mean
as a state, province, or country and no list of of the distribution.
the population is available, then one develops Normative bias: Value judgments are made
a sample by stages. At each stage of the sam- by the investigator regarding normalcy and
pling process, every individual (or unit) must the appropriateness of phenomenon-related
have a known chance of being selected. feelings and behaviors based on current con-
Multivariate analysis of variance (MANOVA): ceptual models, theories, and available em-
Used in examining relationships between a ra- pirical data.
tio level dependent variable and independent NUD*IST: A software program developed
variables measured at any level. MANOVA specically for qualitative data analysis. It
techniques also permit the researcher to ex- stands for non-numerical unstructured data
amine multiple dependent variables simulta- indexing searching and theorizing.
neously. Null hypothesis: States that there will be no
Multivariate models: Models that involve relationship between the variables.
numerous variables. Nuremberg code: Formulated as a result of
Mutually exclusive categories: Situations in unethical medical research experiments con-
which category boundaries are established so ducted by Nazi scientists predominantly on
that no category overlaps another one. inmates of concentration camps before and
Natural setting experiments: Experiments during World War II. The code addresses is-
in which everyday situations can be manipu- sues of informed consent, protection of sub-
lated, allowing for the researcher to observe jects from risk or harm, the right of partici-
reactions of people to the intervention. pants to withdraw from experimentation, and
Naturalistic observational studies: When adequate qualications of those conducting
the researcher records observations without research.
the subjects being aware that they are being Objective: Observations that are free from
observed. bias.
Needs assessment: A form of inquiry that as- Observational measurement: Frequently con-
sesses the needs, problems, concerns, or conducted in nursing research to determine how
ditions of a group, community, or organiza- participants respond in specic situations. It is a
tion that should be addressed in future data collection method that is particularly well
planning activities. suited to phenomena that are best viewed from
Nodes: Qualitative data containers for thoughts a holistic rather than a reductionistic perspec-
about a project. The nodes store the index cate- tive. For example, direct observation of a childs
gories developed by the qualitative researcher. separation anxiety behavior would most likely
Nominal measurement: A quantitative mea- yield better data than what would be achieved
sure in which the numbers are arbitrarily as- through a survey.
signed to categories of the variable. One-tailed test: Situations in which we are
Nonmalecence: The duty to not inict predicting which particular tail of the normal
harm, either emotional or physical. distribution the result will fall into if the null
Nonparallel constructions: Particularly prob- hypothesis is to be rejected.
lematic if used in gender references. They may One-way analysis of variance: Provides a
lead to confusion or misrepresentation by vio- test of signicance for the effect of one inde-
lating the principle that parts of a sentence that pendent variable on a dependent variable.
are parallel in meaning should be parallel in Open-ended question: Asks a respondent to
structure. answer a question or to offer a suggestion or
Nonprobability sampling techniques: Tech- opinion but to do so without any preset cate-
niques that do not provide potential respon- gories provided for the answer.
dents with a known chance of being asked to Operant conditioning theory: The process
participate in a study. Convenience, quota, and by which behavior is modied by the rewards
710 GLOSSARY
and punishments that result from the behav- and a client. It is concerned with the knowing
ior being displayed. and encountering of the individual self.
Operationalization: The selection of indica- Phenomenological approach: A qualitative
tors (measures) to reect conceptual vari- research method that describes the meaning of
ables used in the implementation of a re- a lived experience from the perspective of the
search project. participant. Lived experiences are the everyday
Operational level: Consists of the measure- human experiences that are real to the individ-
ment of variables as well as the collection and uals who experience them. Phenomenology
analysis of data. seeks to achieve a deep understanding of the
Ordinal measurement: Consists of a mea- phenomenon being studied through a rigorous,
sure in which there is an order in the values systematic examination of it. Its purpose is to
assigned but in which the intervals between describe the essences of lived experiences.
the values are not equal. Phone survey: Survey that relies on infor-
Overgeneralization: A statement that claims mation reported by the respondent over the
to refer to all people but is based on evidence telephone.
that does not represent all people. Pilot study: In survey research, a study that
Overspecicity: Involves situations in which involves having a small sample of respon-
single gender terms are used when members of dents complete a questionnaire or undergo an
both genders are involved (e.g., the doctor interview. Pilot studies are used to determine
. . . he). items to be included in indexes and to deter-
Panel studies: Studies that monitor specic mine, from open-ended questions, the cate-
organizations or individuals over time. gories that should be used in a xed choice
Partial correlation: Measures the strength format.
of association between two ratio level vari- Plagiarism: The unacknowledged borrow-
ables while simultaneously controlling for the ing of other authors ideas or words.
effects of one or more additional variables. Planned change: An intentional process
Partial theory: Explains an assumed or that involves a change agent who works to
known relationship by specifying a testable bring about a demonstrable improvement in
causal model. the status quo.
Participant: Used to refer to the individuals Platykurtic distribution: Has a great deal of
who inform the study in qualitative investiga- variability that tends to be at and wide.
tions. These individuals are viewed as active Population: A collection of individuals, com-
participants in the research process and as munities, or nations about which one wishes to
equal partners with the investigator. make a general statement.
Participant observation studies: Those in Positivist approach: The approach used in
which the researcher is involved in the daily the physical sciences. It emerged from a
lives of the study group, observing and record- branch of philosophy known as logical posi-
ing as much as possible about their lives. tivism, which operates on strict rules of logic,
Participatory action research: Uses criti- truth, axioms (general principles), and pre-
cal social theory methods; it is research for dictions.
the purpose of taking action and creating Power analysis: A procedure used to deter-
change. mine the power of a statistical test; in other
Path model: A graphic representation of a words, to determine the probability that an in-
complex set of proposed interrelationships ferential statistical test will detect a signi-
among variables. cant difference that is real or correctly reject
Patriarchy: The domination of the social a null hypothesis.
groups by men, who have greater power and Practice theories: Theories that are nar-
privilege than women and children. rower in scope and more specic than middle-
Percentage: Represents a proportion multi- range theories. Concepts are specically and
plied by 100; thus, a percentage represents narrowly dened and readily measured. Prac-
how many for every 100 of something. tice theory propositions produce clear direc-
Personal knowledge: The interpersonal in- tives for application in practice. Much of what
teractions and relationships between a nurse nurses follow today in hospital procedure
GLOSSARY 711
manuals is actually practice theory that has that a particular case will be similar to others
been developed inductively by expert nurses in the same general category.
who have observed and described ways of Probability levels: The test statistic com-
practicing based on trial and error. putes the chance that a result is caused by
Precise communication: Unambiguous in- sampling uctuations. This is reported in var-
formation. ious ways, normally labeled as the p value or
Precision matching: A method of achieving signicance value.
equivalence between control and experimen- Probability sampling procedures: Involve
tal groups by ensuring that the groups are techniques for selecting sampling units so that
matched on certain key variables. each has a known chance of being included.
Precoded, single-choice questions: Ques- Problem statement: A narrative that elabo-
tions that ask a respondent to indicate which rates on the research problem and identies
category applies to him or to her. the specic area of concern. It provides direc-
Prediction: In regression analysis, the ex- tion for the entire study and guides the study
tent to which variation in a dependent vari- toward a quantitative or a qualitative design.
able can be accurately estimated with knowl- Process consent: The researcher renegoti-
edge of the independent variables. ates the consent with the participants as un-
Predictive research: Moves beyond expla- foreseen circumstances arise. This allows the
nation to the prediction of precise relation- participants to be part of the decision-making
ships between dimensions or characteristics process as the study unfolds.
of a phenomenon or differences between Pronoun problem: Encourages stereotypic
groups. Predictive research typically engages thinking by referring to doctors and managers
the empirical method of experimentation. as he and nurses as she.
Predictive validity: If you were attempting to Proportion: Represents the part of one that
develop a measure to predict success in nurse is represented by a category (e.g., how many
registration examinations, you could assess females there are in a population compared
the validity of your measure by correlating it with the total population). Proportions al-
with success rates on RN examinations. A high ways begin with a decimal point.
correlation would indicate high predictive va- Proportionate reduction in error: Associ-
lidity. ated with the statistic Lambda, which mea-
Pre-experimental design: A design that does sures the reduction in error in estimating vari-
not permit clear causal inferences about the im- ation in a dependent variable, given knowledge
pact of a treatment on the dependent variable. of an independent variable. If two variables are
Presenceabsence questions: Questions that strongly associated, then errors in predicting
request respondents to check off which items variations in the dependent variable will be
in a list do or do not apply to them. considerably reduced if information on the in-
Primary sampling units: May be census tract dependent variable is taken into account.
areas, or other similar units, normally several Propositions: Statements of proposed rela-
hundred of them; these units are numbered tionships between two or more concepts in a
and, using an equal probability technique, a se- theory. Propositions link concepts of a theory
lection of at least 30 units is made from them. together so that something can be described,
Principal components method: Includes an explained, or predicted. Propositions may
analysis of all the variance in the observed also be referred to as axioms or theorems.
variables, not just the shared variance among Prospective: Reference to data that are col-
the variables in factor analysis. lected from the present to a future point in time.
Principal factors method: Known in the Sta- Provincialism: The tendency to see things
tistical Package for the Social Sciences (SPSS) as ones culture sees them.
as the principal axis factoring method, this ap- Proxemics: The study of the norms sur-
proach relies on the shared variances of the rounding personal space and the conditions
variables included in the analysis and appears under which such space will or will not be vi-
to be the most commonly used factor analysis olated.
approach. Pure research: Research that focuses on un-
Probabilistic explanation: Involves arguing derstanding phenomena of interest.
712 GLOSSARY
Purposive sampling: Selecting and interview- Randomization: A process of assigning sub-

ing or observing participants who have had ex- jects to a treatment or a control group so that
perience with the phenomena of interest. each subject has an equal chance of being as-
Qualitative research: Research that uses signed to either group.
concepts, classications, and attempts to in- Randomized clinical trial: Uses one or more
terpret human behavior that reects not only control groups and experimental groups, de-
the analysts views but the views of the people pending on the number of interventions. Sub-
whose behavior is being described; the em- jects are randomly allocated to the groups.
phasis is on verbal descriptions as opposed to Range: Indicates the gap between the low-
numerical ones. est and highest value in a distribution.
Quality control monitoring system: Proce- Rank-ordering questions: Questions in which
dures to ensure that interviewers are following a respondent is asked to indicate an ordering of
established procedures for selecting respon- response items, usually from most preferred to
dents, asking questions, and entering data. least preferred.
Quantitative research: Seeks to quantify, or Rate: The frequency of a phenomenon for a
reect with numbers, observations about hu- standard sized unit (such as incidence per
man behavior. 1000 or per 100,000).
Quasi-experimental design: Design in which Ratio measurement: Quantitative measure-
it has not been possible to do any or all of the ment in which intervals are equal and there is
following: (1) randomly assign subjects to a a true zero point.
treatment or control group or (2) control the Ratios: Used to compare rates or other mea-
timing or nature of the treatment. The quasi- sures across categories.
experiment design comes as close as possible Reexive critique: A process that enables
to the experimental design in order to measure participants and researchers to make explicit,
the impact of a treatment. alternative explanations for events or experi-
Questionnaire: Made up of a series of set ences. It facilitates discussion between the re-
questions that either provide a space for an searchers and participants and leads to
answer or offer a number of xed alternatives greater insights and acceptance of multiple
from which the respondent makes a choice. explanations for events.
Quota sample: Respondents are selected on Regression analysis: A method for analyz-
the basis of meeting certain criteria. The rst ing the relationship between a ratio level de-
respondent to meet the requirement (or re- pendent variable and independent variables.
quirements) is asked to participate and sam- This form of analysis provides weightings that
pling continues until all the categories have may be used in an equation to describe the re-
been lleduntil the quota for each has been lationship. Standardized weightings provide a
reached. means for estimating the relative impact of in-
r2: A measure of the amount of variation in dependent variables on the dependent one.
the dependent variable is explained by an in- Regression line: A straight line describing
dependent variable. the relationship between an independent and
R2: A measure of the amount of variation in dependent variable drawn so that the vertical
the dependent variable that is explained by deviations of the points above the line equal
the combination of independent variables (re- the vertical deviations below the line.
gression analysis). Reliability: The extent to which, on re-
Random error: Inconsistencies that enter peated measures, an indicator yields similar
into the coding process but that have no pat- readings.
tern in the errors. Reliable knowledge: Knowledge you can
Random variable: Varies without control count on; knowledge that allows you to pre-
but is taken into account by the way groups dict outcomes.
are set up in an experiment. Repeated measures analysis of variance:
Random sample: Provides each unit (usually To test for differences among the mean scores
a person) in the population with an equal of the treatment conditions when the re-
chance of being selected for participation in a searcher has exposed the same subject to two
study. or more treatment conditions.
GLOSSARY 713
Replacement of terms: Replacing general the problem in the interrogative form is that a
theoretical concepts by specic instances of question invites an answer.
these concepts. Research report: A comprehensive descrip-
Research: The systematic inquiry into a tion of key aspects of the project that includes
subject to discover new knowledge or to vali- a minimum of four major sections: Introduc-
date or rene existing knowledge. tion, Methods, Results, and Discussion.
Research bias: Dened as systematic distor- Research topic: The broad general area you
tions in research outcomes. wish to investigate.
Research critic: The role is to provide an ob- Research utilization: A complex process that
jective evaluation of the studys merits. To do includes critically analyzing the literature, se-
this, the critic must read the report compre- lecting appropriate interventions, implement-
hensively and offer constructive comments ing them, and evaluating the outcome.
on all aspects of the report, as well as on the Response bias: Exhibited if respondents who
specic strengths and limitations of the study. want to appear consistent give the same re-
Research critique: A critical appraisal of a sponses as they did the rst time; others might
piece of completed research. It involves a high want to help the researchers and, suspecting
level of critical thinking and analysis of each the study is to demonstrate how good some-
component of the research study. Critiques thing is, respond by being more positive after
may be conducted on research reports, manu- the experimental stimulus has been given.
scripts, or published articles based on research Response rate: The percentage of delivered
studies, or they may be done on research pro- questionnaires that are completed and re-
posals. A critique provides constructive criti- turned.
cism of a specic piece of work for the purpose Response set: A situation in which a respon-
of promoting excellence in research. dent tends to answer similarly to all items.
Research design: Similar to a blueprint for a Retrospective: Reference to data collection
study, it guides an investigator in planning and about behavior that occurred in the past.
implementing a study. It provides a detailed Review of the literature: A section of a pa-
plan for data collection and analysis and is the per that tries to provide an overview of the
critical element linking the theoretical frame- state of scientic knowledge on the topic be-
work and questions with the resultant data. ing researched.
Researcher affect: Reference is made to a Role modeling: Emulating someone elses
process whereby the researcher, having attitudes and behaviors.
fallen in love with a particular explanation Salience: The degree of interest in the topic
for a particular relationship or view of the of the research by the respondent; the greater
world, may inadvertently engage in proce- the salience of the topic to respondents, the
dures that lead to conclusions supporting the greater the response rate.
preferred explanation. Sample: Selected from a population and in-
Research hypothesis: The statement of rela- terpreted to represent that population.
tionship between variables. Sampling error: The difference between the
Researchpractice gap: The gap between values emerging from the sample and those
what we know in the eld of nursing as a result emerging from the population on a particular
of research and the actual application of this phenomenon.
knowledge in practice. Sampling fraction: The sample size in rela-
Research problem: A situation or circum- tion to the population.
stance that requires a solution to be de- Sampling frame: The list from which a sam-
scribed, explained, or predicted. ple is drawn.
Research process: The research process is Sample interval: See Skip interval.
simply a series of actions taken by the re- Saturation: Used in qualitative investiga-
searcher to discover new knowledge or vali- tions, situation in data collection in which the
date or rene existing knowledge. participants descriptions become repetitive
Research question: The research problem and conrm previously collected data.
stated in the interrogative form. It is stated in Secondary data: Information collected by
the present tense. The advantage to stating persons other than the researcher.
714 GLOSSARY
Scalable items: Items in which there is a hi- coherent questions about situations that are
erarchy among them. This means that when limited in scope and focus.
items are arranged from low intensity to high Skip interval: Determined by dividing the
intensity, if a person responds positively to an total sample requirement into the total num-
item at the high end, they are expected to re- ber of units in the population being surveyed;
spond positively to all lower intensity items. this number should then be rounded to the
Scale: A complex combination of indicators nearest, but lower, whole number.
in which the pattern of the responses is taken Snowball sampling: A name for a referral
into account. sampling procedure. As a researcher com-
Scientic method: Viewed in Western cul- pletes one interview, he or she asks if there is
tures as the most sophisticated method of ac- anyone else known to the respondent who
quiring knowledge, it combines the process of might be appropriate for the study.
logical reasoning with systematic planned in- Source of spuriousness variable: A variable
vestigation, data collection, analysis, and eval- that is viewed as possibly inuencing both the
uation. independent and the dependent variable in
Selected evidence: Evidence that results such a way that it accounts for the relation-
when one chooses to report only the studies ship between them.
that support a particular point of view, ignor- Spearman correlation: Statistic used to
ing the evidence that runs counter to what the measure the strength of association between
researcher is attempting to demonstrate. two ordinal level variables.
Selection: Subjects selecting themselves Split-half method: A method for testing relia-
into a study. bility involves randomly splitting index items
Semantic differential: Items are used to into two groups, computing the indexes, and
study subjective feelings toward objects or then correlating the resulting scores. Internal
persons. reliability would be indicated by a high corre-
Sexism: Discrimination on the basis of gen- lation between the two indexes.
der. SPSS (Statistical Package for the Social Sci-
Sexist language: Problems that include the ences): A collection of procedures for pro-
pronoun problem, the man problem, and the cessing social science and nursing data.
nonparallel construction problem. Stability: Concerned with the consistency of
Sexual dichotomism: Treating the genders the results with repeated measures.
as discrete social, as well as biological, co- Standard deviation: The average amount of
horts rather than two cohorts with shared deviation from the mean value of the variable.
characteristics. Standard error of the means: A measure of
Signicant others: Help dene the world for variability in the means of repeated samples.
individuals by serving as models for attitudes Statement of purpose: Flows from the prob-
and behavior. lem statement and is included in it. One re-
Simple random sample: Provides each unit search problem statement may be the basis
(usually a person) in the population with an for several research purposes. The research
equal chance of being selected for participa- purpose is usually a single declarative state-
tion in a study. ment that focuses the study and clearly iden-
Simultaneity paradigm: Reality is char- ties what the researcher intends to do.
acterized by a mutual process of human- Statistical regression: Demonstrated when
environment interaction. Persons exist in a sample is selected on the basis of extreme
open participation with the universe and are scores and retesting shows a tendency to-
more than and different from the summed ward less extreme scores.
parts studied in the totality paradigm. The Stratied sample: A sample that gives indi-
wholeness or unitary nature of the human be- viduals within designated categories an equal
ing is primary in this paradigm. chance of selection.
Single blind: Study in which the subjects Structural equation modeling: An ad-
are not aware of the type of intervention that vanced statistical method designed to test
will be administered. theories.
Situation-specic theory: Answers a set of Structured interviews: Face-to-face inter-
GLOSSARY 715
views in which questions are read to the re- theoretical framework describes the basic
spondents. Such interviews ordinarily do not structure of ideas (theories, concepts, propo-
provide for in-depth probes on some of the sitions) within which the study is to be con-
questions. ducted and the results interpreted. It is a ver-
Summary tables: Tables that compress the bal description of existing theory relevant to
results of many analyses into one table. the phenomena under investigation based on
Summated rating indexes: Indexes in which a review of the literature.
the scores for each item are summed, or Theoretical memos: The ideas a researcher
summed and averaged, to yield each individ- holds about codes and the relationships as
uals score. they strike him or her during analysis.
Summative evaluation: Conducted to deter- Theoretical level: The most abstract, gen-
mine the effectiveness, value, and worth of an eral conceptualization of the research prob-
innovation. For many evaluation projects, it lem.
includes evaluation of the costs as well as the Theoretical substruction: A process of dis-
effectiveness of the program or intervention. covery in which a researcher begins with ob-
Surveys: A method for collecting informa- servations and tries to make sense of them
tion by having respondents complete a ques- and to identify the concepts the data seem to
tionnaire. reect. As observations are made, the con-
Symbolic interactionist theory: Pays atten- cepts are continuously identied and rened.
tion particularly to how ones self-concept is Theoretical triangulation: A strategy for the-
formed. People develop a sense of self that is ory testing that involves the testing of two com-
inuenced by how others see them and by peting or alternative theories within one study.
how others react to them. Theory: A systematic vision of reality that
Systematic errors: Errors that distort the describes, explains, or predicts something.
data in one particular direction. Theory development: Factor analysis is used
Systematic sample: Provides each unit to identify structures of variables or constructs
(usually a person) in the population an equal that can be interpreted meaningfully by the re-
chance of being selected for participation in a searcher. The researcher explains why vari-
study by choosing every nth unit, starting ran- ables that have been dened mathematically
domly. are clustered together the way that they have
t Test: A test of signicance usually used emerged. It can also lead to the conceptualiza-
with small samples. tion of research phenomena and their underly-
Tally sheets: Used to record information ing dimensions and show important intercor-
during the data collection phase of observa- relations and relationships among constructs.
tional studies. These are important steps in theory develop-
Taxonomy: A set of categories organized on ment.
the basis of a single semantic relationship. 3M approach: A method used to identify
Test of signicance: Reports the probability which SPSS procedure to use; the Model/
that an observed association or difference is Measurement/Method.
the result of sampling uctuations and not re- Totality paradigm: Reality exists indepen-
ective of a real difference in the population dent of the knower, humans are viewed as the
from which the sample has been taken. sum of their parts, persons adapt to an exter-
Testing: When questions are repeated, nal environment through cause-and-effect re-
some respondents may want to appear con- lationships, and health is viewed along a con-
sistent and, therefore, give the same re- tinuum from illness to wellness.
sponses in both tests. Transferability: Concerned with the gener-
Themes: Recurrent ideas or patterns that alizability or ttingness of study ndings to
emerge in the data representing common other settings, populations, and contexts. In
threads of meaning in the narrative of partici- assessing transferability of ndings, the re-
pants. search consumer hopes to show that the re-
Theoretical frameworks: Brief descriptions sults are not context bound.
or explanations of a theory or portions of a Treatment level: The number of different
theory to be tested in a research project. A categories of a variable that will be exposed to
716 GLOSSARY
the subject in an experiment. A study, for ex- Unwarranted conclusion: An error in rea-
ample, that when using three treatment levels, soning; confusing correlation and cause is an
might compare the effects of seeing a short, example of this type of error.
medium, or long lm. Utilitarian perspective: Ethical judgments
Treatment variable: Variable whose effect about a research project should be made by
on some dependent variable is being assessed evaluating its consequences for the partici-
in an experiment. pants, for society, and for the academic disci-
Triangulation: Using a variety of techniques pline. This view entails the belief that the good
to test research questions. of a research project is dened by the conse-
True value: The underlying exact quantity quences of the results.
of a variable at any given time. Validity of a measurement: The extent that
Trustworthiness: The process of establish- a measure reects the concept, reecting
ing the validity and reliability of qualitative re- nothing more or nothing less than that im-
search. plied by the conceptual denition.
Two-tailed test: Situations in which the re- Variables: Characteristics of research con-
searcher has not predicted into which partic- cepts that are directly measurable and that
ular tail of the normal distribution the result vary.
will fall into if the null hypothesis is to be re- Variance: The average amount of deviation
jected. from the mean value of the variable; it is the
Two-way analysis of variance: A multifactor standard deviation squared.
analysis of variance, when the design has two Veriable: Information that could be con-
or more independent variables. rmed by tests conducted by others.
Type I error: Results when a researcher re- Verstehen: A German word meaning the em-
jects a null hypothesis that should be ac- pathetic understanding of behavior.
cepted. Visual analog scale: Measures the intensity
Type II error: Results when a researcher ac- of participants sensations and feelings about
cepts a null hypothesis that should be re- the strength of their attitudes, beliefs, and
jected. opinions about specic stimuli such as pain,
Unilineal model: A model in which the same fatigue, nausea, quality of life, health status,
patterns of development are followed by all appetite, self-care ability, and so on.
societies. Within-subject design: Study design that ex-
Unit of analysis: The basic type of object un- poses one subject to the different experimen-
der investigation. Usually researchers dene tal treatments; because the subject is the same
the unit and deal with only one unit at a time; person, background characteristics, attitudes,
it may be a speech, it may be a paragraph from and intelligence are all perfectly controlled.
the speech, or it may be a number of individu- Z score: Expresses an observations loca-
als or a number of groups. The information col- tion relative to the mean ( in standard devia-
lected describes the unit under investigation. tion units) within a normal distribution.
Bibliography
Aamodt, A.M. ( 1991). Ethnography and epis- American Psychological Association (1994).
temology: Generating nursing knowledge. Publication Manual of the American Psycho-
In J.M. Morse, (Ed.), Qualitative Nursing Re- logical Association (4th ed.). Washington, D.C.
search (pp. 4053). Newbury Park, CA: Anderson, N. (1961). The Hobo. Chicago: Phoe-
Sage. nix Edition, [1923].
Abrahamson, M. (1983). Social Research Meth- Anderson, N.L. (1996). Decisions about sub-
ods. Englewood Cliffs, N.J.: Prentice-Hall. stance abuse among adolescents in juve-
Aish, A.E. & Isenberg, M. (1996). Effects of nile detention. Image: Journal of Nursing
Orem-based nursing interventions on nu- Scholarship, 28(1), 6570.
tritional self-care myocardial infarction pa- Arathuzik, D., & Aber, C. (1998). Factors asso-
tients. International Journal of Nursing Stud- ciated with national council licensure ex-
ies, 33(3), 259270. amination: Registered nurse success. Jour-
Alcock, D., Lawrence, J., Goodman, J., & Ellis, nal of Professional Nursing, 14(2), 119126.
J. (1993). Formative evaluation: Implemen- Asch, D.A. (1996). The role of critical care
tation of primary nursing. Canadian Journal nurses in euthanasia and assisted suicide.
of Nursing Research, 25(3), 1527. The New England Journal of Medicine, May
Alliex, S. (1998). Hurdling: The process of nurse- 23, 13741379.
patient interaction in the presence of tech- Atkinson, P., & Hammersley, M. (1994). Ethnog-
nology. Qualitative Health Research Con- raphy and participant observation. In N.K.
ference, Hotel Vancouver, Vancouver, BC. Denzin & Y.S. Lincoln, (Eds.), Handbook of
February, 1998. Qualitative Research (pp. 248261). Thou-
Alpers, R. (1998). The changing self-concept of sand Oaks, CA: Sage.
pregnant and parenting teens. Journal of Azen, I. (1985). From intentions to action: A
Professional Nursing, 14(2), 111119. theory of planned behavior. In J. Kuhl & J.
American Nurses Association (1995). Code for Beckman (Eds.), Action Control: From Cog-
nurses with interpretive statements. Wash- nition to Behavior (pp. 1139). New York:
ington, DC: American Nurses Association. Springer.
American Nurses Association (1989). Commis- Azen, I. (1991). The theory of planned behav-
sion on nursing research: Education for prep- ior. Organizational Behavior and Human
aration in nursing research. Kansas City, Decision Processes, 50, 179211.
MO: Author.
American Nurses Association (1985). Human Babbie, E. (1992). The Practice of Social Re-
Rights Guidelines for Nurses in Clinical and search, (6th ed.). Belmont: Wadsworth.
Other Research. Kansas City, MO: Author. Bailey, P., Maciejewski, J., & Karen, I. (1993).
American Nurses Association (1976). Prepara- Combined mother and baby care: Does it
tion for Nurses for Participation in Research. meet the needs of families? Canadian Jour-
Kansas City, MO: Author. nal of Nursing Research, 25(3), 2939.
American Psychological Association (1998). Baker, C. (1996). Subjective experience of symp-
Publication Manual of the American Psycho- toms in schizophrenia. Canadian Journal of
logical Association (5th ed.). Washington Nursing Research, 28(2), 1937.
D.C.: American Psychological Association. Banik, B.J. (1993). Applying triangulation in
717
718 BIBLIOGRAPHY
nursing research. Applied Nursing Research, Canadian Journal of Nursing Research, 27(4),
6(1), 4752. 1333.
Banks-Wallace, J. (1998). Emancipatory poten- Berman, H., McKenna, K., Arnold, C., & Mac-
tial of storytelling in a group. Image: Journal Quarrie, B. (2000). Sexual harassment:
of Nursing Scholarship, 30(1), 1721. Everyday violence in the lives of girls and
Barnsley, J., & Ellis, D. (1992). Research for women. Advances in Nursing Science, 22(4),
change. Participatory action research for com- 3246.
munity group. Vancouver, BC: Womens Re- Bertrand, L.D., & Abernathy, T.J. (1993). Pre-
search Center, 101-2245 West Broadway. dicting cigarette smoking among adoles-
Baron, S.W. (1989). The Canadian West Coast cents using cross-sectional and longitudi-
punk subculture: A eld study, Canadian nal approaches. Journal of School Health,
Journal of Sociology, 14(3), 289316. 63(2), 98103.
Barragan, L. (1998). Nursing practice draws Betz, C.L., & Beal, J. (1996). Use of nursing
upon several different ways of knowing. models in pediatric nursing research: A
Journal of Clinical Nursing, 7, 209217. decade of review. Issues in Comprehensive
Beal, J. (2000). A nurse practitioner model of Pediatric Nursing, 19, 153167.
practice. The American Journal of Maternal Blalock, H.M., Jr. (1964). Causal Inference in
Child Nursing, 25(1), 1824. Nonexperimental Research. Chapel Hill: The
Beauchamp, T.L., & Childrens, J.F. (1989). Prin- University of North Carolina Press.
ciples of Premedical Ethics (3rd ed.). Oxford: Blalock, H.M., Jr. (1979). Social Statistics.
Oxford University Press. Toronto: McGraw-Hill.
Beck, C.T. (1996). Grounded theory: Overview Blishen, B.R., & McRoberts, H. (1976). A revised
and application in pediatric nursing. Issues socioeconomic index for occupations in
in Comprehensive Pediatric Nursing, 19, 115. Canada, Canadian Review of Sociology and
Beck, C.T. (1994). Reliability and validity is- Anthropology, 13(1), 7179.
sues in phenomenological research. West- Blumer, H. (1969). Symbolic Interactionism: Per-
ern Journal of Nursing Research, 16(3), spective and Method. In W. Chenitz and J.
254267. Swanson, From Practice to Grounded Theory:
Beck, C.T. (1993). Qualitative research: The Qualitative Research in Nursing. Menlo Park,
evaluation of its credibility, ttingness, and CA: Addison-Wesley.
auditability. Western Journal of Nursing Re- Borenstein, M., & Cohen, J. (1988). Statistical
search, 15(2), 263265. Power Analysis: A Computer Program. Hills-
Becker, H.S., Blanche, G., Hughes, E.C., & dale, NJ: Erlbaum Associates.
Strauss, A.L. (1961). Boys in White. Chicago: Bostrom, J., & Wise, L. (1994). Closing the gap
The University of Chicago Press. between research and practice. Journal of
Beecher, H.K. (1966). Ethics and clinical re- Nursing Administration, 24(5), 2227.
search. New England Journal of Medicine, Bottomore, T.B., & Maximilien, R. (1988). Karl
274(24), 13541360. Marx: Selected Writings in Sociology and So-
Begley, S. (1993). The meaning of junk, News- cial Philosophy. London: Penguin Books.
week (March 23), 6264. Boutlier, M., Mason, R., & Rootman, I. (1997).
Bell, L. (1995). Effective Writing: A Guide for Community action and reective practice
Health Professionals. Toronto: Copp Clarke. in health promotion research. Health Pro-
Bellack, J.P., Graber, D.R., ONeil, E.H., Mu- motion International, 12(1), 6978.
sham, C., & Lancaster, C. (1999). Curricu- Bradburn, N.M. & Sudman, S. (1980). Improving
lum trends in nurse practitioner programs: Interview Method and Questionnaire Design.
Current and ideal. Journal of Professional San Francisco: Jossey-Bass.
Nursing, 15(1), 1528. Bramwell, L., & Hykawy, E. (1999). The Delphi
Benner, P. (1984). From Novice to Expert. Menlo technique: A possible tool for predicting fu-
Park, CA: Addison-Wesley. ture events in nursing education. The Cana-
Bennett, J.A. (1997). A case for theory triangu- dian Journal of Nursing Research, 30(4),
lation. Nursing Science Quarterly, 10(2), 97 4759.
102. Brandt, A.M. (1978). Racism and research: The
Berg, B.L. (1995). Qualitative Research Meth- case of the Tuskegee syphillis study. Hast-
ods for the Social Sciences. Boston: Allyn ings Center Report, 8(6), 2129.
and Bacon. Brent, E. E., Scott, J. K., & Spencer, J.C. (1988).
Berland, A., Whyte, N.B., & Maxwell, L. (1995). Ex-Sample TM: An Expert System to Assist in
Hospital nurses and health promotion. Designing Sampling Plans. Users Guide and
BIBLIOGRAPHY 719
Reference Manual, Version 2.0. The Idea Canadian Nurses Association (1994). Ethical
Works Inc., 100 West Briarwood, Columbia: Guidelines for Nurses in Research Involving
MI, 65203. Human Participants. Ottawa: Author.
Brentro, M., & Hegge, M. (2000). Nursing faculty: Canadian Nurses Association (1997). Code of
One generation away from extinction? Jour- Ethics for Registered Nurses, Ontario: CNA.
nal of Professional Nursing, 16(2), 97103. Caplan, A.L., Edgar, H., King, P.A., & Jones, J.H.
Brink, P., & Wood, M. (1998). Advanced Design (1992). Twenty years afterThe legacy of
in Nursing Research (2nd ed.). Thousand the Tuskegee syphilis study. Hastings Cen-
Oaks CA: Sage. ter Report, 22(6), NovemberDecember.
Brockopp, D.Y., & Hastings-Tolsma, M.T. (1995). Carmines, E.G., & Zeller, R.A. (1979). Reliabil-
Fundamentals of Nursing Research (2nd ed.). ity and Validity Assessment. Beverly Hills:
Boston, MS: Jones & Bartlett. Sage.
Brown, M.N., & Keeley, S.M. (1990). Asking the Carper, B.A. (1978). Fundamental patterns of
Right Questions: A Guide to Critical Think- knowing in nursing. Advances in Nursing
ing. Englewood Cliffs, NJ: Prentice Hall. Science, 1(1), 1323.
Brown, S.J. (1999). Knowledge for Health Care Carper, D.A. (1992). Response to perspectives on
Practice: A Guide to Using Research Evi- knowing: A model of nursing knowledge (pp.
dence. Philadelphia, PA: W.B. Saunders. 297299) In Leslie H. Nicoll (Ed.) Perspectives
Brundt, J.H., Lindsey, E., & Hopkinson, J. (1997). on Nursing Theory (2nd ed). Philadelphia: Lip-
Health promotion in the Hutterite commu- pincott.
nity and the ethnocentricity of empower- Carruth, A.K. (1996). Development and testing
ment. Canadian Journal of Nursing Research, of the caregiver reciprocity scale. Nursing
26(1), 1729. Research, 45(2), 9297.
Bunkers, S.S., Petardi, L.A., Pilkington, F.B., & Centers for Disease Control and Prevention
Walls, P. (1996). Challenging the myths sur- (CDCP) (1999). HIV/AIDS Surveillance Re-
rounding qualitative research in nursing. port, 10(1), 139.
Nursing Science Quarterly, 9(1),3337. Chalmers, K., Bramadat, I., & Sloan, J. (1997).
Bunn, H., & OConnor, A. (1996). Validation of Development and testing of the primary
client decision making instruments in the health care questionnaire (PHCQ). Canadian
context of psychiatry. Canadian Journal of Journal of Nursing Research, 29(1), 7996.
Nursing Research, 28(3), 1329. Chenitz, W.C., & Swanson, J. (1986). From Prac-
Bunting, S. M., & Campbell, J.C. (1994). Through tice to Grounded Theory: Qualitative Re-
a feminist lens: A model to guide nursing research in Nursing. Menlo Park, CA: Addison
search. In. Chinn, P. (Ed.). Advances in Meth- Wesley.
ods of Inquiring for Nursing. Gaithersburg, Cherry, B., & Jacob, S. (1999). Contemporary
Maryland: Aspen Publications. Nursing: Issues Trends and Management.
Bunton, R., & MacDonald, G. (1992). Health Philadelphia: Mosby.
Promotion: Discipline and Diversity. Lon- Cheyn, E. (1972). The effect of spatial and in-
don: Rortledge Press. terpersonal variables on the invasion of
Burns, N., & Grove, S.K. (1999). Understanding group controlled territories, Sociometry,
Nursing Research (2nd ed.), Philadelphia: 477488.
Saunders. Chinn, P.L., & Kramer, M. (1991). Theory and
Burns, N., & Grove, S.K. (1997). The Practice of Nursing: A Systematic Approach (3rd ed.).
Nursing Research: Conduct, Critique & Uti- St. Louis, MI: Mosby.
lization. Philadelphia: W.B. Saunders. Chinn, P.L., & Kramer, M. (1995). Theory and
Butler, L. (1995). Valuing research in clinical Nursing: A Systematic Approach, (4th ed.).
practice. Canadian Journal of Nursing Re- St. Louis, MI: Mosby.
search, 27(4), 3349. Chow, J.D. (1999). Interruption to research de-
Butler, L., & Ginn, D. (1998). Canadian nurses sign: Substance driven research. Advances
views on assignment of publication credit in Nursing Science, 22(2), 3947.
for scholarly and scientic work. Canadian Clarke, E.G., & Danbolt, N. (1955). The Oslow
Journal of Nursing Research, 30(1), 171 study of the natural history of untreated
185. syphilis. Journal of Chronis Diseases, 2, 311
344.
Campbell, D.T., & Stanley, J.C. (1966). Experi- Clark, D., Clark, P., Day, D., & Shea, D. (2000).
mental and Quasi-Experimental Designs for The relationship between health care re-
Research. Chicago: Rand McNally. form and nurses interest in union repre-
720 BIBLIOGRAPHY
sentation: The role of workplace climate. Practice: A Primer for Health Professionals.
Journal of Professional Nursing, 16(2) 92 London: Churchill-Livingston.
96. Deagle, G., & McWilliam, C. (1992). Developing
Ciliska, D. (1998). Evaluation of two nursing Healthy and Supportive Community Environ-
interventions for obese women. Western ments: A Strategy for Action. Presented at
Journal of Nursing Research, 20(1), 119 Prevention Congress V. London, Ontario,
135. 27 April.
Clairmont, D.H., & Jackson, W. (1980). Seg- Declaration of Helsinki (1986). In R.J. Levive
mentation and the Low Income Blue Collar (Ed.), Ethics and Regulations of Clinical Re-
Worker: A Canadian Test of Segmentation search (2nd ed., pp. 427429). Baltemore-
Theory, Halifax: Institute of Public Affairs, Munide Urban & Schwarzenberg.
Dalhousie University. Deets, C. (1990). Nursings paradigm and a
Cobb, G.W. (1999). Introduction to Design and search for its methodology. In N. Chaska
Analysis of Experiments. New York: Springer. (Ed.) The Nursing Profession: Turning Points,
Cobb, A.K., & Hagemaster, J. (1987). Ten crite- St. Louis, MO: Mosby.
ria for evaluating qualitative research pro- Deets, C. (1998). NursingA maturing disci-
posals. Journal of Nursing Education, 26(4), pline? Journal of Professional Nursing, 14(2),
138143. 65.
Coffey, A., & Atkinson, P. (1996). Making Sense of Deloughery, G.L. (1995). Issues and Trends in
Qualitative Data. Thousand Oaks, CA: Sage. Nursing (2nd ed.). St. Louis, MI: Mosby.
Cohen, J. (1977). Statistical Power Analysis for Denzin, N., & Lincoln, Y. (1994). Handbook of
the Behavioral Sciences. New York: Acade- Qualitative Research. London: Sage.
mic Press. Department of Health, Education and Welfare
Colaizzi, P.F. (1978). Psychological research (1973). Final Report of the Tuskegie Syphilis
as the phenomenologist views it. In R.Valle Study Ad Hoc Advisory Panel. Washington,
& M. King (Eds.), Existential Phenomenolog- D.C.
ical Alternative for Psychology. New York: Deutscher, I. (1966). Words and deeds: Social
Oxford University Press. action and social policy, Social Problems,
Connelly, L., Bott, M., Goffart, N., & Taunton, 13, 235254.
R. (1997). Methodological triangulation in a Dickoff, J., James, P., & Weednbach, E. (1968).
study of nurse retention. Nursing Research, Theory in a practice discipline: Practice
46(5), 299302. oriented theory (Part 1). Nursing Research,
Cook, A.C. (1997). Investigator bias in bereave- 17, 415435.
ment research: Ethical and methodological Dickson, G. (2000). Aboriginal grandmothers
implications. Canadian Journal of Nursing Re- experience with health promotion and par-
search, 29(4), 8793. ticipatory action research. Qualitative
Corbin, J., & Strauss, A. (1990). Grounded the- Health Research, 10(2), 188213.
ory research: Procedures, canons, and Diekleman, N., Allen, D., & Tanner, C. (1989).
evaluative criteria, Qualitative Sociology, The National League for Nursing Criteria for
13(1), 321. Appraisal for Baccalaureate Programs: A
Cormack, F.S. (1996). The Research Process in Critical Hermeneutic Analysis. Publ. No. 15-
Nursing (3rd ed.). London: Blackwell Scien- 2253. New York: NLN Press.
tic Publications. Diers, D. (1979). Research in Nursing Practice.
Creswell, J. (1994). Research Design: Qualita- Philadelphia: J.B. Lippincott.
tive and Qualitative Approaches. Thousand Dillman, D.A. (1999). Mail and Electronic Sur-
Oaks, CA: Sage. veys: The Tailored Design Method. New York:
Creswell, J.W. (1998). Qualitative Inquiry and John Wiley & Sons.
Research Design: Choosing Among Five Tra- Dillman, D.A. (1978). Mail and Telephone Sur-
ditions. London: Sage. veys: The Total Design Method. New York:
Crookes, P., & Davies, S. (1998). Research into John Wiley & Sons.
Practice. Edinburgh: Billire Tindall. Dillon, J.T. (1990). The Practice of Questioning.
New York: Routledge.
Davidoff, H. (1953). A World Treasury of Pro- Downie, R.S., & Calman, K.C. (1987). Healthy
verbs. London: Cassell & Company. Respect: Ethics in Health Care. London: Faber
Davis, D.L., & Joakinsen, L. M. (1997). Nerves & Faler.
as status, and nerves as stigma. Qualitative Downs, F. (1984). Elements of a research cri-
Health Research, 7(3), 370391. tique. In Downs, F., & Newman, M. (Eds). A
Dawes, M., Davies, P., Gray, A., Mont, J., Seers, Sourcebook of Nursing Research (2nd ed.).
K., & Snowball, R. (1999). Evidence-based Philadelphia: F.A. Davis.
BIBLIOGRAPHY 721
Duffy, M. (1986). Primary prevention behav- evaluation. Community Development Jour-

iors: The female-headed, one-parent family. nal, 23, 1625.
Research in Nursing and Health, 9, 115122. Field, W.E. (1983). Clinical nursing research: A
Dulock, H. L., & Holzemer, W. L. (1991). Sub- proposal of standards. Nursing Leadership,
struction: Improving the linkage from the- December 6, 117120.
ory and method. Nursing Science Quarterly, Finfgeld, D. (1998). Courage in middle-aged
4(2), 8387. adults with long-term health concerns. Cana-
dian Journal of Nursing Research, 30(1), 153
Edwards, M. (1998). The Internet for Nurses and 169.
Allied Health Professionals. New York: Fisher, D. (1993). Fed up with too many ques-
Springer-Verlag. tions, Canadians are hanging up on poll-
Edwards, N., Sims-Jones, M., & Breithaupt, K. sters, Montreal Gazette, (September 24, A8).
(1998). Maternal smoking/non-smoking and Fitzpatrick, J., & Whall, A.L. (1996). Concep-
feeding choice. Canadian Journal of Nursing tual Models of Nursing: Analysis and Appli-
Research, 30(3), 8399. cation. (3rd ed.), Stamford, CT: Appleton &
Eichler, M. (1988). Nonsexist Research Meth- Lange.
ods. Boston: Allen & Unwin. Flaskerud, J., & Nyamathi, A. (2000). Attaining
Eichner, K., & Habermehl, W. (1981). Predict- gender and ethnic diversity in health inter-
ing response rates to mailed question- vention research: Cultural responsiveness
naires, American Sociological Review, 46, versus resource provision. Advances in Nurs-
361363. ing Science, 22(4), 115.
Emami, A., Torres, S., Lipson, J.G., & Ekman, S. Flynn (1992). The development of the healthy
(2000). An ethnographic study of a day cities project in the US. In J. Ashton (Ed.),
care center for Iranian immigrant seniors. Healthy Cities (pp. 4348). Buckingham:
Western Journal of Nursing Research, 22(2), Open University Press.
169188. Forcese, D.P., & Richer, S. (1973). Social Re-
Erdos, P. L. (1983). Professional Mail Surveys. search Methods. Englewood Cliffs, N.J.: Pren-
Malabar, Florida: Robert E. Krieger Pub- tice-Hall.
lishing. Forchuk, C. (1994). The orientation phase of
Estabrooks, C.A. (1998). Will evidence-based the nurse-client relationship: Testing Pe-
nursing practice make practice perfect? plaus theory. Journal of Advanced Nursing,
Canadian Journal of Nursing Research, 30(4), 20, 532537.
273295. Forchuk, C., & Roberts, J. (1993). How to cri-
tique qualitative research articles. Canadian
Fain, J.A. (1999). Reading Understanding and Ap- Journal of Nursing Research, 25(4), 4757.
plying Nursing Research: A Text and Work- Ford-Gilboe, M., & Campbell, J. (1996). The
book. Philadelphia: F.A. Davis. mother headed single parent family: A fem-
Fawcett, J., & Downs, F. (1986). The Relation- inist critique of the nursing literature. Nurs-
ship of Theory and Research. Norwalk, CT: ing Outlook, 44, 173183.
Appleton-Century Crofts. Fougere, A. (1992). Effects of eating breakfast
Fawcett, J. (1999). The Relationship of Theory on grade performance, Antigonish: St. Fran-
and Research. Philadelphia: F.A. Davis. cis Xavier University, Research Methods
Fay, B. (1987). Critical Social Science: Libera- Paper, Sociology 300.
tion and Its Limits. Ithaca, N.Y.: Cornell Uni- Fowler, F.J. (1993). Survey Research Methods
versity Press. (2nd ed). Newbury Park, CA: Sage.
Featherman, D.L., & Stevens, G. (1982). A re- Frank-Stromberg, M., & Olsen, S. (1997. Mea-
vised index of occupational status: Appli- surement in Nursing Research (2nd ed.).
cation in analysis of sex differences in at- Boston, MA: Jones & Bartlett.
tainment. In Social Structure and Behavior:
Essays in Honor of William Hamilton Sewell. Gale, A., Gardner, L., Greer, C., Jackson, N.,
New York: Academic Press. Jasper, M., Rolfe, G., & Sherwood-Rogers, S.
Ferketich, S., & Muller, M. (1990). Factor analy- (1998). An Action Research Project to De-
sis revisited. Nursing Research, 39(1), 59 velop the Role of the Generic Health Care
62. Support Worker. 4th Qualitative Health Re-
Festinger, L., Reiken, H.W., & Schachter, S. search Conference, Vancouver, B.C., Febru-
(1956). When Prophecy Fails. New York: ary 1921, 1998.
Harper & Row. Gates, K.M. (2000). The experience of caring
Feurstein, M. (1988). Finding the methods to for a loved one: A phenomenological study.
t the people: Training for participatory Nursing Science Quarterly, 13(1), 5459.
722 BIBLIOGRAPHY
Gates, M.F., & Lackey, N.R. (1998). Youngsters spectives & Practice. Milton Keynes, Eng-
caring for adults with cancer. Image: Jour- land: Crown Copyright.
nal of Nursing Scholarship, 30(1), 1116. Goyder, J., & McKenzie-Leiper, J. (1985c). The
George, D., & Mallery, P. (2000). SPSS for Win- decline in survey response: A social values
dows Step by Step: A Simple Guide and Ref- interpretation, Sociology, 19(1), 5571.
erence 9.0 Update (2nd ed.). Boston: Allyn Goyder, J.C. (1982). Further evidence on fac-
& Bacon. tors affecting response rates to mailed ques-
George, J.B. (1995). Nursing Theories: The Base tionnaires, American Sociological Review,
for Professional Nursing Practice (4th ed.). 47, 550553.
Norwalk, CT: Appleton & Lange. Goyder, J.C. (1985a). Nonresponse on surveys:
Georgi, A. (1985). Phenomenology and Psycho- A Canada-United States comparison, Cana-
logical Research. Pittsburgh: Duquesne Uni- dian Journal of Sociology, 10, 231251.
versity Press. Goyder, J.C. (1985b). Face-to-face interviews
Gillis, A.J. (1990). Nurses knowledge of growth and mailed questionnaires: The net differ-
and development principles in meeting the ence in response rate, Public Opinion Quar-
psychosocial needs of hospitalized children. terly, 49, 23425.
Journal of Pediatric Nursing, 5(2), 7887. Gray, M., & Smith, L. (1999). The professional
Gillis, A.J., MacLellan, M., & Perry, A. (1998). socialization of diploma of higher educa-
Competencies of liberal education in post tion nursing students (Project 2000): A lon-
RN baccalaureate students. Journal of Nurs- gitudinal qualitative study. Journal of Ad-
ing Education, 37(9), 408412. vanced Nursing, 29(3), 639647.
Gillis, A.J. (1998). The trends and variations in Greenberg, M. (1998). Therapeutic playing: A
smoking. Course paper for Nursing 300, St. grounded theory of the utilization of humor
Francis Xavier University. within the nurse-client relationship in an
Gillis, A.J. (1993). The relationship of denition acute care setting. Qualitative Health Re-
of health, perceived health status, self-efcacy, search Conference. Hotel Vancouver, Van-
parental health promoting lifestyle, and se- couver, BC. February, 1998.
lected demographics to health-promoting Grunburg, Martha. (1998). Therapeutic play-
lifestyle in adduces out females. Doctoral Dis- ing: A grounded theory of the Welzation of
sertation, University of Texas at Austin. Humor within the nurse-client relationship in
Gillis, A.J. (1994). Determinants of health-pro- an acute care setting. 4th QHRC-UBC School
moting lifestyle in adolescent girls. Canadian of Nursing, Vancouver, BC. Feb 1921.
Journal of Nursing Research, 26(2), 1329.
Gillis, A.J. (1997). The adolescent lifestyle ques- Hahn, E.J., Bryant, R., Peden, A., Robinson, K.,
tionnaire: Development and psychometric & Williams, C.A. (1998). Entry into commu-
testing. Canadian Journal of Nursing Re- nity based nursing practice: Perceptions of
search, 29,(1), 2947. prospective employers. Journal of Profes-
Gillon, R. (1986). Philosophical Medical Ethics. sional Nursing, 14(5), 305314.
Chichester: Wiley. Hall, J.M., & Stevens, P.E. (1991). Rigor in fem-
Glaser, B., & Strauss, A. (1967). The Discovery inist research. Advances in Nursing Science,
of Grounded Theory. Chicago: Aldine. 13(3), 1629.
Glaser, B. (1978). Theoretical Sensitivity. Niel Halldrsdtter, S., & Hamrin, E. (1997). Caring
Valley, CA: Sociology Press. and uncaring encounters within nursing
Goffman, E. (1962). Asylums. Chicago: Aldine. and health care from the cancer patients
Gold, D. (1958). Comment on A Critique of perspective. Cancer Nursing, 20(2), 120
Tests of Signicance, American Sociologi- 128.
cal Review, 23 (February), 8586. Hamblin, R.L. (1971). Ratio measurement for
Goode, C., Butcher, L., Cipperley, J., Ekstrom, the social sciences, Social Forces, 50, 191
J., Hayes, J., Lovett, M., & Wellendorf, S. 206.
(1996). Research utilization: A study guide. Hammersley, M., & Atkinson, P. (1995). Ethnog-
Ida Grove, Iowa: Horne Video Productions. raphy: Principles in Practice (2nd ed). New
Gooding, B.A., Sloan, M., & Gagnon, L. (1993). York: Routledge.
Important nurse caring behaviors: Percep- Hanna, B. (1997). From raw data to chapters
tions of oncology patients and nurses. Cana- via the support of NUD*IST. Health Info-
dian Journal of Nursing Research, 25(3), matics Journal, 3, 6165.
6577. Hanson, M.J. (1997). The theory of planned be-
Gott, M., & OBrien, M. (1990). The Role of the havior applied to cigarette smoking in
Nurse in Health Promotion: Policies, Per- African-American, Puerto Rican, and Non-
BIBLIOGRAPHY 723
Hispanic, white teenage females. Nursing Homans, G.C. (1964). Bringing men back in.
Research, 46(3), 155162. American Sociological Review, 29, 809818.
Harding, S. (1987). Is there a feminist method? Horne Video Productions (1996). Research uti-
In Harding S. (Ed.). Feminism and Method- lization: A study guide (2nd ed.). Ida Grove,
ology. Bloomington, IN: Indiana University Iowa: Author.
Press. Hoskins, C.N. (1988). The Partner Relationships
Harding, S. (1991). Whose Knowledge? Whose Inventory. Palo Alto, CA: Consulting Psy-
Science? Thinking from Womens Lives. Ith- chologists Press.
ica, NY: Cornell University Press. Howell, D.C. (1997). Statistical Methods for Psy-
Hart, E., & Bond, M. (1998). Action Research for chology (4th ed.). Belmont, CA: Duxbury
Health and Social Care. Philadelphia: Open Press, Wadsworth Publishing.
University Press. Huber, P.W. (1991). Galileos Revenge: Junk-
Harton, H.C., & Latane, B. (1997). Social inu- yard Science in the Courtroom. New York:
ence and adolescent lifestyle attitudes. Jour- Basic Books.
nal of Research on Adolescence, 7(2), 197 Hughes, D. (1998). The invisible burden: Wo-
220. mens lives with bromyalgia. Qualitative
Hartrick, G.A. (1997). Women who are moth- Health research Conference, Hotel Vancou-
ers: The experience of dening self. Health ver, Vancouver, BC, February, 1998.
Care for Women International, 18, 263277. Humenick, S., Hill, P., Thompson, J., & Hart,
Havelock, R., & Zlotolow, S. (1995). The Change A.M. (1998). Breast-milk sodium as a predic-
Agents Guide to Innovation in Education (2nd tor of breastfeeding patterns. Canadian
ed.) Englewood Cliffs, NJ: Educational Tech- Journal of Nursing Research, 30(3), 6783.
nology Publications. Humphreys, L. (1970). Tearoom Trade. Chicago:
Havelock, R. (1972). Bibliography on knowl- Aldine Press.
edge Utilization and Dissemination. Ann Ar- Humphris, D. (1999). Types of evidence. In S.
lor, MI: Institute for Social Research. Hamer & G. Collinson (Eds.), Achieving Ev-
Healthcare Quest. (1997). Getting Audit Ready idence-based Practice: A Handbook for Prac-
to Benet Patients. Romsey. titioners. Edinburgh: Harcourt.
Heberlein, T.A., & Baumgartner, R. (1978). Fac- Hyman, H. (1955). Survey Design and Analysis.
tors Affecting Response Rates to Mailed New York: Free Press.
Questionnaires. American Sociological Re-
view, 43. Ibbotson,T. (1999). An Ethnographic Study of
Hinds, P., Vogel, R., & Clarke-Sleffen, L. (1997). the Diffusion Process of Telemedicine in
The possibilities and pitfalls of doing a sec- Scotland. Middlesex: Brunel University.
ondary analysis of a qualitative data set. Im, E. (2000). A feminist critique of research
Qualitative Health Research, 7(3), 408425. on womens work and health. Health Care
Hilton, B., Thompson, R., & Moore-Dempsey, for Women International, 21, 105119.
L. (2000). Evaluation of the AIDS Preven- Im, E., & Meleis, A. (1999). Situation-specic
tion Street Nurse Program: One step at a theories: Philosophical routes, properties,
time. Canadian Journal of Nursing Research, and approach. Advances in Nursing Science,
32 (1), 1738. 22(2) 1124.
Hochschild, A.R. (1983). The Managed Heart: Iowa Model for Research-Based Practice to
Commercialization of Human Feeling. Berke- Promote Quality Care (1994). The Univer-
ley: University of California Press. sity of Iowa.
Hodges, J.C., Whitten, M.E., Brown, J., & Flick, Israel, G.A. & Shurman, S.U. (1990). Social sup-
J. (1994). Harbrace College Handbook: for port, control, and the stress process. In K.
Canadian Writers, (4th ed.). Toronto: Har- Glanz, F. Lewis, & B. Rimer (Eds.) Health Be-
court Brace, Canada. havior and Health Education. San Franciso,
Holden, C. (1979). Ethics in social science re- CA: Jossey-Bass.
search, Science, 206 (November), 537540. Iversen, G.R., & Norpoth, H. (1976). Analysis of
Hollingshead, A. B. (1975). Four Factor Index of Variance. Beverly Hills: Sage.
Social Status. Unpublished manuscript.
Yale Station, New Haven, CT. 06520. Jackson, W. (1973). University Preferences and
Holroyd, E., Katie, F., Chun, L., & Wai Ha, S. Perceptions of Pictou County Students. Antigo-
(1997). Doing the month: An exploration of nish: St Francis Xavier University.
postpartum practices in Chinese women. Jackson, W., & Poushinsky, N.W. (1971). Mi-
Health Care for Women International, 18, gration to Northern Mining Communities:
301313. Structural and Social Psychological Dimen-
724 BIBLIOGRAPHY
sions. Winnipeg: Center for Settlement and gratication. Canadian Journal of Nurs-
Studies, University of Manitoba. ing Research, 30(3), 5367.
Jackson, W. (1999). Methods: Doing Social Re- Kimchi, J., & Simmons, H. (1994). Developing
search (2nd ed). Scarborough, Ont.: Pren- a Philosophy of Nursing. Lincoln: Sage.
tice Hall (Canada). Kinnear, P., & Gray, C. (1999). SPSS for Win-
Janke, J. (1999). The effects of relaxation ther- dows Made Simple (3rd ed.). East Sussex,
apy on preterm labor outcomes. Journal of UK: Psychology Press.
Obstetric, Gynecologic, and Neonatal Nurs- Kinzel, C. (1992). The Alberta Survey 1992:
ing, 28, 255263. Sampling Report. The University of Alberta:
Janson-Bjerklie, S., Carrieri, V., & Hudes, D. Population Research Laboratory.
(1986). The sensation of pulmonary dsy- Kish, L. (1965). Survey Sampling. New York:
pnea. Nursing Research, 35(3), 154159. Wiley.
Jeans, M.E. (1992). Editorial: Clinical signi- Klein, R.D. (1983). How to do what we want to
cance of research: A growing concern. do: Thoughts about feminist methodology.
Canadian Journal of Nursing Research, 24(1), In Bowles, G., & Klein, R.D. (Eds.) Theories of
12. Womens Studies (pp. 88104). London: RKP.
Jeans, M.E. (1990). Editorial: Research dollars: Klostermann, B., Perry, C., & Britto, M. (2000).
Whos passing the buck? Canadian Journal Quality improvement in a school health
of Nursing Research, 22(3), 12. program. Evaluation and the Health Profes-
sions, 23(1), 91106.
Kachigan, S.K. (1986). Statistical Analysis: An Kna, K., & Howard, M. (1984). Interpreting
Interdisciplinary Introduction to Univariate and reporting qualitative research. Re-
& Multivariate Methods. New York: Radius search in Nursing and Health, 7, 1724.
Press. Konrad, T. R., & DeFriese, G. (1990). On the
Kalisch, B.J., Kalisch, P.A., & Belcher, B. subject of sampling. American Journal of
(1985). Forecasting for nursing policy: A Health Promotion, 5(2), 147153.
news-based image approach. Nursing Re- Krahn, H. (1991). Sociological methods of re-
search, 34(1), 4449. search. In L. Tepperman and R. Jack Richard-
Kalischuk, R. (1998). Family healing following son (Eds.), The Social World: An Introduction
adolescent suicide. Qualitative Health Re- to Sociology. Toronto: McGraw-Hill Ryerson,
search Conference, Hotel Vancouver, Van- pp. 3466.
couver, BC, February, 1998. Kulig, J., & Thorpe, K. (1996). Teaching and
Kang, D., Coe, C., Karaszewski, J., & McCarthy, learning needs of culturally diverse post
D. (1998). Relationship of social support to RN students. Canadian Journal of Nursing
stress responses and immune function in Research, 28(2), 119125.
healthy and asthmatic adolescents. Re-
search in Nursing and Health, 21, 117128. LaPiere, R.T. (1934). Attitudes vs. actions, So-
Kaviani, N., & Stillwell, Y. (2000). An evaluative cial Forces, 13, 230237.
study of clinical preceptorship. Nurse Edu- Larson, P. (1981). Oncology patients and pro-
cation Today, 20, 218226. fessional nurses perception of important
Kerlinger, F.N. (1973). Foundations of Behav- caring behaviors. (Doctoral dissertation,
ioral Research. New York: Holt, Rinehart & University of California, San Francisco,
Winston. 1981). Dissertation Abstracts International,
Kerlinger, F.N. (1986). Foundations of behav- 42, 568B.
ioral research (3rd ed.). New York: Holt, Larson, P. (1984). Important nurse caring be-
Rinehart, and Wilson. haviors perceived by patients with cancer.
Kerr, J., & MacPhail, J. (1996). Canadian Nurs- Oncology Nurses Forum, 11, 4650.
ing: Issues and Perspectives (3rd ed.). St. Lastrucci, C.L. (1967). The Scientic Approach.
Louis: Mosby. Cambridge, MA: Schenkman Publishing.
Kidd, P., & Parshall, M. (2000). Getting the fo- Lather, P. (1993). Fertile obsession: Validity af-
cus and the group: Enhancing analytical ter post structuralism. Sociological Quar-
rigor in focus group research. Qualitative terly, 34, 673693.
Health Research, 10(3), 293308. LeCompte, M.D., & Goetz, J.P. (1982). Prob-
Kikchi, J., Polivka, B., & Stevenson, J. (1991). lems of reliability and validity in ethno-
Triangulation: Operational denitions. Nurs- graphic research. Review of Educational
ing Research, 40, 364366. Research, 52(1), 3160.
Killien, M.G. (1998). Postpartum mothers re- Lederman, R., & Miller, D. (1998). Adaptation
turn to work: Mothering stress, anxiety, to pregnancy in three different ethnic
BIBLIOGRAPHY 725
groups: Latin American, African-American, Lincoln, Y.S. (1995). Emerging criteria for qual-
& Anglo-American. Canadian Journal of ity in qualitative and interpretive research.
Nursing Research, 30(3), 3753. Qualitative Inquiry, 1, 275289.
Lee, K.H. (1997). Korean urban womens expe- Lindeman, C.A. (1975). Priorities in clinical
rience of menopause: New life. Health Care nursing research. Nursing Outlook, 23, 693
for Women International, 18, 139148. 698.
Lee, M. (1992). Smoking Behaviours., Antigo- Lindsey, E., & Stafduhar, K. (1998). From rhe-
nish: St. Francis Xavier University (Re- toric to action: Establishing community par-
search methods paper). ticipation in AIDS-related research. Canadian
Leenerts, M., & Magilvy, J. (2000). Investing is Journal of Nursing Research, 30(1), 137153.
self-care: A midrange theory of self-care LoBiondo-Wood, G., & Haber, J. (1998). Nurs-
grounded in the lived experience of lowing Research: Methods Critical Appraisal
income HIV-positive white women. Ad- and Utilization (4th ed). St. Louis: Mosby.
vances in Nursing Science, 22(3), 5875. Lodge, M. (1981). Magnitude Scaling: Quantita-
Leininger, M. (1970). Nursing and Anthropol- tive Measurement of Opinions. Beverly Hills:
ogy: Two Worlds to Blend. New York: John Sage.
Wiley & Sons. Long, E.R., Suppe, F., Gift, A., Pugh, L.C., & Mil-
Leininger, M. (1978). Transcultural Nursing: Con- ligan, R.A. (1995). Collaborative develop-
cepts Theories and Practices. New York: John ment of middle-range nursing theories: To-
Wiley & Sons. ward a theory of unpleasant symptoms.
Leininger, M. (1985). Qualitative Research Advances in Nursing Science, 17, 113.
Methods in Nursing. Orlando, FL: Freene &
Atratton. MacDonald, L. (1991). Attitudes toward the el-
Leininger, M. (1991). Culture Care Diversity and derly (Research Methods Paper). Antigo-
Universality: A Theory of Nursing. New York, nish: St. Francis Xavier University.
NY: National League for Nursing Press. MacKinnon, M. (1997). What a difference a
Leininger, M. (1990). Ethnomethods: The philo- nurse makesThen and now. Western Jour-
sophic and epistemic bases to explicate nal of Nursing Research, 19(6), 795801.
transcultural nursing knowledge. Journal of MacLeod, C.J., & Hockey, L. (1989). Further Re-
Transculteral Nursing, 1, 4051. search for Nursing. London: Scutari-Press
Lenz, E.R., Pugh, L.C., Milligan, R.A., Gift, A., & Malinozwski, B. (1925). Magic, Science and Re-
Suppe, F. (1995). Collaborative develop- ligion. New York: Doubleday.
ment of middle-range nursing theories: To- Manson-Singer, S. (1994). The Canadian health
ward a theory of unpleasant symptoms. Ad- communities project: Creating a social
vances in Nursing Science, 17, 113. movement. In A. Pederson, M. ONeil & I.
Lenz, E.R., Pugh, L.C., Milligan, R.A., Gift, A., & Rootman (Eds.), Health Promotion in Canada,
Suppe, F. (1997). The middle range theory Toronto: WB Saunders.
of unpleasant symptoms: An update. Ad- Marshall, C., & Rossman, G.B. (1995). Design-
vances in Nursing Science, 19, 1427. ing Qualitative Research (2nd ed.). Thou-
Levin, J., & Fox, J.A. (1991). Elementary Statis- sand Oaks, CA: Sage.
tics in Social Research, (5th ed.). New York: Martof, E. (2000). Personal communication with
HarperCollins. authors.
Lewin, K. (1951). Field Theory in Social Sci- Martin, D. W. (1991). Doing Psychology Experi-
ences. New York: Harper & Row. ments (3rd ed.). Pacic Grove: Brooks/Cole
Likert, R. (1931). A technique for the mea- Publishing Company.
surement of attitudes. Archives of Psychol- Massey, V.H. (1995). Nursing Research (2nd
ogy. New York: Columbia University Press. ed.). Springhouse, PA: Springhouse.
Lincoln, Y., & Guba, E.G. (1994). Competing Mason, C., Orr, J., Harrisson, S., & Moore, R.
paradigms in Qualitative Research. In N.K. (1999). Health professionals perspective on
Denzin & Y.S. Lincoln (Eds.), Handbook of service delivery in two Northern Ireland
Qualitative Research (pp. 105117). Thou- communities. Journal of Advanced Nursing,
sand Oaks, CA: Sage. 30(4), 827834.
Lincoln, Y., & Guba, E.G. (1985). Naturalistic In- Masterson, A. (1996). Clarifying Theory for
quiry. Beverly Hills, CA: Sage. Practice: Study Guide. Distance Learning.
Lincoln, Y.S., & Guba, E.G. (1981). Effective Eval- London: Royal College of Nursing.
uation: Improving the Effectiveness of Evalua- Mawdsley, C. (1997). Study says critical care
tion Results Through Responsive and Natural- nurses quick to aid in death of terminally
istic Approaches. San Francisco: Josey-Bass. ill: A research critique. CACCN, 8(2), 811.
726 BIBLIOGRAPHY
Maynard-Tucker, G. (2000). Conducting focus Montgomery, K. (2000). Getting organized:

groups in developing countries: Skill train- Qualitative data collection. Applied Nursing
ing for local bilingual facilitators. Qualita- Research, 13(2), 103104.
tive Health Research, 10(3), 396410. Moody, L. E., Wilson, M. E., Smyth, K., Tittle,
McDonald & Daly cited in Parahoo. (1997). M., & Vancott, M. L. (1988). Analysis of a
McEwen, M. (1993). The health motivation as- decade of nursing practice research: 1977
sessment and inventory. Western Journal 1986. Nursing Research, 37, 374379.
Nursing Research, 15(6), 770776. Moore, P.A. (1995). The utilization of research
McKay, L., & Diem, L. (1995). Health concerns in practice. Professional Nurse, 10, 536537.
of adolescent girls. Journal of Pediatric Nurs- Morris, R. (1996). The culture of female cir-
ing, 10(1), 1927. cumcision. Advances in Nursing Science,
McQueen, D. (1994). Health Promotion Research 19(2), 4353.
in Canada: A European/British Perspective. In Morris, J., Penrod, J., & Hupcey, J. (2000).
A. Pederson, M. ONeill, & I. Rootman (Eds.). Qualitative outcome analysis: Evaluating
Health Promotion in Canada. Toronto: W.B. nursing interventions for complex clinical
Saunders. phenomena. Journal of Nursing Scholarship,
Mead, G.H. (1934). Mind, Self, and Society. Chi- 32 (2), 125130.
cago: University of Chicago Press. Morrison, E.F. (1998). Enormous beliefs about
Mead, M. (1935). Sex and Temperament in Three research held by staff nurses. The Journal
Primitive Societies. New York: Morrow. of Continuing Education in Nursing, 29(5),
Meehan, T.C. (1998). Therapeutic touch as a 196203.
nursing intervention. Journal of Advanced Morse, J. (1991). Qualitative Nursing Research:
Nursing, 28(1), 117125. A Contemporary Dialogue. Newbury Park,
Meerabeau, L. (1992). Tacit nursing knowledge: CA: Sage.
An untapped resource or a methodological Morse, J. (1992). The power of induction. Qual-
headache? Journal of Advanced Nursing, 17, itative Health Research, 2(1), 36.
108112. Morse, J. (1998). Qualitative Methods: The State
Meleis, A. (1991). Theoretical Nursing: Devel- of the Art. Keynote address presented at the
opments and Progress (2nd ed.). Philadel- 4th Qualitative Health Research Confer-
phia: Lippincott. ence, Vancouver, British Columbia, Febru-
Meleis, A. (1997). Theoretical Nursing: Develop- ary 19 21, 1998.
ment and Progress (3rd ed.). Philadelphia: Morse, J. M., & Field, P.A. (1995). Qualitative
Lippincott. Research Methods for Health Professionals.
Meyer, J.E. (1993). New paradigm research in Thousand Oaks: Sage.
practice: The trials and tribulations of ac- Morris, J., Penrod, J., & Hupcey, J. (2000).
tion research. Journal of Advanced Nursing, Qualitative outcome analysis: Evaluating
18, 10661072. nursing interventions for complex clinical
Miles, M., & Huberman, M. (1994). Qualitative phenomena. Journal of Nursing Scholarship,
Data Analysis. Thousand Oaks, CA: Sage. 32 (2), 125130.
Mill, J.S. (1925). A System of Logic (8th ed.). Morton, J. (1994). The clinical usefulness of
London: Lognmans, Green and Co. breast milk sodium in the assessment of
Miller, D. C. (1977). Handbook of Research De- lactogenesis. Pediatrics, 93, 802886.
sign and Social Measurement (3rd ed). New Moustaka, C. (1994). Phenomenological Re-
York: Longman. search Methods. Thousand Oaks, CA: Sage.
Mishel, M., & Murdaugh, C. (1987). Family ad- Mueller, C. (2000). Personal communication
justment to heart transplantation: Redesign- to authors.
ing the dream. Nursing Research, 36(6), 332 Munhall, P.L., & Boyd, C.O. (1993). Nursing Re-
338. search. A Qualitative Perspective. New York:
Mitchell, G. (1994). Intuitive knowing: Expos- National League for Nursing Press.
ing a myth in theory development. Nursing Munhall, P.L. (1988). Ethical considerations in
Science Quarterly, 7(1), 23. qualitative research. Western Journal of Nurs-
Mitchell, M., & Jolley, J. (2000). Research Design ing Research, 10(2), 150162.
Explained (4th ed.). Orlando: Harcourt Brace Munhall, P.L., & Oiler, C.J. (1986). Nursing Re-
Jovanovitch College Publishers. search: A Qualitative Perspective. Norwalk,
Monette, D.R., Sullivan, T.J., & DeJong, C.R. CT: Appleton-Century-Crofts.
(1990). Applied Social Research: Tool for the Munro, B.H. (2001). Statistical Methods for
Human Services (2nd ed.). Fort Worth: Holt, Health Care Research (4th ed.). Philadelphia:
Rinehart and Winston. Lippincott.
BIBLIOGRAPHY 727
Munro, B., & Page, E. (1993). Statistical Meth- in the Social Sciences (pp. 4462). San Fran-
ods for Health Care Research. Philadelphia: cisco: West View Press.
J.B. Lippincott. Oates, K. (1997). Models of planned change
Murray (1999). Unwanted intimacy in female and research to product evaluation. Clini-
university students (Research Methods Pa- cal Nurse Specialist, 1(6), 270273.
per). Antigonish: St. Francis Xavier Univer- OConnell, K. (2000). If you call me names, Ill
sity. call you numbers. Journal of Professional
Musial, C.M., Jones, S.L., & Warner, C.D. (1998). Nursing, 16(2), 74.
Structural equation modeling and its rela- OConnell, A. (2000). Sampling for Evaluation.
tionship to multiple regression and factor Evaluation & the Health Professions, 23(2),
analysis. Research in Nursing & Health, 21, 212234.
271281. Ogden-Burke, S., Kauffman, E., Costello, E.,
Wiskin, N., & Harrison, M. (1998). Stressors
National Health Services (NHS) (1998). Chang- in families with a child with a chronic condi-
ing Clinical Practice. Leeds, England: De- tion. Canadian Journal of Nursing Research,
partment of Health. 30(1), 7197.
National Health Services (NHS) (1998). Achiev- Olesen, V. (1994). Feminisms and models of
ing effective practice: A clinical effectiveness qualitative research. In NK. Denzin & Y.S.
and research information pack for nurses, Lincoln (Eds.). Handbook of Qualitative Re-
midwives and health visitors. Leeds, Eng- search (pp. 158174). Thousand Oaks, CA:
land: Department of Health. Sage.
Nelms, T. (2000). The practices of mothering Oliver, S., & Redfern, S. (1991). Interpersonal
in caregiving an adult son with AIDS, Ad- communication between nurses and elderly
vances in Nursing Science, 22(3), 4657. patients: Renement of an observation sche-
Neter, J., Wasserman, W., & Kutner, M. (1985). dule. Journal of Advanced Nursing, 16, 3038.
Applied Linear Statistical Models. Home- Onyskiw, J. (1996). The meta-analytic approach
wood, IL: Irwin. to research integration. Canadian Journal of
Neuberger, J. (1992). Ethics and health care. Nursing Research, 28(3), 6985.
The role of researcher ethics committees in Omery, A. (1983). Phenomenology: A method
the U.K. London: Kings Fund Institute. for nursing research. Advances in Nursing
Nilson, K., Nordstrom, G., Krusebrant, A., & Science, 5(2), 4963.
Bjorvell, H. (2000). Perceptions of research Orem, D. (1995). Nursing: Concepts of Practice
utilization: Comparisons between health (5th ed.). St Louis: Mosby.
care professionals, nursing students, and a Orne, M. T. (1962). On the social psychology
reference group of nurse clinicians. Journal of the psychological experiment: With par-
of Advanced Nursing, 31(1), 99109. ticular reference to demand characteris-
Norbuck, J.S., Lindsay, A.M., & Carrieri, V.L. tics and their implications. American Psy-
(1982). Further development of the nor- chologist, 17, 7683.
brick social support questionnaire: Norma- Orne, R., Fishman, S., Manka, M., & Pagnozzi, M.
tive data and validity testing. Nursing Re- (2000). Living on the edge: A phenomeno-
search, 32(1), 49. logical study of medically uninsured work-
Norwood, S.L. (2000). Research Strategies for ing Americans. Research in Nursing and
Advanced Practice Nurses. Upper Saddle Health, 23(23), 204212.
River, NJ: Prentice Hall Health. Osgood, C.E., Suci, G.J., & Tannenbaum, P.H.
Nunnally, J.C. (1978). Psychometric Theory (1957). The Measurement of Meaning. Ur-
(2nd ed). New York: McGraw-Hill. bana: University of Illinois Press.
Nuremburg Code (1949). In R.J. Levine (Ed.),
(1986). Ethics and Regulation of Clinical Re- Pallikkathayil, L., Crighton, F., & Aaronson, L.
search, (2nd ed., pp 425426). Baltimore, (1998). Balancing ethical quandaries with
Munich: Urban & Schwarzenberg. scientic rigor: Part I. Western Journal of
Nursing Research, 20(3), 388393.
Oakley, A. (1989). Whos afraid of the random- Palo-Bengtson, L., & Ekman, S. (1997). Social
ized control trial? Some dilemmas of the dancing in the care of persons with de-
scientic method and good research prac- mentia in a nursing home setting: A phe-
tice. Women and Health, 15(4), 2559. nomenological study. Scholarly Inquiry for
Oakley, A. (1981). Interviewing women: A con- Nursing Practice: An International Journal,
tradiction in terms. In H. Roberts (Ed.), Fem- 11(2), 101118.
inist Research Methods: Exemplary Readings Paltiel, F., Ross, E., & Neill, M. (1998). Coming of
728 BIBLIOGRAPHY
age in the metropolis: The Toronto experi- prestige in Canada. Canadian Review of So-
ence. The Canadian Nurse, 94(10), 2230. ciology and Anthropology, 2440.
Palys, T. (1992). Research Decisions: Quantita- Polit, D. (1996). Data Analysis and Statistics for
tive and Qualitative Perspectives. Toronto: Nursing Research. Stanford, CT: Appleton-
Harcourt Brace Jovanovitch Canada. Lange.
Parahoo, K. (1997). Nursing Research: Princi- Polit, D., & Hungler, B. (1999). Nursing Re-
ples, Process and Issues. London: MacMillan search: Principles and Methods (6th ed.)
Press. Philadelphia, PA: J. B. Lippincott.
Parahoo, K. (2000). Barriers to and facilitators Polit, D.F., & Sherman, R.E. (1990). Statistical
of research utilization among nurses in power in nursing research. Nursing Re-
Northern Ireland. Journal of Advanced Nurs- search, 39(6), 365369.
ing, 31(1), 8998. Porter, C., Oakley, D., Ronis, D., & Neal, W.
Parse, R. (1987). Nursing Science: Major Para- (1996). Pathways of inuence on fth and
digms, Theories and Critiques. Philadelphia, eighth graders report about having had
PA: Saunders. sexual intercourse. Research in Nursing and
Parse, R., Coyne, A.B., & Smith, M.J. (1985). Nurs- Health, 19, 193204.
ing Research: Qualitative Methods. Bowie, Pranulis, M. (1997). Nurses roles in protecting
MD: Brady. human subjects. Western Journal of Nursing
Parse, R.R. (1992). Human becoming: Parses Research, 19(1), 130136.
theory of nursing. Nursing Science Quar- Pringle, D. (1999). Another twist on the double
terly, 5, 3542. helix: Research and practice. Canadian
Parse, R.R. (1993). The experience of laughter: Journal of Nursing Research, 30(4), 165181.
A phenomenological study. Nursing Science Public Citizen (1997). Dangerously awed
Quarterly, 6(1), 3943. AIDS research criticized. http://japan.cnn.
Parse, R.R. (1994). Quality of life: Sciencing com/HEALTH/9704122/AIDS.
and living the art of human becoming. Nurs- Punch, M. (1986). The Politics and Ethics of Field-
ing Science Quarterly, 7, 1620. work. Sage: London.
Parsons, K. (1997). The male experience of care- Punches, N. (1996). QSR NUD*IST User Guide
giving for a family member with Alzheimers (2nd ed.). Thousand Oaks, CA: Sage.
disease. Qualitative Health Research, 7(3),
391408. QSR NUD*IST User Guide (1997). Thousand
Patterson, G., & Zderad, L.T. (1976). Humanis- Oaks, CA: Sage.
tic Nursing. New York: John Wiley & Sons. Querker, A. (1997). Longitudinal research using
Pederson, A., ONeill, M., & Rootman, I. (1994). computerized clinical databases: Caveats
Health Promotion in Canada: Provincial, Na- and constraints. Nursing Research, 46(6),
tional and International Perspectives. Phila- 353355.
delphia, PA: W.B. Saunders.
Pence, G.E. (1990). Classic Cases in Medical Ratner, P., Johnson, J., & Jeffery, B. (1998). Ex-
Ethics. London: McGraw-Hill. amining emotional, physical, social, and spir-
Pender, N.J. (1987). Health Promotion in Nurs- itual health as determinants of self-rated
ing Practice (2nd ed.). Norwalk, CT: Apple- health status. American Journal of Health Pro-
ton & Lange. motion, 12(4), 275282.
Pender, N.J. (1996). Health Promotion in Nurs- Reifsnider, E. (1998). Follow-up study of chil-
ing Practice (3rd ed.). Stanford, CT: Apple- dren with growth deciency. Western Jour-
ton & Lange. nal of Nursing Research, 20(1), 1429.
Peplau, H. (1988). Interpersonal Relations in Reinharz, S. (1992). Feminist Methods in Social
Nursing (reissued). London: MacMillan Ed- Research. New York, NY: Oxford University
ucation. Press.
Peplau, H.E. (1987). Psychiatric skills, tomor- Reynolds, P.D. (1982). Ethics and Social Sci-
rows world. Nursing Times, 83, 2933. ence Research. Englewood Cliffs: Prentice-
Peplau, H.E. (1992). Interpersonal relations: A Hall.
theoretical framework for application in Ribbens, J. (1989). InterviewingAn unnatural
nursing practice. Nursing Science Quarterly, situation? Womens Studies International Fo-
5, 1318. rum, 12(6), 579592.
Piliavin, J.A., & Piliavin, I.M. (1972). Effect of Richards, T., & Richards, L. (1994). Using com-
blood on reactions to a victim, Journal of Per- puters in qualitative analysis. In N. Denzin &
sonality and Social Psychology, 23, 353361. Y. Lincoln (Eds.), Handbook of Qualitative
Pineo, P.C., & Porter, J. (1967). Occupational Research. Thousand Oaks, CA: Sage.
BIBLIOGRAPHY 729
Rittman, M., Page, P., Rivera, J., Stuphin, L., & Sandelowski, M. (1995). On the aestehtics of
Godown., I. (1997). Phenomenological study qualitative research. Image: Journal of Nrus-
of nurses caring for dying patients. Cancer ing Scholarship, 27(3), 205209.
Nursing, 20(2), 115119. Sandelowski, M. (2000). Combing qualitative
Ritzer, G. (1988). Sociological Theory (2nd and quantitative sampling, data collection,
ed.). New York: Alfred A. Knopf. and analysis techniques in mixed method
Roberts, C.A., & Burke, S.O. (1989). Nursing re- studies. Research in Nursing and Health, 23,
search: A quantitative and qualitative ap- 246255.
proach. Boston, MA: Jones & Bartlett. Savage, J. (2000). Participant observation:
Robinson, R.R. (1997). Issues in clinical nurs- Standing in the shoes of others? Qualitative
ing research: You & research nursing Health Research, 10(3), 324339.
practice program. WJNR, 19(2), 265269. Sayre, J. (2000). The patients diagnosis: Ex-
Roethlisberger, F.J., & Dickson, W.J. (1939). planatory models of mental illness. Quali-
Management and the Worker. Cambridge, tative Health Research, 10(1), 7183.
MA: Harvard University Press. Schmitt, R. (1972). Phenomenology. In Paul
Rogers, E.M., & Shoemaker, F.F. (1971). Com- Edwards (Ed.). The Encyclopedia of Philos-
munication of Innovations: A Cross-Cultural ophy, 6, (pp. 135151). New York: Nacul-
Approach. New York: The Free Press. lian/Free Press.
Rogers, E.M. (1995). Diffusion Innovations (4th Schwartz, H., & Jacobs, J. (1979). Qualitative
ed.). New York: Free Press. Sociology: A Method to the Madness. New
Rosenbaum, J., & Carty, L. (1996). The sub cul- York: The Free Press.
ture of adolescence: Beliefs about care Schwarz, M., Landis, S.E., Rowe, J.E., Janes, C.L.,
health and individuation within Leiningers & Pullman, N. (2000). Using focus groups to
theory. Journal of Advanced Nursing, 23, assess primary care patients satisfaction.
741746. Evaluation & the Health Professions, 23(1),
Rosenthal, R. (1966). Experimenter Effects in 5871.
Behavioral Research. New York: Century. Seaman, C. (1987). Research Methods: Princi-
Rosenthal, R., & Fode, K.L. (1963). The effect of ples, Practice and Theory for Nursing. Nor-
experimenter bias on the performance of the walk, CT: Appleton & Lange.
albino rat. Behavioral Science, 8, 183189. Sennott-Miller, L., & C. Miller (1986). Magnitude
Rosenthal, R. (1984). Meta-analytic Procedures estimation: Issues and practical applica-
for Social Research: Applied Social Research tions. Western Journal of Nursing Research, 8,
Methods. London: Sage. 3140.
Rosenthal, R., & Rosnow, R.L. (1991). Essentials Sheppard, R. (1993). Yes, no, undecided or just
of Behavioural Research: Methods and Data hangs up? Globe and Mail, August 23, p. A11.
Analysis (2nd ed.). New York: McGraw-Hill. Sherman, D.W. (1996). Nurses willingness to
Rossman, G.B., & Wilson, B.L. (1985). Num- care for AIDS patients and spirituality, social
bers and words: Combining quantitative support, and death anxiety. Image, 28(3),
and qualitative methods in a single large- 205213.
scale evolution study. Evaluation Review, Sherrod, R. A. (1998). Infertility education in
9(5), 627643. baccalaureate schools of nursing. Journal
Rothe, J.P. (1994). Qualitative Research: A Prac- of Nursing Education, 37(9), 412415.
tical Guide. Heidelberg, Ontario: RCI Publi- Shields, S.A. (1988). Functionalism, Darwin-
cations. ism, and the psychology of women: A study
Rothman, D. J. (1982). Were Tuskegee and Wil- in social myth. In Ludy T. Benjamin, Jr.,
lowbrook studies in nature? Hastings Cen- (1988) A History of Psychology. New York:
ter Report, 12(2), 57. McGraw-Hill.
Silva, M. (1986). Research testing nursing the-
Sandelowski, M. (1986). The Problem of Rigor ory: State of the art. Advances in Nursing
in Qualitative Research, Advances in Nurs- Sciences, 90, 111.
ing Science, 8(3), 2737. Silverman, D. (1993). Interpreting Qualitative
Sandelowski, M., & Pollack, S. (1986). Qualita- Data: Methods for Analysing Talk, Text, and
tive analysis: What it is and how to begin. Interaction. London: Sage.
Research in Nursing and Health, 18, 371375. Silverman, D. (1997). Qualitative Research:
Sandelowski, M. (1993). Rigor or rigor mortis: Theory, Method and Practice. London: Sage.
The problem of rigor in qualitative research Simon, J.L., & Burstein, P. (1885). Basic Re-
revisited. Advances in Nursing Science, 16 search Methods in Social Science (3rd ed.).
(2), 18. New York: Random House.
730 BIBLIOGRAPHY
Smeenk, F. W., deWitte, L. P., vanHaastregt, J. ndings to practice. Nursing Outlook, 42(1),
C., Schipper, R. M., Biezemanus, H. P., & 1820.
Crebolder, H. F. (1998). Transmural care of Stetz, K.M., & Brown, M.A. (1997). Taking care:
terminal cancer patients: Effects on the Care giving to persons with cancer and
quality of life of direct caregivers. Nursing AIDS. Cancer Nursing, 20(1), 1222.
Research, 47(3), 129136. Stevens, B., Stockwell, M., Browne, G., Dent, P.,
Smith, D. E. (1987). The Everyday World as Gani,A., Martin, R., & Anderson, M. (1995).
Problematic: A Feminist Sociology. Toronto: Evaluation of a home-based traction pro-
University of Toronto Press. gram for children. Canadian Journal of Nurs-
Smith, P. (1997). Research Mindedness for Prac- ing Research, 27(4), 133151.
tice: An Interactive Approach for Nursing and Stevens, B.J. (1984). Nursing Theory: Analysis,
Health Care. New York: Churchill Living- Application and Evaluation (2nd ed.). Boston:
stone. Little, Brown.
Smith. (1991). Strategies of Social Research (3rd Stevens, P.E., & Hall, J.M. (1992). Applying crit-
ed.). St. Louis: Holt, Rivehart, & Winston. ical theories to nursing in communities.
Solchany, J.E. (1998). Anticipating the adopted Public Health Nursing, 9(1), 29.
child: Womens preadoptive experiences. Stevens, S.S. (1966a). A metric for the social
Canadian Journal of Nursing Research, 30(3), consensus, Science, 151, 53041.
123129. Stevens, S.S. (1966b). Matching functions be-
Sparkes, S., & Rizzoloo, M. (1998). World wide tween loudness and ten other continua.
web search tools. Image: Journal of Nursing Perception and Psychophysics, 1, 58.
Scholarship, 30(2), 167171. Stevens, S.S. (1951). Mathematics, measure-
Spiegelberg, H. (1975). The Phenomenological ment, and psychophysics. In S.S. Stevens
Movement. The Hague, Netherlands: Marti- (ed.), Handbook of Experimental Psychol-
nus Nijhoff. ogy. New York: Wiley.
Spradley, J.P. (1979). The Ethnographic Inter- Stevenson, B., Mills, E. M., Welin, L., & Beal, K.
view. New York: Holt, Rinehart and Winston. (1998). Falls risk factors in an acute care
Spradley, J.P. (1980). Participant Observation. setting: A retrospective study. Canadian
San Diego, CA: Harcourt Brace College Pub- Journal of Nursing Research, 30(1), 97113.
lishers. Stewart, M.J., Gillis, A., Brosky, G., Johnson, G.,
SPSS. (1999). SPSS Advanced Models 10.0. Chi- Kirkland, S., Leigh, G., Persaud, V., Rootman,
cago: SPSS Inc. I., Jackson, S., & Pawliw-Fry, B.A. (1996).
SPSS. (1999). SPSS Base 10.0 Applications Guide. Smoking among disadvantaged women:
Chicago: SPSS Inc. Causes and cessation. Canadian Journal of
SPSS Inc. (1999). SPSS Base 10.0 Users Guide. Nursing Research, 28, 4160.
Chicago: SPSS Inc. Stewart. M.J., Doble, S., Hart, G., Langille, L., &
SPSS. (1999). SPSS 10.0 Syntax Reference Guide. MacPherson, K. (1998). Peer visitor support
Chicago: SPSS Inc. for family caregivers of seniors with stroke.
Statistical Abstract of the United States (1994). Canadian Journal of Nursing Research, 30(2),
Washington, DC: US Government Printing 87117.
Ofce. Strauss, A., & Corbin, J. (1990). Basics of Quali-
Sterling, T.D. (1959). Publication decisions and tative Research: Grounded Theory Procedures
their possible effects on inferences drawn and Techniques. Newbury Park, CA: Sage.
from tests of signicanceor vice versa. Strauss, A., & Corbin, J. (1997). Grounded The-
Journal of the American Statistical Associa- ory in Practice. Thousand Oaks, CA: Sage.
tion, 54, 3034. Streubert, H., & Carpenter, D. (1995). Qualita-
Sterling, T.D., Rosenbaum, W.L., & Weinkam, J.J. tive Research in Nursing: Advancing the Hu-
(1995). Publication decisions revisited: The manistic Imperative. Philadelphia: J.B. Lip-
effect of the outcome of statistical tests on pincott.
the decision to publish and vice versa. The Streubert, H., & Carpenter, D. (1999). Qualita-
American Statistician, 49, 108112. tive Research in Nursing: Advancing the Hu-
Stern, P. (1998). Grounded theory workshop II. manistic Imperative (2nd ed.). Philadelphia:
Qualitative Health Research Conference, Van- J. B. Lippincott.
couver, British Columbia, February 1921. Streubert, H.J. (1991). Phenomenological re-
Stern, P.N. (1980). Grounded theory method- search as a theoretic initiative in commu-
ology: Its uses and process. Image, 12(1), nity health nursing. Public Health Nursing,
2023. 8(2), 119123.
Stetler, C. B. (1994). Renement of the Stetler/ Strunk, W., & White, E.B. (2000). The Elements
Marram Model for application of research of Style. Galt, Ontario: Brett-MacMillan.
BIBLIOGRAPHY 731
Sudman, S. (1967). Reducing the Cost of Surveys. Varcoe, C., & Hilton, A. (1995). Factors affect-
Chicago: Aldine Publishing. ing acute care nurses use of research nd-
Sudman, S. (1976). Applied Sampling. New York: ings. Canadian Journal of Nursing Research,
Academic Press. 27(4), 5173.
Sudman, S., & Bradburn, N. (1983). Asking Ques-
tions. San Francisco: Jossey-Bass. Walker, L., Fleschler, R. G., Heaman, M. (1998).
Sussman, S., Burton, D., Dent, C.W., Stacy, A.W., Determinants of a healthy lifestyle in new
& Flay, B.R. (1991). Use of focus groups in de- fathers. Canadian Journal of Nursing Re-
veloping an adolescent tobacco use cessa- search, 30(3), 2137.
tion program: Collection norm effects. Jour- Walker, L. (1998). Weight related distress in
nal of Applied Social Psychology, 21, the early months after childbirth. Western-
17721782. Journal of Nursing Research 20(1), 3044.
Sykes, G.M. (1968). The Society of Captives. Waltz, C., & Bausell, R.B. (1981). Nursing Re-
New York: Atheneum. search: Design, Statistics and Computer Analy-
sis. Philadelphia: F.A. Davis.
Teevan, J. J. (2000). Introduction to Sociology: Waltz, C., Strickland, O., & Lenz, E. (1991).
A Canadian Focus, (7th ed.). Scarborough: Measurement in Nursing Research (2nd ed.).
Prentice-Hall Canada. Philadelphia: FA Davis.
Thibaut, J.W., & Kelley, H.H. (1959). The Social Warwick, D.P., & Osherson, S. (1973). Compar-
Psychology of Groups. New York: Wiley. ative Research Methods. Englewood Cliffs,
Thomas, S. (2000). How to Write Health Sci- N.J.: Prentice-Hall.
ences Papers, Dissertations and Theses. New Waterman, A. Webb, L., & Williams, W.
York: Harcourt. (1995). Parallels and contradictions in the
Travelbee, J. (1971). Interpersonal Aspects of theory and practice of action research in
Nursing (2nd ed.). Philadelphia: F.A. Davis. nursing. Journal of Advanced Nursing, 22,
Travers, K. (1997). Reducing inequities through 779784.
participatory research and community em- Webb, C. (1984). Feminist methodology in nurs-
powerement. Health Education and Behavior ing research. Journal of Advanced Nursing, 9,
24(3), 344356. 249256.
Tri-Council (1998). TriCouncil Policy State- Webb, C. (1993). Feminist research. Denitions,
ment: Ethical Conduct for Research Involv- methodology, methods and evaluation. Jour-
ing Humans. Ottawa: Author. nal of Advanced Nursing, 18, 416423.
Trost, S., Pate, R., Saunders, R., Ward, D., Weitz, R. (1990). Living with the stigma of
Dowda, M., & Felton, G. (1997). A prospec- AIDS, Qualitative Sociology, 13(1), 2338.
tive study of the determinants of physical Weitzman, E., & Miles, J. (1995). Computer Pro-
activity in rural fth-grade children. Pre- grams for Qualitative Data Analysis. Thou-
ventive Medicine, 26, 257263. sand Oaks, CA: Sage.
Tuck, I., Harris, L., Renfro, T., & Lexvold, L. Whyte, W.F. (1955). Street Corner Society: The
(1998). Care: A value expressed in philoso- Social Structure of an Italian Slum (2nd ed.).
phies of nursing services. Journal of Profes- Chicago: University of Chicago Press.
sional Nursing, 14(2), 9297. Whyte, W.F., Greenwood, D., & Lazes, P. (1991).
Participatory Action Research: Through Prac-
UKCC(1993). Code of Professional Conduct for tice to Science in Social Research. Newbury
the Nurse, Midwife, and Health Visitor. Lon- Park, CA: Sage.
don: United Kingdom Central Council. Wilson, H. (1993). Introducing Research in Nurs-
Unger, J., Kipe, M., Simon, T., Johnson, C., Mont- ing (2nd ed.). Redwood City, CA: Addison-
gomery, S., & Iverson, E. (1998). Stress, cop- Wesley Nursing.
ing, and social support among homeless Wolf, Z.R., & Heinzer, M. M. (1999). Substruc-
youth. Journal of Adolescent Research, 13(2), tion: Illustrating the connections from re-
154157. search question to analysis. Journal of Pro-
fessional Nursing, 15(1), 3337.
Van Kaam, A. (1959). A phenomenological Woods, N. F., & Catanzaro, M. (1988). Nursing
analysis exemplied by the feeling of being Research: Theory and Practice. St. Louis, MI:
really understood. Individual Psychology, C. V. Mosby.
15, 6672.
Van Manen, M. (1990). Researching the Lived Ex- Zerwekh, J. (2000). Caring on the ragged edge:
perience: Human Science for an Action Sensi- Nursing persons who are disenfranchised.
tive Pedagogy. Ontario, Canada: Atthouse. Advances in Nursing Science, 22(4), 4761.

Index
Accessible population, dened, 497 Agency for Health Care Policy and Anecdotal evidence, as explanation
in sampling, 497498 Research, in research utiliza- aw, 43
Accuracy, dened, 435 tion, 637 Anglo-American participants, in com-
sample size and, 515 Aggregation, as unit of analysis, 34 parative study, 160, 161b, 162
Action research, challenges in, Agreement, in Likert-type questions, Annual Review of Nursing Research, in
268269 479b, 479480 research utilization, 637638
characteristics of, 264265 AIDS studies, bias and, 301 ANOVA See also Analysis of variance
choice of topics and, 74 as in-depth interview example, 232, SPSS procedure for, 676677, 678b
dened, 261 233b234b, 235 Antecedent variable, dened, 89
example of, 269, 270b271b outline of report of, 608 Antecedent variable model, as causal
in new-wave research, 264269 unethical, 337, 340341 model, 89, 89f
participatory, 265 Alabama, Tuskegee Syphilis Study in, Anxiety, Peplaus theory and, 56, 58f,
steps in, 265269 331b332b, 337339 59b
Adaptation to pregnancy study, as Alberta Association of Registered Applicability to practice, of research
example of comparative study, Nurses, survey of, 143b, 151 utilization, 636
160, 161b Alberta Survey, as sampling example, Applied research, dened, 33
Administrators, research utilization 504, 505b pure versus, 3233
by, 644 Alpha, in power analysis, 512 Appraisal, critical, 612 See also Re-
Adolescent attitudes and social inu- Alternative explanation, in testing of search critique
ence, as study example, 129, theories, 6365 Areas under normal curve, in de-
130b Alternative hypothesis, dened, 391 scriptive statistics, 366367,
Adolescent identity awareness study, Alternative therapy/family income 368t369t, 369370, 370f
as F test sample, 410, study, as chi-square sample Arithmetic mean, dened, 358
411b412b, 412414 test, 400t, 400401, 401b Art of nursing, as way of knowing, 11
Adolescent Lifestyle Questionnaire, American Nurses Association, code Articiality, as study limitation,
437438, 439t440t, 467469, of ethics of, 326, 327b, 329b 132133
468b nursing research dened by, 7 Asch, critique of study by, 620626
Adoption study, as example of phe- Analysis, data See Data analysis Asia, AIDS research in, ethics and,
nomenological perspective, factor, 577583 See also Factor 341
192, 194, 195b analysis Assent, to participate in research
Advertising in relation to cigarette in qualitative research, 2829 study, 333334, 335f, 345
smoking, linear regression in quantitative research, 27 Assessment of situation, in action re-
equations of, 561563 specifying methods of, 92 search, 266
562f565f7 units of, individuals versus aggrega- Associate degree, in nursing re-
Advocacy, pure research and, ethics tions, 34 search, 35
in, 347 Analysis of variance, correlation coef- Attitudinal continuum, bias and, 318
pure research versus, bias and, cient versus, 576, 576b Attributes, dened, in ethnographic
313317 dened, 376 perspective, 211
Aesthetics, dened, 11 in F test, 410 Audience, for research reports,
in questionnaire presentation, one-way, 410, 411b412b, 590591
491492 412414 Audit trail, in qualitative research,
as way of knowing, 11 repeated measures, 416 bias and, 299
Affect, researcher, bias and, 302 two-way, 414, 415b dened, 216
Africa, AIDS research in, ethics and, means comparisons in, 376377, Authority, illegitimate appeal to, as
341 377t explanation aw, 4142
African-Americans, in comparative multivariate, 573577 See also Mul- as source of nursing knowledge, 9
study, 160, 161b, 162 tivariate analysis of variance Average, dened, 358
in secondary data analysis, SPSS procedure for, 676677, 678b Axiomatic derivations, in testing of
171172, 173b Analytic les, in eld studies, 231 theories, 6162, 62b
Tuskegee Syphilis Study in, Androcentricity, bias and, 300301 Azidothymidine, in unethical AIDS
331b332b, 337339 dened, 300 studies, 337, 340341
734 INDEX
b coefcients, in linear regression need for awareness of, 316 Categorical imperative, dened,
equations, 561563, 562f565f, research, dened, 298 325326
569, 570t research process and, 302313 ethics and, 325326
Baccalaureate, in nursing research, data analysis in, 306308 Causal explanation, dened, 40
35 data collection in, 303306 Causal models, in data analysis, SPSS
research utilization and, 638 funding in, 311312 in, 539
Background le, in eld studies, 231 problem selection in, 302303 dened, 84
Background information questions, publication decisions and, development of, 8590, 86f91f
in questionnaire development, 308310, 309b four-variable, 556558
487488 reporting of ndings in, 308311, multivariate, 8990, 90f91f
BACKWARD solution, in multiple re- 309b in research report example, 597f
gression analysis, 567 sample selection in, 303 specifying, 8489, 86f91f
Bar graphs, in research reports, 604f use of ndings in, 312313, 314b testing of, 545558
Baseline assessment, in within-sub- study participation and, ethics of, four-variable, 556558
jects research design, 122 344 three-variable, 545556
Baseline data, in between-subjects re- understanding of, 297321 three-variable, 87f89f, 8789,
search design, 117 Biased sample, dened, 498 545556 See also Three-vari-
dened, 117 Bimodal distribution, dened, 356 able causal models
Baseline stability, dened, 117 in descriptive statistics, 356, 357f two-variable, 8687
Basic social psychological process, Birth order and sexual permissive- Causal relationship, in between-sub-
in data analysis, in grounded ness, axiomatic derivation jects research design, 118120
theory perspective, 198, 200, and, 62b Causal sequence, in causal model
200t Blocking, in between-subjects retesting, 545
Behavior, critical perspective on, 22 search design, 117 Censuses, 136 See also Survey de-
ethnographic perspective on, 208 dened, 117 signs
interpretive perspective on, 2021 Blood pressure/touch therapy study, Central tendency, dened, 357
positivist assumptions on, 1719 as t test sample, 404, in descriptive statistics, 357359
theory of planned, 5657, 60f, 61b 405b406b, 406408, mean in, 358359, 359t
Benecence, ethics and, 330, 332 408b410b median in, 358, 358t
Bereavement study, investigator bias Blumer, symbolic interactionism and, mode in, 357t, 357358
in, 313, 314b 196 Challenges, in research, preliminary
Beta, in power analysis, 513 Booth, surveys of, 136 considerations in, 9698
Beta coefcients, in linear regression Borrowed theories, in theoretical Change agents, dened, 641
equations, 562, 569, 570t frameworks, 54 in research utilization, 641642
Between-subjects research design, Bracketing, dened, 183 Charts, in research reports, 600, 605f
112120 in qualitative research, 182184 Chinese postpartum practices study,
advantages and limitations of, bias and, 299 as ethnographic example,
131133 Branching of questions, in question- 206b, 207, 213214
baseline data in, 117 naire formatting, 491, 491b Chi-square test, 396403
blocking in, 117 Breast milk sodium study, as sec- dened, 396
causal relationship in, 118120 ondary data analysis example, elements of, critical value in, 397
control group in, 113f, 113114 165, 172173, 173b formula for, 397
data analysis in, 117t, 117118 Bricolage, qualitative research as, observed and expected frequen-
dened, 112 180, 180b cies in, 396397
example of, 121b hand computation in, 397403
experimental group in, 113f, Campbell and Stanley, study of, sample tests of, 397, 398b399b,
113114 109112 399t, 399401, 401b403b, 403
precision matching in, 116117 Canadian Institute for Health Re- Cigarette smoking See also Smoking
randomization in, 114, 115t116t, search, in research utilization, advertising and, linear regression
116 637 equations of, 561563
synonyms for, 112 Canadian Nurses Association, code of 562f565f
within-subjects design versus, ethics of, 326, 327b in planned behavior theory exam-
123124, 125t, 126 Candidate variable model, as causal ple, 61b
Between-subjects t test, 404, model, 89, 90f Clairmont and Jackson, study of, 163,
405b406b dened, 89 163b
Bias, 295348 Caregivers, effects in terminal cancer Clarity, in questionnaire develop-
advocacy versus pure research on, as study example, ment, 482b, 483485
and, 313316 128b129b Classic experimental research de-
dened, 298 Caregiving study, as grounded theory signs, 112126
elite, 299 perspective example, 201, advantages and limitations of,
ethical issues in, 322348 See also 202b 131133, 133t
Ethics Caribbean, AIDS research in, ethics Cleaning of data, SPSS in, 537539
guidelines for minimizing, 316318 and, 341 Clever Hans, bias and, 304
investigator, in bereavement study, Caring and uncaring experience study, Client safety, research utilization
313, 314b as example of phenomenologi- and, 636
in Likert-based indexes, 458 cal perspective, 192, 194b Clinical nurse specialists, in research
myth versus reality and, 318320, Case numbers, in multiple regression utilization example, 646647
319t analysis, 568 Clinical practice, research topics
nature of, 298302 Case studies, dened, 178 and, 69, 71
in qualitative and quantitative re- evaluation, 274 Clinical signicance of ndings, in re-
search, 299 Casual touch, therapeutic touch ver- search reports, 599
sexism as, 300302 sus, in study comparison ex- Closed-ended questions, dened, 475
triple, 299300 ample, 123124, 125t, 126 in questionnaires, 474b475b, 475
INDEX 735
Cluster sampling technique, 503504, Comparative surveys, comparative Conceptual models, as grand theo-
550b studies versus, 157 ries, 47
Cochrane Database of Systematic Re- Comparison group research design, Conceptual utilization, dened, 633
views, in research utilization, pre-experimental, 110112, Conceptual variable, dened, 25
637 111f Conceptualizing, of research, 68, 7b
Codes, 326, 327b survey design versus, 136 Conclusions, in eld study approach,
of conduct, dened, 324 Compatibility of innovation, in re- computers and, 249
ethics as, 323324 search utilization, 641 in research reports, 605
ethical, 326330 Complexity of innovation, in research critical appraisal of, 616b, 619b,
American Nurses Association, utilization, 641 625626, 629630
326, 327b, 329b Componential data analysis, in ethno- unwarranted, as explanation aw,
Canadian Nurses Association, graphic perspective, 211t, 44
326, 327b 211212 Concurrent validity, dened, 428
Declaration of Helsinki, 328, 330 COMPUTE, SPSS procedure for, test for, 429t
International Council of Nurses 677679, 679f Conditional variables, dened, 52
Code as, 326 Computer-assisted telephone inter- Conditioning of respondents, in ques-
Nuremberg, 326, 328, 329b viewing, mechanics of, tionnaire formatting, 489
United Kingdom Nursing Associ- 153154 Conduct and Utilization of Research
ation, 326, 328b Computers, in eld studies, advan- in Nursing model, of research
missing value, in data entry, 533, tages and disadvantages of, utilization, 634b, 643
535536 252253 Conduct codes, 324
Coding, in data analysis, bias-related data analysis with, 248253 ethics as, 323324
errors in, 306307 data entry in, 231, 248 Condence limits, sample size deter-
in grounded theory perspective, NUD*IST software in, 249251, mination and, 509510, 515
198 250b, 251f, 252t Condentiality, ethics and, 330, 342
dummy, in multiple regression qualitative data analysis in, Conrmability, in qualitative re-
analysis, 566 248253 search, dened, 216217
in eld studies, computers and, 249 software available for, 248 test for, 217b
questionnaire formatting and, software functions in, 249 Conrmatory factor analysis, dened,
489491, 490b in nursing research, 1517 579
Coefcients, in discriminant function in power analysis, 515 Confounding variables, dened, 53,
analysis, 572, 572t in quantitative data analysis, 107
in linear regression equations, 526542 in experimental research designs,
561563, 562f565f, 569570 analysis techniques in, 539540, 107108
of reliability, in content analysis, 540t between-subjects, 119
255256 data cleaning in, 537539 Conscious partiality, in feminist re-
Collaboration, in action research, 266 data entry in, 531533 search, 282
Combined mother and baby care errors and warnings in, 540542, Consciousness, in feminist research,
study, as evaluation research 542t 282
example, 275, 277b le creation in, 534537 in phenomenology, 189
Coming of age study, as example of overview of, 526527 Consent, informed See Informed con-
comparative study, 157, 158b, SPSS in, 527528, 528f529f, 530, sent
159160 534542, 537b, 540t, 542t See Consent forms, ethics and, 333336,
Common sense reasoning, as source also SPSS for Windows 334f335f
of nursing knowledge, 10 questionnaire formatting and, in questionnaire development, 487,
Communication, of ndings, as quan- 489491, 490b 488f
titative research stage, 1314 research problems concerning, 97 in secondary data analysis, 170,
in grounded theory perspective, Concept integration, in data analysis, 171f
196 in grounded theory perspec- Consistency, critical appraisal of, 614
precise, quantitative evidence and, tive, 201 in data collection, 542
41 Concept modication, in data analy- Constant comparative method, of
Community, in action research, 266 sis, in grounded theory per- data analysis, in grounded the-
challenges of, 268 spective, 200201 ory perspective, 198
in health promotion research, Concepts, constructs as, 49 Construct validity, dened, 428
279280 continuum of, 50f test for, 429t
Community empowerment study, as dened, 49, 49f, 427 Constructs, in concept continuum,
action research example, 269, research project level and, 427 50f
270b271b in theory, 4953 as concepts, 49
Community health, discriminant variables as, 4953, 50f52f See also dened, 427
function analysis in, 571 Variables in instrument development, 467
Comparative analysis, questionnaire Conceptual denitions, in question- in qualitative research, validity
wording and, 485 naire development, 482483 and, 434
Comparative judgments, magnitude Conceptual framework, dened, 54 research project level and, 427
estimation procedures and, selection of, as quantitative re- Contemporary applied research,
458459 search stage, 13 263293 See also New-wave re-
Comparative studies, challenges in, Conceptual hypothesis, dened, 25, search
160, 162164, 163b 425 dened, 15
comparative surveys versus, 157 research project level and, 425 example of, 15, 16b
examples of, 157, 158b, 159160, Conceptual level, in quantitative re- goals of, 15
161b search, 25 Content analysis, dened, 253
in feminist research, 281282 of research project, 425, 425f in eld study approach, 253256,
questions asked in, 157158 Conceptual maps, dened, 5455 254b
in survey designs, 156164 examples of, 56f, 58f, 60f basic decisions in, 255256
736 INDEX
Content analysis, denedContinued Credibility, in qualitative research, publicly available, as solution to

computers and, 249 dened, 215216, 433 research problem, 97
examples of, 256, 257b258b test for, 217b raw See Raw data
Content validity, dened, 428 Creswell, qualitative research de- secondary, 136 See also Secondary
test for, 429t ned by, 181 data
Context, feminist perspective and, Criterion validity, dened, 428 Data analysis, in action research,
284, 286287 test for, 429t planning and, 266
in qualitative research, 187 Critic, research, 612 in between-subjects research de-
Contradictory sayings, explanations Critical appraisal, criteria for, in qual- sign, 117t, 117118
and, 38, 39b itative research reports, 614, bias and, 306308
Contrast questions, in in-depth inter- 616617, 617b619b, 619620 coding errors in, 306307
views, in eld study approach, in quantitative research reports, data massaging in, 307
232 613614, 615b616b guidelines to minimize, 318
Control by constancy, dened, 120 dened, 612 hunting in, 307308
in within-subjects research design, Critical perspective, in nursing sci- in content analysis, 256
120 ence, 2223, 24t critique of, criteria for, 614, 615b,
Control group, in between-subjects Critical value, in chi-square test, 397 618b
research design, 113f, 113 dened, 397 examples of, 623, 628
114 Critique, research, 611631 See also in evaluation research, 273274
Control variables, in cross-tabulation Research critique in feminist research, 286287
tables, 373 Cronbachs alpha, factor analysis ver- in eld studies, 231
dened, 51f52f, 5153, 107 sus, 582 in focus group interviews, in eld
in experimental research designs, internal consistency and, 437438, study approach, 236
107 439t manual techniques in, 522f523f,
Controlled observations, quantitative Crossover designs, example of, 120, 522526, 524t
evidence and, 41 122f, 122123 NUD*IST software program and,
Convenience sampling, 506 in within-subjects research design, 252t
Core category variable, in data analy- 120, 122f, 122123 in qualitative research, 185
sis, in grounded theory per- Cross-sectional studies, dened, 94, ethnographic, 210213, 211t
spective, 201 139 in eld study approach, 248
Core variable, in data analysis, in in survey designs, 139 256
grounded theory perspective, CROSSTABS See also Cross-tabula- grounded theory, 198, 199f, 200t,
198, 200, 200t tion 200201, 202b
Correlation coefcient, analysis of Cross-tabulation, in causal model phenomenological, 191
variance versus, 576, 576b testing, 545550, 546t, 549t, 555 in quantitative research, 185 See
dened, 382 in chi-square sample test, 401, also Quantitative data analysis
in descriptive statistics, 382, 402b403b, 403 techniques in, 519584
383f384f, 384385 in descriptive statistics, 371375 questionnaire development and,
calculation of, 379b, 382, 384f, gamma in, 374t, 374375 481482
384385 lambda in, 373t, 373374 secondary See also Secondary data
Correlation matrix, in factor analysis, rules for constructing, 371373 analysis
579 summaries of, 600, 601t in survey designs, 164173
Correlational analysis, dened, 377 Cultural bias, in bereavement study, Data cleaning, SPSS in, 537539
in descriptive statistics, 377385 314b Data collection, bias and, 303304
linear equation in, 377, 380f381f, Cultural contradictions, as universal demand characteristics in,
380t, 380382 themes, in ethnographic per- 305306
Correlations, in causal model testing, spective, 212 environmental and cultural
552, 552t, 557 Cultural diversity, in comparative norms in, 305
partial, 385386 studies, 161163, 163b expectancy in, 304305
in summary tables, 600, 603t Cultural scene, dened, in ethno- halo effect in, 306
Cost, bias and, 311312 graphic perspective, 210 critique of, criteria for, 614, 615b,
ethics and, utilitarian view of, Cultural themes, dened, in ethno- 618b
325 graphic perspective, 212 examples of, 622623, 628
of focus group interviews, 239 Culture, bias and, 299300 See also dened, 450
of phone surveys, 153 Bias equipment list for, 452
sampling and, rationale for, 496 data collection in, 305 in evaluation research, 273
sample size and, 511 dened, 204 measurement and, 423469 See
of secondary data analysis, 175 in ethnography, 204205 also Data measurement; spe-
Counterbalanced research design See Current issues, research topics and, cic methods
also Crossover designs 7273 procedures for, 450466
dened, 122 Curvilinear relationship, in correla- example of, 450, 451b
Counterbalancing, dened, 119 tional analysis, 382 indexes in, 452458, 453t, 456b
in experimental research designs, Customs, as source of nursing knowl- interviews in, 466
between-subjects, 119 edge, 89 magnitude estimation in,
Court testimony, bias and, 312313 458461, 460b
Covariates, in multivariate analysis of Data, analysis of See Data analysis observational measurement in,
variance, 574, 574b borrowed, as solution to research 465466
Cover letter, in questionnaires, 151, problem, 97 physiological measures in,
487, 487b, 488f cleaning of, SPSS in, 537539 463465
Covert research methods, ethics of, collection of See Data collection questionnaires in, 466
346 entry of, 534537 See also Data en- research reports and, 596597
in participant observation studies, try semantic differential, 458, 458b,
228229 error checking, 442 459t
INDEX 737
visual analogue scales in, Deciding what to study, 68101 See data distributions in, 354t355t,
461463, 462f, 464b also Research 354356, 355f357f
in qualitative research, 184 Decision making, ethical, 324336 data standardizing in, 360364,
ethnographic, 209210 assessment of acceptability and, 362t364t
grounded theory, 197198 324326 dispersion measures in, 359360,
phenomenological, 191 codes and guidelines for, 326, 360t361t
in questionnaires See also Ques- 327b329b, 328, 330, normal distribution in, 364370
tionnaires 331b332b See also Normal distribution
precision in, 473474 protection methods in, 333336, measurement levels in, 351353
in research design, description of, 334f335f nominal, 353
93 research principles in, 330333 ordinal, 353
in research process, nursing Declaration of Helsinki, 328, 330 ratio, 353
process and, 4 Deductive explanation, dened, relationships between variables in,
in survey designs, 136 3839 370388, 371t
Data display, in eld study approach, Deductive reasoning, dened, 10 correlational analysis in, 377385
computers and, 249 as source of nursing knowledge, 10 cross-tabulation in, 371375,
Data distributions, dened, 354 Denitions, conceptual, in question- 373t374t
in descriptive statistics, 354t, naire development, 482483 data plotting in, 385
354356 Degrees of freedom, dened, 396 means comparisons in, 375377,
bimodal distribution in, 356, 357f in multivariate analysis of variance, 376t377t
egalitarian scores in, 355, 356f 574b partial correlations in, 385386
frequency distribution tables in, in tests of statistical signicance, Spearman correlations in, 385
354, 355f, 355t 396 Designs, research See Research de-
leptokurtic distribution in, 356 Delphi technique, in expert sampling, signs
normal distribution in, 355356 507 Deutscher, views on attitudes and be-
platykurtic distribution in, 356 in visual analogue scales, 462463, havior of, 138, 139b
Data entry, 531534 464b Developing countries, AIDS research
creation of raw data le in, 531534 Demand characteristics, bias and, in, ethics and, 341
point-and-click method in, 534 305306 Diagrams, in hypothesis specica-
dened, 531 dened, 305 tion, 9091, 91f
double, 534 Denzin and Lincoln, qualitative re- Dialectic critique, in action research,
error checking in, 534 search dened by, 180b, 267
Data linking, in eld studies, comput- 180181 dened, 267
ers and, 249 Deontological perspective, dened, Differential, semantic, 458, 458b
Data massaging, bias and, 307, 318 325 Dilemma, false, as explanation aw,
dened, 307 ethical decision making and, 42
Data measurement See also Data col- 325326 Disclosure, study participation and,
lection; specic methods Department of Health and Human ethics of, 344
errors in, 440443, 443f Services, ethical guidelines of, Discomfort, study participation and,
instrument development for, 337 ethics of, 343
466469, 468b Dependability, in qualitative re- Discrepancy, dened, 71
levels in, 443446, 444t446t, 447b search, dened, 216 research topics and, 71
effects of reduced, 446447, test for, 217b Discriminant function analysis,
448b449b, 450 Dependent variables, in candidate 570573
research project levels and, variable model example, 90f dened, 570
424427, 425f426f in cross-tabulation tables, 371372 multiple regression analysis versus,
linkages between, 427440, 429t, dened, 5051, 107 571, 572t
436t, 439t440t in experimental research designs, presenting and interpreting results
Data plotting, in descriptive statis- between-subjects, 118120 of, 571573, 572t573t
tics, 383f, 385 in manual data analysis, 524t, rationale for, 570571
Data recording, in content analysis, 524525 Discussion, in research reports,
256 in multivariate analysis, SPSS in, 604605
in focus group interviews, in eld 583, 584t critical appraisal of, 616b, 619b,
study approach, 236 of variance, 573 624625
Data reduction, in factor analysis, 578 in questionnaire development, Dispersion measures, in descriptive
Data sampling, in data analysis, in 480481, 581t statistics, 359360, 360t
grounded theory perspective, research topics and, 73 Diversity, in action research, chal-
201 summarizing mean values of, 600, lenges of, 268
Data sets, in secondary data analysis, 601t in feminist research, 283284
168171, 169f171f validity and, 430431 bias and, 300 See also Bias
Data standardization, in descriptive Derivations, axiomatic, in testing of Doctoral degree, in nursing research,
statistics, 359360, 362t364t theories, 6162, 62b 35
Databases, for literature review, Descriptions, in eld notes, 230 Document system, in NUD*IST soft-
80b82b Descriptive questions, in in-depth in- ware program, 250, 250b
published, as solution to research terviews, in eld study ap- Domain data analysis, in ethno-
problem, 9798 proach, 232 graphic perspective, 210
in research utilization, 637 Descriptive research, dened, 31 Dominant/less-dominant design, in
in secondary data analysis, predictive versus, 3132 research studies, 3031
167168 Descriptive statistics, 351388 Double-blind study, dened, 123, 305
Debrieng of study subjects, ethics dened, 349, 351 Dracup-Breu model, of research uti-
and, 346 individual variables in, 354370 lization, 634b
Deception, study participation and, central tendency measures in, Drop-box returns, in questionnaire
ethics of, 344 357t359t, 357359 surveys, 145
738 INDEX
Drop-off form, in questionnaire sur- random, 306 Evidence selection, in comparative

veys, 145, 146t sampling, 498 studies, 163164
Drug use by gender study, as chi- in SPSS, 540542, 542t Evidence, anecdotal, as explanation
square sample test, 401, systematic, 306307 aw, 43
402b403b, 403 types I and II, dened, 395 missing, as explanation aw, 4243
Duffy and Hedin, feminist research in tests of statistical signicance, quantitative, in nursing science ex-
dened by, 282283 395396 planation, 4041
Dummy coding, in multiple regres- Essences, in phenomenological per- selected or suppressed, as expla-
sion analysis, 566 spective, 188 nation aw, 43
Dummy variables, in multiple regres- Ethical codes, 326330 See also spe- Evidence-based practice, dened, 4,
sion analysis, 565567, 566t cic codes 633
Duty to care, duty to advance knowl- Ethics, 322348 research utilization in, 633
edge versus, ethics and, 332 of covert group entry, in partici- survey study on, 143b
Dynameter, dened, 110 pant observation studies, Ex post facto study, in survey de-
228229 signs, 138
Editing, in eld studies, computers decision making and, 324336 Existentialism, dened, 189
and, 249 assessment of acceptability in, Expectancy, bias and, 304305
of research reports, 591592 3240326 dened, 304
Education, research and, 3435 codes and guidelines in, 326, Expectations, measurement error
utilization of, 638639 327b329b, 328, 330, and, 442
Educators, research utilization by, 331b332b Expected frequencies, in chi-square
643 principles of, 330, 332333 test, 396397
Effect size, in power analysis, protection methods in, 333336, Expenses See Cost
513t514t, 513515 334f335f Experience, in phenomenology, 189
Egalitarian scores, in data distribu- dened, 323324 Experimental group, in between-sub-
tion, 355, 356f in naturalistic observational stud- jects research design, 113f,
Egalitarianism study, as multivariate ies, 242 113114
analysis of variance example, in research design choice, 95 Experimental research designs,
574, 574b575b research problems concerning, 97 105126, 134
Eigenvalue, in factor analysis, 579, research rules and, 341348 advantages and limitations of,
581t for participants, 342346 131133, 133t, 174t, 221t
Elite bias, dened, 299 for socially responsible nursing basics of, 106108
in qualitative research, validity research, 346348 classic, 112126
and, 434 studies showing dilemmas in, between-subjects, 112120
Emancipatory potential of story- 336341 comparisons among, 123126
telling study, as secondary AIDS research as, 340341 within-subjects, 120123
data analysis example, Tuskegee Syphilis Study as, dened, 14, 106
171172, 173b 337339 ethics and, 346
Emergency room waiting study, as Willowbrook Study as, 339340 key elements in, 106108
data analysis example, as way of knowing, 11 manual data analysis in, 525526
522f523f, 522525, 524t Ethics review board, dened, 336 pre-experimental, 108112
Emic view, dened, 203 purpose of, 336 rationale for, 106
Emotional bias, in bereavement Ethnicity, in comparative studies, tests of signicance and, 417
study, 314b 161163, 163b Experimenter effect, bias and,
Empathetic explanation, dened, 40 Ethnographic interviews, in-depth in- 303304
Empathy, as way of knowing, 11 terviews in, 232 Experiments, dened, 106
Empirics, as way of knowing, 1112 Ethnographic perspective, partici- eld, 239
Empowerment, in feminist research, pant observation studies in, in eld study approach, 239241,
289 225 241b
as nursing research goal, 8 in qualitative research, 203214 Expert sampling, 507
Entry to study group, in participant application of, 205213, 206b, Expert witnesses, bias and, 312313
observation studies, 227229 207t, 211t Explanations, 3844
Envelopes, in questionnaire surveys, denitions in, 203204 alternative, in testing of theories,
145, 151 examples of, 206b, 207, 207t, 6365
Environment, bias and, 305 213214, 214b contradictions and, 38, 39b
Epidemiology, discriminant function philosophy of, 205 aws in, 4144
analysis in, 571 Ethnographic research, phases of, nursing science, 4041
Epistemology, dened, 10 214b theory and, 3844
as way of knowing, 1011 Ethnography, dened, 203 types of, 3840
Equal interval, as measurement level, Ethnonursing, dened, 204 Explanatory research, dened, 31
446, 447b Etic view, dened, 204 predictive versus, 3132
Equivalence, dened, 438 Evaluation, as action research step, Exploratory factor analysis, dened,
in quantitative research, 438 267 578579
tests of, 436t research topics and, 7374 Exposed group research design, pre-
Equivalence of concepts, in compara- Evaluation case study, dened, 274 experimental, 110112, 111f
tive studies, 162 in evaluation research, 274 survey design versus, 136
Equivalence of indicators, in compar- Evaluation research, dened, 269 External validity, dened, 106, 428, 604
ative studies, 162 examples of, 275, 276b277b in research report discussion,
Equivalence of language, in compara- in new-wave research, 269, 271 604605
tive studies, 162163, 163b 275 as study limitation, 132
Error checking, in data entry, 534 pluralistic approaches to, 274275
Errors, coding, bias and, 306307 purpose of, 269272 F distribution, dened, 408
measurement, 440443, 443f steps in, 272274 F test, 408416
INDEX 739
analysis of variance in, one-way, in-depth interviews in, 232234, Functional explanation, dened, 40
410, 411b412b, 412414, 413t 233b234b Funding, bias and, 311312
repeated measures, 416 naturalistic observational stud- research problems concerning, 97
two-way, 414, 415b ies in, 241248, 243b244b, Funnel technique, in ordering of
F distribution in, 408, 410 246f247f questionnaire items, 488489
F value, in multivariate analysis of participant observation studies Future planning, in action research,
variance, 574b, 577 in, 225232 267268
Face validity, dened, 428 qualitative data analysis in,
test for, 429t 248256, 250b, 252t, 254b, Gallup, polls of, 136
Faceism study, as example of content 257b258b Gamma, in cross-tabulation tables,
analysis, 256, 257b258b dened, 178, 225 374t, 374375
Factor, in factor analysis, 577 limitations of, 258260, 259t Gatekeepers, in ethnographic per-
Factor analysis, 577583 in research design, 15 spective, 208
conrmatory, 579 Fieldwork phase, of ethnographic re- Gender bias, feminism and, 281289
Cronbachs alpha versus, 582 search, 214b See also Feminist research
dened, 577 Figures, in qualitative reports, 600, Gender insensitivity, bias and,
exploratory, 578579 604f 301302
presenting and interpreting results File types, in eld studies, 231 dened, 301
in, 580582, 581t Findings, bias and, 312313 Gender-neutral terminology, 594595,
rationale for, 578579 critique of, criteria for, 614, 616b, 595t
regression analysis versus, 582583 619b Generalizability, in research report
terminology of, 579580 examples of, 624, 629 discussion, 604
Factor extraction, in factor analysis, as qualitative research report com- Generalizations, bias and, 298
579 ponent, 608 examples of, 40
Factor loadings, in factor analysis, reporting of See Reporting of nd- Glaser, theoretical coding of, 201
579, 581t, 582 ings Goal-directed research designs See
Factor scores, in factor analysis, Flexibility, in research design choice, also specic design types
579580, 582 96 advantages and limitations of, 290t
Factorability, in factor analysis, 579 Focus group See also Focus group in- Goals, dened, 8
Factor-isolating theory, dened, 48 terviews of nursing research, 8
Factor-relating theory, dened, 48 dened, 234 Goyder model, of estimating ques-
Fallacies, about research among Focus group interviews, in eld study tionnaire response rate to,
nurses, 638640, 639t approach, 234239 148, 149b150b, 150151
False dilemma, as explanation aw, 42 advantages and disadvantages Grand theories, dened, 47
Familism, bias and, 301302 of, 236, 239 examples of, 55b
dened, 301 data analysis in, 236 in theory continuum, 46f
Family, in health promotion research, data recording in, 236 Grandmothers in health promotion
277279, 279b evolution of, 235 study, in summary table exam-
Family adjustment to heart transplan- example of, 236, 237b238b ple, 603t
tation, as grounded theory ex- mechanics of, 235 Graphic mapping, in eld study ap-
ample, 202203, 203b, 218, 218b member selection in, 236 proach, computers and, 249
Feasibility, in research design choice, rationale for, 235 Graphs, in research reports, 600, 604f
95 Folk wisdom, testing of, as research Grounded theory perspective, eld
of research utilization, 636 topic, 72 notes and, 230
Feminism, dened, 281 Folklore, as source of nursing knowl- in qualitative research, 194,
Feminist perspectives, bias in, 300 edge, 89 196203
See also Bias Follow-up communications, question- application of, 196202, 199f,
dened, 283 naire response rate and, 148, 200t, 202b
in feminist research, 283 151 dened, 194
Feminist research, dened, 282283 Food purchasing study, as action re- example of, 202203, 203b
examples of, 16b, 287289, 288b search example, 269, philosophy of, 196
methodology of, 284287 270b271b Group storytelling study, as sec-
in new-wave research, 281288 Forchuk and Roberts, critical ap- ondary data analysis example,
overview of, 281283 praisal criteria of, 617b 171172, 173b
philosophical tenets of, 283284, Formative evaluation, dened, 272 Group-administered questionnaires,
284b in evaluation research, 272 in survey designs, 145147
Feminist theory, critical perspective Formative evaluation study, as evalu- Guba and Lincoln, quality standards
and, 22 ation research example, 275, of, 218219, 219b
Field, data collection in, 184 276b
Field experiments, dened, 239 Formatting of questions, in question- Halo effect, bias and, 306
in eld study approach, 239241 naire development, 489491, dened, 306
examples of, 240, 241b 490b491b Hand calculations, in quantitative
Field notes, computers and, 249 evaluation of, 492t data analysis, 522f523f,
dened, 229230 Four-variable causal models, compo- 522526, 524t
in participant observation studies, nents of, 556, 556f Harding, feminist philosophy of, 282
229231 testing of, 556558, 557f Harm, study participation and, ethics
Field studies, advantages of, 256, 258, Frameworks, theoretical, 5358 of, 343344
259t Freedom from organizational con- Havelock, Theory of Planned Change
approaches to, 224261 straints, research utilization of, 642, 644, 646, 646b
eld experiments in, 239241, and, 640 Hawthorne effect, in within-subjects
241b Frequency distribution table, in de- research design, 123
focus interviews in, 234239, scriptive statistics, 354, 355f, Health care, as nursing research goal,
237b238b 355t 8
740 INDEX
Health care agencies, value of re- dened, 53 dened, 452

search utilization to, 635b formulation of, as quantitative rein NUD*IST software program, 250,
Health promotion, discriminant func- search stage, 13 250b
tion analysis in, 571 in naturalistic observational stud- in research reports, 597598
research on See Health promotion ies, 245 Indexes of periodicals, for literature
research null, 391392, 392b review, 80b82b
Health promotion in grandmothers positivism and, 18 Individual, as unit of analysis, 34
study, in summary table exam- research, 391392 Individual wellness, in health promo-
ple, 603t in research reports, 596, 597f tion research, 277278
Health promotion research, charac- evaluation in, 600, 603604 Individualism liberalism study, as fac-
teristics of, 277278 in secondary data analysis, 165 tor analysis example, 580582,
dened, 275276 secrecy regarding, in questionnaire 581t
example of, 278, 279b surveys, 142, 144 Individually delivered question-
future directions in, 280281 specifying of, in preliminary re- naires, in survey designs,
methods used in, 280 search stages, 9092 144145, 146t
in new-wave research, 275281 testing of, theory testing versus, Inductive reasoning, dened, 10
scope of, 278280 60 as source of nursing knowledge, 10
Health promoting lifestyle prole, in Inequity reduction study, as action
health promotion research, Identication of issue, in action re- research example, 269,
279 search, 266 270b271b
Healthy Cities Initiative, as health Identication of objectives, in evalua- Inferential statistics, 389419
promotion research, 279280 tion research, 273 dened, 349, 390
Heart transplantation adjustment Identication of problem See Prob- overview of, 390391
study, as grounded theory ex- lem identication tests of statistical signicance in,
ample, 202203, 203b, 218, Identication with respondents, in appropriateness of, 416419
218b qualitative research, validity chi-square, 396403 See also Chi-
Hedin and Duffy, feminist research and, 434 square test
dened by, 282283 Identity awareness in adolescents degrees of freedom in, 396
Hepatitis, in Willowbrook Institution study, as F test sample, 410, F test as, 408416 See also F test
study, 337, 339340 411b412b, 412414 key elements in, 391396
Hidden coding, in questionnaires, Illegitimate appeal to authority, as ex- null hypothesis in, 391392, 392b
ethics of, 344 planation aw, 4142 one- and two-tailed, 393394
Hierarchical regression, in multiple Impartiality, need for, 315316 See probability levels in, 395
regression analysis, 567568 also Bias research hypothesis in, 391392
Hierarchical tree, in NUD*IST soft- Impersonal social relations, as uni- sampling in, 392393, 393f394f
ware program, 250251, 251f versal themes, in ethnographic signicance levels in, 394395,
High school lifestyle study, in causal perspective, 212 395f
model testing, 545556, 546t, Implementation, in action research, steps in, 396
548t549t, 551t552t, 554t555t 266267 t test as, 403408 See also t test
Hii&trade, as resource, 100 Implications, in research reports, 605 type I and II errors in, 395396
History, in qualitative research, valid- critical appraisal of, 616b, 619b, Informants, dened, 184
ity and, 434 625626, 629630 in ethnographic perspective, 208
as threat to internal validity, 109 Inadvertent bias, in research reports, Information, informed consent and,
History le, in eld studies, 231 314 333334
Holistic ethnography, dened, 204 Independent variables, in candidate Informed consent, ethics and,
Holistic fallacy, in qualitative re- variable model example, 90f 333336, 345346
search, validity and, 434 in cross-tabulation tables, 372 in naturalistic observational stud-
Holistic perspective, in qualitative re- dened, 51, 107 ies, 242
search, 178 in experimental research designs, sample forms for, 334f335f
Homeless youth study, as data col- 107108 in questionnaire development,
lection example, 451b in manual data analysis, 524t, 487, 488f
Homogeneity, internal consistency 524525 Innovation, in research utilization,
and, 437 in multiple regression analysis, 636637, 641
Hong Kong postpartum study, as 567, 569 Insensitivity, gender, 301302
ethnographic example, 206b, in multivariate analysis, SPSS in, Insiders, researcher as, in qualitative
207, 213214 583, 584t research, 187
Horne model, of research utilization, of variance, 574 Insiders perspective, in ethnogra-
634b in questionnaire development, phy, 203204
Hospital surveys, participation in, 480481, 581t research topics and, 72
144145 in summary tables, 601t602t Institutional review board, dened,
Human behavior See also Behavior validity and, 430431 336
positivist assumptions on, 1719 In-depth interviews, dened, 232 purpose of, 336
Human nature, positivist assump- in eld study approach, 232234, Instrument decay, as threat to inter-
tions on, 18 233b234b nal validity, 110
Human rights, critique of, criteria for, ethnographic, 232 Instrument development, for data
615b, 618b example of, 232, 233b234b, 235 measurement, 466469, 468b
examples of, 623, 628629 questions in, 232 in factor analysis, 578
Hunting for data, bias and, 307308 Indexes, in data collection, 452458 Instrument validation, in factor analy-
ethics and, 347 item analysis in, 453454, 454t sis, 578
Husserl, philosophical ideas of, 189 item selection in, 454455 Instrumental utilization, dened,
Hypotheses, alternative, 391392 Likert-based, 456b, 456458 633
bias and, 317318 questionnaires in, 480, 492t Interaction effects, in multivariate
in causal model, 8586 rationale for using several items analysis of variance, 574b, 577
conceptual, 25 in, 455 Internal consistency, dened, 437
INDEX 741
in quantitative research, 437438, Item analysis, in data collection, in- Likert-based indexes, in data collec-
439t440t dexes in, 453454, 454t tion, 456458
SPSS procedure for, 704, 705b, Item selection, in data collection, in- advantages and disadvantages
706707 dexes in, 454455 of, 457458
tests of, 436t Item-total correlations, internal con- construction of, 456457
Internal validity, dened, 106, 428 sistency and, 437 evaluation of, 457458
in research designs, threats to, example of, 456b
109110 Jackson and Clairmont, study of, 163, Likert-type questions, dened, 479
International adoption study, as ex- 163b in questionnaires, 478480, 479b
ample of phenomenological Jacksons rule of thirds, in causal example of, 711716
perspective, 192, 194, 195b model testing, 548, 548t Limitations, in research critique ex-
International Council of Nurses Code, Jerusalem-Toronto girls study, as examples, 626, 630
326 ample of comparative study, Lincoln, critical appraisal criteria of,
Internet, in literature review, 80b82b 157, 158b, 159160, 163 617b
in research utilization, 648 Journals, critiques published by, 620 Denzin and, qualitative research
Interpersonal relationships, research research reports in, 590591 dened by, 180b, 180181
utilization and, 640 qualitative, 607608 Guba and, quality standards of,
as way of knowing, 11 in research utilization, 637638, 218219, 219b
Interpretation of ndings, in discrimi- 648 Linear correlations, in correlational
nant function analysis, Justice, dened, 332 analysis, 382, 383f
571573, 572t573t ethics and, 332333 Linear equation, in descriptive statis-
in factor analysis, 580582, 581t Juvenile detention facility study, as tics, 377, 380f381f, 380t,
in multiple regression analysis, example of ethnographic 380382
569570, 570t study, 209210 calculation of, 377, 378b,
in multivariate analysis of variance, 380f381f, 380t, 380382
577 Kant, categorical imperative of, Linear regression equation, in multi-
as quantitative research stage, 13 325326 ple regression analysis,
in research process, nursing Key character les, in eld studies, 561563, 562f565f
process and, 5, 5t 231 Literacy, as questionnaire prerequi-
Interpretations, in eld notes, 230 Key informants, in ethnographic per- site, 139
Interpretive perspective, in nursing spective, 208 Literary Digest 1936 presidential elec-
science, 2021, 24t Knowledge, feminist perspective and, tion poll, 496, 497b
Intersubjectivity, in phenomenology, 283284 Literature See also Literature review
189 nursing See also Nursing knowl- bias and, publication decisions in,
Intervening variables, as causal edge 308310, 309b
model, 87 varieties of, 812 use of ndings in, 312313
testing and, 546547, 547f, positivist assumptions on, 18 Literature review, bias and, guide-
550552, 551t552t in research design choice, 94 lines to minimize, 317
dened, 51, 87 Knowledge of Growth and Develop- critique of, criteria for, 614, 615b,
Interview schedules, dened, 155 ment Test, 455 618b
Interviewers, selection and training Knowledge-focused triggers, in re- examples of, 621, 627
of, 156 search utilization, 636, 644, dened, 595
Interviewing, oral history, in feminist 645f in naturalistic observational stud-
research, 286 Kuder-Richardson coefcient, inter- ies, 245
Interviews, classication of, 466 nal consistency and, 437 as preliminary research step, 76
in data collection, 466 79
dened, 466 Lambda, in cross-tabulation tables, article summary in, 82, 83t
ethnographic, in-depth interviews 373t, 373374 conducting of, 79, 80b82b
in, 232 Language equivalence, in compara- research design and, 96
focus group, in eld study ap- tive studies, 162163, 163b writing of, 79, 83t, 84, 85b
proach, 234239, 237b238b LaPiere, eld study of, 138 in qualitative research, 182
in-depth, in eld study approach, Lastrucci, science dened by, 17 grounded theory, 201202
232234, 233b234b Latin-American participants, in com- phenomenological, 192
for questionnaires, in survey de- parative study, 160, 161b, 162 in quantitative research, 13
signs, 154, 156 Lazarsfeld, survey of, 136137 reports in, 595596
worksheet for estimating re- Leadership, research utilization and, in research utilization, 636
sponse rate in, 152b 640 Literature sampling, in data analysis,
structured, 154 Legitimacy, in participant observa- in grounded theory perspec-
Introduction, in research report, 595 tion studies, 227228 tive, 201
Intuition, dened, 10 of questionnaires, 142 Lived experiences, in phenomenolog-
in phenomenology, 189 Leiningers criteria, in research re- ical perspective, 188
as source of nursing knowledge, 10 ports, critical appraisal of, Log function variables, in multiple re-
Investigator bias, in bereavement 617b gression analysis, 569
study, 313, 314b Leptokurtic distribution, dened, 356 Logical positivism, in nursing sci-
Investigators, in research design, de- in descriptive statistics, 356 ence, 17
scription of, 94 Lewin, action research described by, Longitudinal study See also Panel
Invitation, to participate in research 261 studies
study, ethics and, 335336, planned change model of, 642 dened, 94, 129
342343 Liberalism study, as factor analysis
in questionnaire development, example, 580582, 581t MacIntyre, ethics and, 326
486488, 487b, 488f Lifestyle study, in causal model test- Macrovariables, dened, 23
Iowa Model for Research-Based Prac- ing, 545556, 546t, 548t549t, Magnitude estimation procedures, in
tice, of research utilization, 551t552t, 554t555t data collection, 458461
633, 634b, 644646, 645f Lifeworld, in phenomenology, 189 evaluation of, 460461
742 INDEX
Magnitude estimation procedures, in Median, in central tendency, 358, Moving phase, of planned change,
data collectionContinued 358t 642, 646b
sample instructions for, 460b dened, 358 Multicollinearity, dened, 567
tips for using, 459460 Memoing, in eld studies, computers in multiple regression analysis, 567
Mailed questionnaires, response rate and, 249 Multiple index items, rationale for,
in, 147148 Meta-analysis, dened, 136, 164 455
model for estimating, 148, in secondary data analysis, Multiple realities, in phenomenology,
149b150b 164165 189
researcher-controllable factors Methodological triangulation, bene- in qualitative research designs, 186
in, 148 ts of, 31 Multiple regression analysis, 560570
researcher-noncontrollable fac- challenges of, 31 dened, 560
tors in, 148149, 150151 dened, 2930 discriminant function analysis ver-
tips for increasing, 151152 perspectives on, 3031 sus, 571, 572t
worksheet for estimating, 152b Methodology, development of, as dummy variables in, 565567, 566t
in survey designs, 147148, quantitative research stage, 13 linear regression equation in,
149b150b, 150152, 152b interpretive, 2021 561563, 562f565f
Mailed returns, in questionnaire sur- nursing perspectives in, 1723, 24t multivariate analysis of variance
veys, 145 critical, 2223 versus, 576b, 576577
Main effects, in multivariate analysis interpretive, 2021 ordinal variables in, 564565
of variance, 574b, 577 positivist, 1720 presence-absence questions in, 565
Man problem, in sexist language, research design versus, 92 presenting and interpreting results
594, 595t in research process, nursing of, 569570, 570t
MANOVA See Multivariate analysis of process and, 4 rationale for, 560561
variance in research report, 596598 strength of association in, 563564
Manual data analysis, 522f523f, Microsoft Project, as resource, 100 tips for using, 567569
522526, 524t Microsoft Windows, SPSS for, Multistage area sampling technique,
Maps, conceptual, 5455 527542 503504, 505b
Marker variable, in factor analysis, Microvariables, dened, 23 Multivariate analysis, techniques in,
579580 Middle-range theories, dened, 47 559584 See also specic tech-
Marx, critical perspective and, 22 examples of, 55b niques
surveys of, 137 testing of, 6062, 62b analysis of variance as, 573577,
Master eld le, in eld studies, in theory continuum, 46f 574b576b
231 Mill, Method of Difference of, 106 comparisons among, 583,
Master table, data analysis and, 523, utilitarian view of, 324 583t584t
523f Mini ethnography, dened, 204 discriminant function as,
in naturalistic observational stud- Missing evidence, as explanation 570573, 572t573t
ies, 246247, 247t aw, 4243 factor, 577583, 581t
Masters degree, in nursing research, Missing value codes, in data entry, multiple regression as, 560570
35 533, 535536 SPSS in, 540t, 583, 583t584t
Matrix, correlation, in factor analysis, Mixed-methodology design, in re- Multivariate analysis of variance,
579 search studies, 31 573577
Maturation, as threat to internal va- Mode, in central tendency, 357t, multiple regression analysis versus,
lidity, 109 357358 576b, 576577
Maxi ethnography, dened, 204 dened, 357 presenting and interpreting nd-
Mead, symbolic interactionism and, Models, antecedent variable, 88, 89f ings in, 577
196 candidate variable, 89, 90f rationale for, 573574, 574b575b,
Mean, in causal model testing, causal, 8490, 86f91f See also 576
550552, 551t552t, 554t555t, Causal models Multivariate models, as causal mod-
554555 conceptual, as grand theories, 47 els, 8990, 90f91f
dened, 358 intervening variable, 8788, 88f dened, 89
in descriptive statistics, central mechanistic, 14 Mutual exclusivity, dened, 475
tendency in, 358359, 359t multivariate, 8990, 90f91f in questionnaires, 475, 475b
comparisons of, 375377, path, 8990, 91f Myths, reality versus, in nursing re-
376t377t in research reports, evaluation in, search, 318320, 319t
MEANS, in causal model testing, 600, 603604
550552, 551t552t, 554t555t, of research utilization, 633, 634b Naming, in research process, nursing
554555 example of, 644646, 645f process and, 4
in multiple regression analysis, source of spuriousness, 88f, 8889 National Institute of Nursing Re-
569 specication of, in preliminary research, funding bias and, 311
Measurement, dened, 424 search stages, 8490, 86f91f National Institutes of Health, nonsex-
in research reports, 597598 unilineal, of comparative studies, ist guidelines of, 300
Measurement error, dened, 440 157 Natural context, in qualitative re-
Measurement levels, in descriptive Moral component, in nursing knowl- search, 187
statistics, 352353 edge, 11 Natural settings, in research design
Measurement of objectives, in evalu- Moral philosophy, ethics as, 323 choice, 9495
ation research, 273 Morals, ethics and, 323324 Naturalistic observational studies,
Measures of dispersion, in descrip- Mortality, as confounding factor, 112 dened, 242
tive statistics, 359360, 360t Mother-child preadoption study, as ethics and, 346
range in, 359 example of phenomenological in eld study approach, 241248
standard deviation in, 359360, perspective, 192, 194, 195b examples of, 242, 243b244b
361t Motherhood and self-denition rationale for, 242
variance in, 360, 361t study, as example of feminist steps in, 242, 245247, 246f247f,
Mechanistic model, dened, 14 research, 287289, 288b 248t
INDEX 743
Nature, positivist assumptions on, 18 der normal curve in, 366370, questionnaire development in,
NCLEX-RN success, in literature re- 368t369t 471494
view example, 85b Z scores in, 365366, 367t sampling in, 495517
Needs assessment, dened, 271 Normative bias, in bereavement myth versus reality in, 318320, 319t
in evaluation research, 271272 study, 314b nursing process and, 46, 5t
Neurosyphilis, Tuskegee Study of, NUD*IST software program, dened, overview of, 1101
331b332b, 337339 1617 philosophy in, 1213
Neutralizing gender terms, 594595, in eld studies, 249251 points of agreement in, 6566
595t data analysis with, 251, 252t as professional mandate, 3436
New Brunswick study, language hierarchical tree diagram quantitative data analysis in,
equivalence and, 163, 163b adapted from, 250251, 251f 519584
Newspapers, for literature review, overview of, 249250, 250b causal model testing in, 544558
80b82b Null hypothesis, dened, 391 initial considerations in, 521543
New-wave research, 263293 See also in tests of statistical signicance, multivariate techniques in,
specic types 391392, 392b 559584
action research in, 264269, Nuremberg Code, 326, 328, 329b reporting and critiquing ndings
270b271b Nurse practitioner model, as qualita- in, 587649
advantages and limitations of, 289, tive research example, 28b research critique in, 611631
290t Nurse-client relationship, Peplaus research report in, 589609
dened, 261 theory and, 56, 57b, 58f, 59b research utilization in, 632649
design of, 15 Nurses, research utilization by, atti- research designs and approaches
evaluation research in, 269, tudes toward, 638640, 639t in, 103293
271275, 276b277b strategies in practice, 643 experimental and quasi-experi-
feminist research in, 281288, 284b, Nurses Network on Violence against mental, 105134
288b Women International, as re- eld study, 224261
health promotion research in, search example, 16b new-wave, 263293
275281, 279b Nursing, art of, as way of knowing, 11 qualitative, 177222
qualitative versus quantitative in, empowerment in, as nursing re- survey, 135175
291, 291t search goal, 8 socially responsible, rules for,
Nodes, in NUD*IST software pro- ethics in, 322348 See also Ethics 346348
gram, 250251 knowledge in, 1217 statistics in, 349419
Nominal measurement, dened, 353 process of, 46, 5t descriptive, 351388
in descriptive statistics, 352353 recognition as science of, as nurs- inferential, 389419
mode and, 357 ing research goal, 8 theoretical frameworks in, exam-
examples of, 443444, 444t research utilization in, 632649 See ples of, 55b, 5558, 56f, 57b,
Nominal variables, in data analysis, also Research utilization 58f, 59b, 60f, 61b
SPSS in, 539540, 540t science of See also Nursing science types of, 2334, 24t25t, 26b, 27b
dummy variables used for, in multi- empirics in, 1112 value of, 4, 7
ple regression analysis, Nursing care, quality improvement Nursing science, critical perspective
565567, 566t in, as nursing research goal, 8 in, 2223
sample size determination and, Nursing Child Assessment Satellite explanations in, 4041
511512 Training model, of research interpretive perspective in, 2021
in summary tables, 599t utilization, 634b, 643 points of agreement in, 6566
Nonequivalent control group design, Nursing knowledge, research in, positivist perspective in, 1720
dened, 127 1217
example of, 128b129b computers and, 1517 Obesity interventions, as between-
as quasi-experimental research de- design and, 1415, 16b subjects study example, 121b
sign, 127 philosophy and, 1213 Objective identication, in evaluation
Nonexperimental research, tests of stages and, 1314 research, 273
signicance and, 418 varieties of, 812 Objective observations, quantitative
Nonmalecence, ethics and, 330, sources of, 810 evidence and, 41
332 ways of knowing in, 1012 Objective phenomena, positivist as-
Nonparallel construction problem, in Nursing process, research process sumptions on, 18
sexist language, 594595, and, 46, 5t Observability, of innovation, in re-
595t Nursing research See also Nursing; search utilization, 641
Nonparticipation, tests of signi- Research; specic entities Observational measurement, in data
cance and, 417418 bias in, 297348 See also Bias collection, 465466
Nonprobability sampling techniques, computers in, 1517 Observations, quantitative evidence
504508 conceptualizing in, 68, 7b and, 41
advantages and disadvantages of, dened, 7 Observed frequencies, in chi-square
507508 designs in, 1415, 16b test, 396397
convenience, 506 ethical issues in, 322348 See also Occupational rating scales, validity
dened, 504 Ethics of, 430432
expert, 507 goals of, 8 Ofce of Research on Womens
purposive, 507 initial considerations in, deciding Health, nonsexist guidelines
quota, 506 what to study in, 68101 of, 300
in sampling overview, 499t introduction to, 336 One-tailed test, dened, 393
snowball, 506507 understanding theory in, 3766 of statistical signicance, 393394
Nonsexist language, 594595, 595t introduction to, 336 Online Journal of Knowledge Synthesis
Nontherapeutic research, in Helsinki knowledge acquired by, 1217, for Nursing, 648
Declaration, 330 16b Open-ended questions, dened, 475
Normal distribution, dened, 355, 364 measurement issues in, 421517 in questionnaires, 475477, 476b
in descriptive statistics, areas un- data collection in, 423469 precoded versus, 477
744 INDEX
Openness, research utilization and, establishing rapport as, 229 Personal contact, with questionnaire
640 eld note creation as, 229231 respondents, 145
Operational level, in quantitative re- observation analysis as, 231 Personal experience, as source of
search, 26 report writing as, 232 nursing knowledge, 910
of research project, 425, 425f Participants See also Respondents Personal knowledge, as way of know-
Operationalization, dened, 26, 426 dened, 184 ing, 11
research project level and, 426 ethics and, assessments in, Personal preference, in research de-
Oral history interviewing, dened, 324326 sign choice, 95
286 codes for, 326, 327b329b, 328, Phenomenological perspective, in
in feminist research, 286 329b, 330, 331b332b, 332 qualitative research, 188194
Ordering of questions, in question- 333 application of, 189192, 193f,
naire development, 486489, protection in, 333336, 334f335f 194b
487b, 488f rules for, 342346 example of, 192, 194, 195b
Ordinal measurement, dened, 353 in feminist research, data analysis other approaches versus, 188t
in descriptive statistics, 352353 and, 287 philosophy of, 188189
median and, 358 researchers and, 284285 Phenomenological Study of Nurses Car-
examples of, 444445, 445t permission from, in secondary ing for Dying Patients, critique
Ordinal variables, in data analysis, data analysis, 170, 171f of, 626630
SPSS in, 539540, 540t in qualitative research, credibility Phenomenology, dened, 189
in multiple regression analysis, and, 433 in feminist research, 287
564565 points of view of, 186187 Philosophy, of feminist research,
in summary tables, 599t researchers as, 187 283284, 284b
Ordinary morality, ethics as, 323324 selection of, in qualitative re- of nursing, common elements in, 13
Orems theory of self-care, 5556, 56f, search, 184185 dened, 12
57b Participation See also Participants; functions of, 12
Outcome prediction, positivism and, Respondents in nursing research, 1213
1819 in questionnaire surveys, pressure in qualitative research, ethno-
Overcrowding, of questionnaire on respondents to, 142 graphic perspective in, 205
items, 492 as voluntary, 146147 grounded theory perspective in,
Overgeneralization, bias and, 301 voluntary, dened, 334335 196
dened, 301 informed consent and, 334335 phenomenological perspective
Overspecicity, bias and, 301 Participatory action research, critical in, 188189
dened, 301 perspective and, 22 Phone surveys, 152154, 155t
dened, 265 advantages and disadvantages of,
Pain, study participation and, ethics examples of, 16b, 269, 270b271b 152153
of, 343 Pascal, experiment of, 106 rules for, 154, 155t
Pairwise solution, in multiple regres- Past tense, in results presentations, sample creation in, 153154
sion analysis, 568569 598 sampling and, 504, 505b
Panel studies, dened, 127 Path model, as causal model, 8990, Physical activity in children study, as
example of, 130b 91f example of health promotion
as quasi-experimental research de- dened, 89 research, 278, 279b
sign, 127131 Path model of initiation pathways, ex- Physiological measures, in data col-
rationale for, 129 ample of, 63f lection, 463465
respondent monitoring in, 129131 Patriarchy, dened, 283 Pilot studies, data collection in, 455
Paradigms, in ethnographic perspec- feminist research and, 283 in questionnaire development, 486
tive, 211, 211t Patterns of incidents, research topics in secondary data analysis,
simultaneity, 29 and, 71 169170
in qualitative research, 179 Payment, of interviewers, 156 Pilot test index, in data collection,
totality, 29 of questionnaire respondents, 142, 454455
in qualitative research, 179 148, 151 Pilot tests, in instrument develop-
Parallel gender terminology, 594595, response rate and, 148, 151 ment, 467468
595t Pediatric intravascular study, as re- Plagiarism, dened, 592
Paraphrased material, in research research utilization example, means of avoiding, 592594
ports, 593 644646, 645f in long quotations, 593
Parse and associates, qualitative re- Peoples Choice, in history of surveys, in paraphrased material, 593
search dened by, 179 136137 in references, 593594
Part correlation coefcient, in multi- Peplaus interpersonal relations the- in secondary source citations, 593
ple regression analysis, ory, 56, 58f, 59b in short quotations, 592593
569570 Percentage change, calculation of, Planned behavior theory, conceptual
Partial corrections, in causal model 362, 362t map of, 60f
testing, 551552, 552t in descriptive statistics, 362, 362t dened, 5657
PARTIAL CORRELATION, in causal Percentages, calculation of, 360361 example of, 61b
model testing, 551552, 552t in manual data analysis, 524525 Planned change, dened, 642
Partial correlations, dened, 385 in descriptive statistics, 360361, in research utilization, 642643
in descriptive statistics, 385386 362t example of, 646b, 646647
Participant observation studies, de- Periodicals See also Journals Havelocks theory in, 644, 646,
ned, 225 for literature review, 80b82b 646b
ethics and, 345 Permission requirements, in natural- steps in, 637, 638f
in eld study approach, 225232 istic observational studies, 242 Planning, in action research, 266
history of, 226 in participant observation studies, Platykurtic distribution, dened, 356
rationale for, 226 228229 in descriptive statistics, 356
steps in conducting, 227232 in secondary data analysis, 168, Plotting of data, in descriptive statis-
entry into group as, 227229 170, 171f tics, 383f, 385
INDEX 745
Policy making, bias and, 316317 advantages and limitations of, 131, example of, 620621
health promotion research and, 133t, 174t dened, 75
277278 dened, 108 Problem-focused triggers, in research
unethical research studies and, 337 exposed or comparison group, utilization, 636, 644, 645f
Popular literature, feminist research 110112, 111f Procedures, critique of, criteria for,
as, 287 same group, 108110, 109f, 110b 614, 615b616b, 618b619b
Population, dened, 496 threats to, 109110 examples of, 623624, 629
in survey designs, 137 Preeld phase, of ethnographic re- Process, nursing, 46, 5t
in sampling, 496 search, 214b research, 46, 5t
Positivist perspective, assumptions Pregnancy adaptation study, as ex- Process consent, ethics and, 336
in, 1719 ample of comparative study, Professional bias, in bereavement
criticisms of, 1920 160, 161b study, 314b
in nursing science, 1720, 24t Presence-absence questions, dened, Professional conduct codes, ethics
qualitative research versus, 178 477 as, 323324
research designs and, 19 in multiple regression analysis, 565 Professional mandate, research as,
values and, 19 in questionnaires, 477b, 477478 3435
Posteld phase, of ethnographic re- Presentation of questions, in ques- Professionalism, as nursing research
search, 214b tionnaire development, goal, 8
Postpartum practices in Chinese 491492 questionnaire response rate and,
women study, as ethnographic Presentation of results See also Inter- 148, 151
example, 206b, 207, 213214 pretation of ndings value of research utilization and,
Posttest research design, 108110, in research reports, 598599 635b
109f See also Between-subjects Presidential 1936 election poll, 496, Project initiation checklist, 98, 99t
research design 497b See also Research project
statistical regression in, 110, 110b Pressure to participate, ethics and, Pronoun problem, in sexist language,
Power, theory and, 45 335336, 342343 594
Power analysis, dened, 512 Prestige indexes, as chart example, Proportionate reduction in error, in
sample size determination and, 600, 605f cross-tabulation tables, 373
512515 socioeconomic status measured Proportions, calculation of, 360
alpha in, 512 by, validity and, 430432 in descriptive statistics, 360
beta in, 513 Pretest index, in data collection, 454 Propositions, dened, 53
effect size in, 513t514t, 513515 Pretest research design, 108110, Prospective study, dened, 94
Power balances, in action research, 109f See also Between-subjects Provincialism, as explanation aw, 42
challenges of, 268 research design Proxemics, dened, 240
Practical considerations, in research statistical regression in, 110, 110b in eld study approach, 240
design choice, 95 Pretesting, in instrument develop- naturalistic observational stud-
Practice, evidence-based, 4 ment, 467468 ies in, 242
theory and, 46 of questionnaires, 486 Psychiatric nursing, discriminant
Practice theories, dened, 4748 Primary nursing study, as evaluation function analysis in, 571
examples of, 55b research example, 275, 276b Psychosocial differences, as topic of
testing of, 62 Primary sampling units, dened, 503 comparative study, 160, 161b
research topics and, 73 Principal components method, in fac- Public Citizen, AIDS research criti-
in theory continuum, 46f tor analysis, 579580 cized by, 340341
Pragmatists, in methodological trian- Principal factors method, in factor Public health, discriminant function
gulation, 30 analysis, 580 analysis in, 571
Precise communications, quantita- Probabilistic explanation, dened, in health promotion research,
tive evidence and, 41 3940 other methods versus, 280
Precision, in measurement levels, in Probability levels, dened, 395 Public Health Service, Tuskegee
questionnaires, 473474 in tests of statistical signicance, Syphilis Study of, 331b332b,
reduction of, 450 395 337339
in questionnaire wording, 485486, Probability sampling techniques, Public policy liberalism study, as fac-
486b 499504 tor analysis example, 580582,
sample size determination and, 510 dened, 499 581t
Precision matching, in between-sub- multistage area sample, 503504, Publication, bias and, 308310, 309b
jects research design, 116117 505b instrument development and,
dened, 116 in sampling overview, 499t 468469
Precoded single-choice questions, simple random, 500501 research report style and, 590591
dened, 474 stratied sample, 502, 503f Published databases, as solution to
in questionnaires, 474b475b, systematic sample, 501502 research problem, 9798
474475 Problem, research See Research Pure research, allied versus, 3233
Predictability limits, in experiments, problem dened, 33
132 Problem identication, in research Purists, in methodological triangula-
Prediction, in correlational analysis, process, nursing process and, tion, 30
381382 4, 5t Purpose, in research design choice, 94
of outcomes, positivism and, 1819 quantitative, 13 statements of, critique of, 620,
in research process, nursing unanswerable through research, 74 626627
process and, 4 Problem selection, bias and, 302 as research preliminary, 76,
Predictive research, dened, 32 303 77b78b
descriptive versus, 3132 Problem solving, as universal theme, Purposeful sample, in qualitative re-
Predictive validity, dened, 428 in ethnographic perspective, search, 184
test for, 429t 213 Purposive sample, in qualitative re-
Pre-experimental research designs, Problem statement, critique of, crite- search, 184
108112 ria for, 614, 615b, 618b Purposive sampling, 507
746 INDEX
Qualitative data analysis, in eld Quality, of data sets, in secondary summary tables in, 599t,
study approach, 248256 data analysis, 169f170f 599600, 601t603t, 604f
computers in, 248253, 250b, in research reports, critical ap- Quantitative sampling techniques,
251f, 252t praisal of, 616b, 626 summary of, 499t
content analysis in, 253256, Quality control, in phone surveys, 154 Quasi-experimental research designs,
254b of questionnaires, 144 105, 126134
Qualitative data sets, in secondary Quality improvement, in nursing advantages and limitations of, 131,
data analysis, 168171, care, as nursing research goal, 133t, 174t
169f171f 8 dened, 14, 126
Qualitative evidence, nursing science Quantitative data analysis, 519584 nonequivalent control group de-
explanations and, 41 computers in, 526542 sign in, 127, 128b129b
Qualitative research See also Qualita- analysis techniques with, panel studies in, 127, 129131, 130b
tive research designs 539540, 540t Questionnaires, Adolescent Lifestyle,
bias in, 299 data cleaning with, 537539 437438, 439t440t, 467469,
as bricolage, 180, 180b data entry with, 531533 468b
computers in, 1517 errors and warnings with, in data collection, 466
core activities in, 181185, 183t 540542, 542t data entry and, 531534
data analysis as, 185 le creation with, 534537 dened, 137, 466, 472
data collection as, 184 overview of, 526527 development of, 471494
explicating researchers beliefs SPSS in, 527528, 528f529f, 530, evaluation in, 492t, 492493
as, 182184 534542, 537b, 540t, 542t formatting in, 489491,
literature review as, 182 hand calculations in, 522f523f, 490b491b
participant selection as, 184185 522526, 524t guidelines for, 141142, 144,
saturation as, 185 Quantitative evidence, nursing sci- 472474, 473t
credibility in, 215216, 433 ence explanations and, 4041 introduction to respondents in,
data collection in, procedure for, Quantitative research, bias in, 299 486488, 487b, 488f
451452 data analysis in, 185 See also Quan- ordering of questions in,
denitions in, 5, 179181, 180b titative data analysis 486489, 487b, 488f
example of, 28b techniques in, 519584 presentation in, 491492
positivist perspective and, 20 data collection in, procedure for, steps in, 480486, 481t, 482b,
purpose of, 27 451452 486b
quantitative versus, 2331, 182t, dened, 6 types of questions in, 474b479b,
183f example of, 26b 474480
choice of strategy in, 29 purpose of, 2324 ethics in hidden coding of, 344
methodological triangulation in, qualitative versus, 2331, 182t, 183f frivolous responses to, 147
2930 conceptual level in, 25 group-administered, 145147
in new-wave research, 291, 291t in new-wave research, 291, 291t individually delivered, 144145,
nursing process and, 56, 6t nursing process and, 56, 6t 146t
types of analysis in, 2829 operational level in, 26 interview, 154, 156
reliability and, 438440 theoretical level in, 2425 length considerations in, 472473,
scientic adequacy of, 214219 types of analysis in, 27 473t
contemporary quality standards validity and reliability in, 2627 mailed, 147148, 149b150b,
and, 218219, 219b reliability in, 435438 150152, 152b
traditional quality standards equivalence in, 438 phone, 152154, 155t
and, 215t, 215218, 217b internal consistency in, 437438, student-designed, 710716
validity and, 433434 439t440t summary of survey study with,
Qualitative research designs, 177222 stability in, 435, 437 142, 143b
See also Qualitative research tests of, 436t in survey designs, 141156
advantages and limitations of, reports in See Quantitative re- Questions, in comparative studies,
219221, 221t search reports 157158
characteristics of, 183f, 186188 scientic adequacy of, qualitative in in-depth interviews, 232
comparisons among, 188t versus, 215t measurement error and, 442
dened, 15, 178179 stages of, 1314 in phone surveys, 154
ethnographic perspective in, validity and, 430433 in questionnaires, 474480
203214 See also Ethnographic Quantitative research reports, formulation of, 480486, 486b
perspective 595605 index development and, 480
grounded theory perspective in, conclusion in, 605 Likert-type, 456458, 478480,
194, 196203 See also critique of, criteria for, 613614, 479b
Grounded theory perspective 615b616b numbers of, 472473, 473t
phenomenological perspective in, examples of, 620626 open-ended, 475477, 476b
188194 See also Phenomeno- discussion in, 604605 precoded single-choice,
logical perspective nal checklist for, 605, 606t 474b475b, 474475
Qualitative research reports, 606609 general considerations in, 592 presence-absence, 477478, 478b
components of, 607608 hypotheses and research ques- rank-ordering, 478, 478b
critique of, criteria for, 614, tions in, 596, 597f research See Research question
616617, 617b619b, 619620 introduction in, 595 in secondary data analysis, 169, 170f
examples of, 626630 literature review in, 595596 Quota sampling, 506
general considerations in, 592 methods in, 596598 Quotations, in eld notes, 230
presentations of ndings in, results in, 598604 in research reports, 592593
598599 graphs and charts in, 600,
summary tables in, 600, 603t, 604f 604f605f Random error, bias and, 306
Qualitative sampling techniques, 508 hypothesis evaluation in, 600, dened, 306, 441
summary of, 500t 602604 Random numbers, in between-sub-
INDEX 747
jects research design, 114, in measurement levels, 446447, bias in, 297348 See also Bias
115t116t, 116 448b449b, 450 dened, 6, 7b
dened, 114 Reference lists, in research reports, descriptive, predictive versus, 31
in simple random sampling, 593594 in evidence-based practice, 4
500501 Reference Manager, as resource, 98, explanatory, predictive versus, 31
in systematic sampling, 501 100 Helsinki Declaration denitions of,
Random sampling, measurement er- Reection, in action research, 267 330
ror and, 441442 Reexive critique, in action research, nursing See also Nursing research
Random selection, in qualitative re- 267 goals of, 8
search, 184185 dened, 267 predictive, explanatory versus, 31
Random variables, dened, 108 Reexivity, dened, 285 preliminary considerations in,
in experimental research designs, in feminist research, 285 68101
108 Refreezing phase, of planned change, choice of topics as, 6974,
between-subjects, 119120 642, 646b 70b71b
Randomization, in between-subjects Refusals, sample size and, 515516 hypothesis and analysis speci-
research design, 114, Regression, statistical, 110, 110b cations as, 9092
115t116t, 116 Regression analysis, factor analysis literature review as, 7679,
dened, 114 versus, 582583 80b82b, 82, 83t, 84, 85b
Randomized clinical trials, dened, Regression line, in correlational readiness as, 98, 99t, 100
272 analysis, 380, 380f research design as, 9296
ethics and, 346 Relaxation therapy study, as quanti- setbacks and challenges as,
in evaluation research, 272, 274 tative research example, 26b 9698
Range, dened, 359 Reliability, dened, 435 specifying the model as, 8490,
in descriptive statistics, 359 in instrument development, 468, 86f91f
Rank-ordering questions, dened, 478 468b stating the problem as, 7476,
in questionnaires, 478, 478b measurement error and, 442443, 77b78b
Rapport, in participant observation 443f problem in, 7476
studies, 229 in qualitative research, 438440 process of, 46, 5t
Rates, calculation of, 362363, 363t in quantitative research, 2627, pure, applied versus, 3233
in descriptive statistics, 362363, 435438 qualitative, 2729, 28b See also
363t equivalence in, 438 Qualitative research
Ratio measurement, dened, 353 internal consistency in, 437438, quantitative, 2327, 26b See also
in descriptive statistics, 352353 439t440t Quantitative research
mean in, 358 stability in, 435, 437 survey, 141 See also Survey de-
examples of, 445446, 446t tests of, 436t signs
in multiple regression analysis, 567 research project level and, 435, theory and, 46
in multivariate analysis of variance, 436t, 437440, 439t440t points of agreement on, 6566
573574 tests of, 436t testing and using of, 5865
Ratio variables, in data analysis, SPSS Reliability coefcients, acceptable topic choice in, 6974 See also Re-
in, 539540, 540t levels of, 435 search topics
sample size determination and, Repeated measures analysis of vari- utilization in nursing of, 632649
509b, 509511 ance, in F test, 416 See also Research utilization
in summary tables, 599t Replacement of terms, dened, 62 value to nurses of, 4
Ratios, calculation of, 363364 Replication, research topics and, 74 values in, critical perspective on,
in descriptive statistics, 363364, of research utilization, 636, 641, 2223
364t 643644 interpretive perspective on, 21
Raw data, in multiple regression Report writing, 589609 See also Re- positivism and, 19
analysis, 567 search reports Research assistants, availability of,
in qualitative research reports, 607 in action research, 267268 research problems concern-
Readiness, as preliminary research in eld study approach, computers ing, 97
consideration, 98, 99t, 100 and, 249 in questionnaire surveys, quality
Reality, in phenomenology, 189 in naturalistic observational control and, 144
Reasoning, dened, 10 studies, 247 Research bias, dened, 298
as source of nursing knowledge, 10 in participant observation stud- Research critic, dened, 612
Reassessment, in action research, ies, 232 Research critique, 611631
267268 NUD*IST software program and, dened, 612
Recognition, of nursing as science, as 252t process of, 612613
nursing research goal, 8 Reporting of ndings, bias and, critic in, 612
Recommendation stage, in research 308311, 309b denitions in, 612
process, nursing process and, guidelines to minimize, 317 stages in, 612613, 613b
5, 5t in feminist research, 287 of research reports, 613620
Record keeping, in qualitative rein qualitative research, ethno- examples of, 620630
search, validity and, 434 graphic, 213 qualitative, 614, 616617,
Recording of data, 236, 256 grounded theory, 202 617b619b, 619620
Red ball/white ball test, as chi-square phenomenological, 192 quantitative, 613614, 615b616b
sample test, 397, 398b399b, to questionnaire respondents, 142 Research designs, choice of, 9596
399t, 399400 Representativeness, sample size and, factors in, 9495
Redesigning the dream, in grounded 516 classic experimental, 112126,
theory perspective example, Research, action, choice of topics 131133 See also Classic exper-
202203, 203b, 218, 218b and, 74 imental research designs
Reduction, in data analysis, in advocacy versus, bias and, 313317 critical perspective and, 23
grounded theory perspective, ethics in, 347 critique of, criteria for, 614, 615b,
200 applied, pure versus, 3233 618b
748 INDEX
Research designs, choice of partipants receiving, ethics and, payment of, 142, 148, 151
Continued 344345 permission and, 147, 487, 488f
examples of, 621622, 627 qualitative, 187, 606609 See also personal contact with, 145
dened, 92 Qualitative research reports pressure to participate on, 142
description of, 9394 guidelines for, 607 reporting of results to, 142
dominant/less-dominant, 3031 sample outlines of, 607609 secrecy regarding study hy-
experimental, 106126 See also Ex- quantitative, 595605 potheses and, 142, 144
perimental research designs conclusion in, 605 Response rate, dened, 498
interpretive perspective and, 21 critique of, 613614, 615b616b, to mailed questionnaires, 147148,
methodology versus, 92 620626 493
mixed-methodology, 31 discussion in, 604605 dened, 147
positivism and, 19 nal checklist for, 605, 606t model for estimating, 148,
pre-experimental, 108112, 131 See hypotheses and research ques- 149b150b
also Pre-experimental research tions in, 596, 597f researcher-controllable factors
designs introduction in, 595 in, 148
preliminary considerations in, literature review in, 595596 researcher-noncontrollable fac-
9296 methodology in, 596598 tors in, 148149, 150151
quasi-experimental, 126131 See presenting results in, 598604, tips for increasing, 151152, 493
also Quasi-experimental re- 599t, 601t603t, 604f605f worksheet for estimating, 152b
search designs Research topics, choice of, 6974 Response to health and illness, as
sample size determination and, in comparative studies, 157, 158b, nursing research goal, 8
508509 159160, 161b Retrospective study, dened, 94
two-phase, 30 dened, 69 in survey designs, 138
types of, 1415 listing of, 70b71b Return boxes, in questionnaire sur-
understanding of, 9296 salience of, response rate and, 148 veys, 145
Research hypothesis, dened, 391 Research utilization, 632649 Review boards, ethics and, 336
in tests of statistical signicance, barriers to, 637641 Rittman et al., critique of study by,
391392 innovation characteristics as, 626630
Research imperative, therapeutic im- 641 Rituals, evidence-based practice ver-
perative versus, ethics and, 332 nurse characteristics as, sus, 633
Research problem, dened, 75 638640, 639t Rogers, planned change model of,
focusing of, 7475 research characteristics as, 642643
limiting of, 7576 640641 Role modeling, interpretive perspec-
statements of, 76 setting characteristics as, 640 tive on, 20
examples of, 77b dened, 633 Role of Critical Care Nurses in Eu-
Research process, data collection in, example of, 644647, 645f, 646b thanasia and Assisted Suicide,
4 future of, 647648 critique of, 620626
nursing process and, 46, 5t importance in nursing of, 633, 635, Roper, polls of, 136
Research project See also Research 635b
initiation checklist for, 98, 99t models of, 633, 634b Safety, study participation and,
levels of, data collection and mea- steps in, 635637, 638f ethics of, 343344
surement and, 424427, strategies to facilitate, 641644 Salience of topic, in phone surveys,
425f426f planned change as, 642643 154
linkages between, 427440, 429t, value of, 633, 635, 635b in questionnaires, response rate
436t, 439t440t Researcher affect, bias and, 302 and, 148
readiness for, 98, 100 Researchers, in feminist research, Same group research design, pre-ex-
stages of, 1314 role of, 284285 perimental, 108110, 109f, 110f
Research question, dened, 75 minimizing of bias by, 316 Sample See also Sampling
example of, 78b in qualitative research, ethno- bias and, selection in, 303
in feminist research, 284285 graphic, 208 size in, 310
formulation of, as quantitative re- grounded theory, 197 cluster, 503504, 505b
search stage, 13 as participants, 187 in comparative studies, 158159
in qualitative research, ethno- personal beliefs of, 182184 convenience, 506
graphic perspective in, 205, phenomenological, 190 creation in phone surveys in,
207t, 207208 research utilization by, 643644 153154
grounded theory perspective in, value of research utilization to, critique of, criteria for, 614, 615b,
196197 635b 618b
in quantitative research, phenome- Research-practice gap, dened, 635 examples of, 622, 627
nological, 189190 Respect, ethics and, 330, 342 dened, 496, 498
reports in, 596, 597f Respondents See also Participants in survey designs, 137
Research reports, 589609 See also monitoring of, in panel studies, expert, 507
Report writing 129131 in feminist research, 285286
bias in, 308316 See also Bias to phone surveys, 154 multistage area, 503504, 505b
critique of, 611631 See also Re- rules for, 154, 155t purposive, 507
search critique in qualitative research, dened, in qualitative research, 184
dened, 590 184 ethnographic, 208209
general orientation of, 590595 identication with, 434 grounded theory, 197
audience in, 590591 to questionnaires, conditioning in, phenomenological, 190191
plagiarism and, 592594 489 quota, 506
report type in, 592, 593b explanations to, 147, 486b, in research reports, 596
sexist language and, 594595, 486488, 488f simple random, 500501
595t imposition and, 472 size of, 508516 See also Sample
style in, 591592 opinions of, 485 size
INDEX 749
snowball, 506507 Scores, factor, in factor analysis, dened, 522

stratied, 502, 503f 579580, 582 Signicance, clinical, in research re-
systematic, 501502 Search and retrieval, in eld studies, ports, 599
Sample interval, dened, 501 computers and, 249 tests of statistical, 391419 See also
Sample size, determination of, Search procedures, in NUD*IST soft- specic tests; Tests of statisti-
508516 ware program, 250b, 251 cal signicance
nominal variable and, 511512 Seat belt study, as example of natu- Signicance levels, dened, 395
overview of, 508509 ralistic observational study, in tests of statistical signicance,
power analysis and, 512515, 242, 245247, 246f247f, 394395, 395f
513t514t 247t248t Simmel, critical perspective and, 22
ratio variable and, 509b, 509511 Secondary data, dened, 136 Simple random sampling technique,
research design and, 508509 in survey designs, 136 500501
Sampling, 495517 See also Sample Secondary data analysis, challenges Simplicity, in phone surveys, 154
accuracy of, 496, 497b, 515 in, 167171, 169f171f of questionnaires, 142, 482b, 485
overview of, 496, 497b data sources in, 167 Simultaneity paradigm, in qualitative
purpose of, 496 examples of, 166b, 171173, research, 179
refusals and, 515516 172b173b in research methodology strategy,
representativeness and, 516 purpose of, 165167 29
sample size and, 508516 in survey designs, 164173 Single-blind study, dened, 124, 305
accuracy and, 515 terms in, 164 Single-choice questions, precoded, in
condence limits and, 515 Secondary source citations, in re- questionnaires, 474b475b,
nominal variable and, 511512 search reports, 593 474475
power analysis and, 512515, Selected evidence, in comparative Situationalists, in methodological tri-
513t514t studies, 163164 angulation, 30
ratio variable and, 509b, 509511 as explanation aw, 43 Situation-producing theory, dened,
techniques in, 498508 Selection, as confounding factor, 112 48
nonprobability, 504, 506508 in qualitative research, 434 Situation-relating theory, dened, 48
probability, 499504, 503f, 505b Self-care, dened, 55 Situation-specic theory, dened, 48
qualitative, 508 Orems theory of, 5556, 56f, 57b Skip interval, dened, 501
summary of qualitative, 500t Self-denition study, as example of Slotted return boxes, in question-
summary of quantitative, 499t feminist research, 287289, naire surveys, 145
terminology of, 496498 288b Smoking See also Cigarette smoking
Sampling distribution, in tests of sta- Semantic differential, in data collec- in candidate variable model exam-
tistical signicance, 392393, tion, 458, 458b ple, 90f
393f394f Semantic relationships, in ethno- in literature review example, 83t
Sampling error, dened, 498 graphic perspective, 210 study in children of, as discrimi-
Sampling fraction, dened, 510 Sensitive questions, questionnaire nant function analysis exam-
sample size determination and, wording and, 485 ple, 572573, 573t
510511 Sensitivity, in action research, chal- study in disadvantaged women of,
Sampling frame, dened, 498 lenges of, 268 as focus group example, 236,
evaluation of, 498 Service provision control, as nursing 237b238b
Saturation, dened, 185 research goal, 8 study of behavior regarding, in
in qualitative research, 185 Set bias, in Likert-based indexes, 458 summary table examples,
Scales, dened, 452 Setbacks, in research, preliminary 601t602t
visual analogue, in data collection, considerations in, 9698 treatment for cessation of, in
461463, 462f Setting, in qualitative research, valid- crossover study example, 120,
Scholarly Inquiry for Nursing Practice, ity and, 434 122f, 122123
critiques published by, 620 in research design, description of, Snowball sampling, 506507
School health, discriminant function 93 Social change, feminist research and,
analysis in, 571 in research utilization, barriers in, 282283
Science, advocacy versus, ethics in, 640 nursing research and, bias in,
347 Sexism, bias and, 300302 315316
bias in, 299 See also Bias androcentricity in, 300301 Social conict, as universal theme, in
critical perspective on, 22 gender insensitivity in, 301302 ethnographic perspective,
interpretive perspective on, 2021 guidelines for minimizing, 316 212
Lastruccis denition of, 17 overgeneralization in, 301 Social control techniques, as univer-
nursing as, empirics in, 1112 dened, 300 sal themes, in ethnographic
positivist assumptions on, 1719 feminist research and, 283 perspective, 212
recognition of nursing as, as nurs- Sexist language, 594595 Social inuence and adolescent atti-
ing research goal, 8 man problem in, 594, 595t tudes, as study example, 129,
Scientic adequacy, of qualitative re- nonparallel construction problem 130b
search, 214219 in, 594595, 595t Social interactionism, in grounded
contemporary quality standards pronoun problem in, 594 theory perspective, 196
in, 218219, 219b Sexual behavior, in examples, of al- Social science, in comparative stud-
quantitative versus, 215t ternative explanation testing, ies, 157
traditional quality standards in, 64, 64f Social situation, in ethnographic per-
215218, 217b218b of axiomatic derivation, 62b spective, 210
Scientic merit, of research utiliza- of path model of initiation path- Social support, as hypothesis speci-
tion, 636 ways, 63f cation example, 91f
Scientic method, dened, 12 Sexual harassment, as research ex- study in homeless youth of, as data
as source of nursing knowledge, ample, 16b collection example, 451b
12 Signaling behavior, in data analysis ex- Socially responsible nursing re-
Scientists for hire, bias and, 312313 ample, 522f523f, 522525, 524t search, rules for, 346348
750 INDEX
Socioeconomic status, bias and, 302 proportions in, 360 Summary tables, in research reports,
rating of, validity and, 430432 rates in, 362363, 363t 599600
Software See also Computers ratios in, 363364, 364t correlations in, 600, 603t
in eld studies, functions of, 249 Stanley and Campbell, study of, cross-tabulation in, 600, 601t
NUD*IST, 249251, 250b, 251b, 109112 mean dependent variables in,
252t Statements, hypotheses as, 53 600, 602t
programs available for, 248 of problem, in preliminary re- qualitative, 600, 603t, 604f
Source acknowledgment, ethics and, search stages, 7476, 77b78b univariate statistics in, 599t,
347 propositions as, 53 599600
Source of spuriousness, as causal of purpose, as research prelimi- Summative evaluation, dened, 272
model, 88f, 8889 nary, 76, 77b78b in evaluation research, 272
testing and, 545, 552556, in theory, 53 Supermarket pricing inequity study,
554t555t Statistical Package for the Social Sci- as example of action research,
Source of spuriousness variables, ences, dened, 17 269, 270b271b
confounding variables versus, Statistical regression, as threat to in- Supportive leadership, research uti-
107 ternal validity, 110, 110b lization and, 640
dened, 52, 88 Statistics, 349419 See also Descrip- Suppressed evidence, as explanation
Spearman correlations, dened, 385 tive statistics; Inferential sta- aw, 43
in descriptive statistics, 385 tistics; specic entities Survey designs, 135175
Specication of model, in preliminary descriptive, 351388 advantages and limitations of,
research stages, 8490, 86f91f dened, 349, 351 173175, 174t, 221t
Split-half method, internal consis- inferential, 389419 comparative studies in, 156164
tency and, 437 dened, 349, 390 challenges in, 160, 162164, 163b
basics of, 526527 Status, in-depth interviews and, in examples of, 157, 158b, 159160,
accessing existing les in, eld study approach, 232 161b
527528, 529f as universal theme, in ethno- dened, 1415, 136
beginning session in, 527, 528f graphic perspective, 212 history of, 136137
point-and-click menu in, 528, Stepwise entry solution, in multiple manual data analysis in, 525526
529f, 530, regression analysis, 568 questionnaires in, 141156
syntax method in, 530, Sterling, reporting of ndings and, general rules for, 142, 143b, 144
bivariate analysis using, 371, 371t 308, 309b group-administered, 145147
creating and saving le in, 534537 Stetler model, of research utilization, individually delivered, 144145,
point-and-click method in, 535f, 634b 146t
535536 Storytelling study, as secondary data interview, 154, 156
syntax method in, 536537, 537f analysis example, 171172, mailed, 147148, 149b150b,
data analysis in, 539540 173b 150152, 152b
3M approach to, 539540, 540f, Stouffer, survey of, 136137 phone, 152154, 155t
540t Strategy, in research methodology, rationale for, 137139
data cleaning in, 537539 29 appropriateness of, 137138
data entry and, 531534 Stratied sampling technique, 502, validity of, 138139, 139b141b
dened, 17 503f secondary data analysis in,
errors and warnings in, 540542, Strength of association, in multiple 164173
542t regression analysis, 563564 challenges in, 167171, 169f171f
multivariate analysis with, compar- Strength of wording, in Likert-type data sources in, 167
isons of, 583t questions, 479480 examples of, 166b, 171173,
procedures for, 583, 584t Strengths, in research critique exam- 172b173b
pilot test index and, 455 ples, 626, 630 purpose of, 165, 167
probability levels using, 395 Structural equation modeling, in test- terms in, 164
MEANS as, 375376, ing of theories, 6263, 63f steps in, 141
in research reports, presentations Structural questions, in in-depth in- Symbolic interactionism, in grounded
of ndings and, 599 terviews, in eld study ap- theory perspective, 196
Spuriousness, source of See Source proach, 232 Syphilis, Tuskegee Study of,
of spuriousness Structured interview, dened, 154 331b332b, 337339
Stability, dened, 435 Student surveys, examples of, 139, Systematic error, bias and, 306307
in quantitative research, 435, 437 140b141b dened, 306307, 441
tests of, 436t Study designs See Research designs Systematic sampling technique,
Staff nurses, attitudes toward re- Study participants See also Partici- 501502
search of, 638640, 639t pants; Respondents
Standard deviation, dened, 359 in research design, description of, t test, between-subjects, 404,
in descriptive statistics, 359360, 93 405b406b, 407t
361t Style, of research reports, 591592 dened, 376, 403
sample size determination and, Subjectivity, feminist perspective overview of, 403404
509b, 510 and, 283284 within-subjects, 404, 406408,
Standard error of the difference, de- Subjects, in qualitative research, 184 408b410b
ned, 404 Substruction, theoretical, research Tables, in manual data analysis, 523f,
in descriptive statistics, means project level and, 425426, 523524, 524t
comparisons in, 376 426f in research reports, style of, 591
Standard error of the means, dened, Suggestions for improvement, in re- summary, 599t, 599600,
392 search critique examples, 626, 601t603t, 604f
Standardizing of data, in descriptive 630 Taking care study, as grounded the-
statistics, 360364 Suicide rates and ratios, statistical ory perspective example, 201,
percentage change in, 362, 362t calculation of, 362364, 202b
percentages in, 360361, 362t 363t364t outline of report of, 608
INDEX 751
Tally sheet, data analysis and, borrowed, in theoretical frame- Total populations, tests of signi-
522523, 523f works, 54 cance and, 417
in naturalistic observational stud- classication of, 46f, 4648 Totality paradigm, in qualitative re-
ies, 245246, 246f concepts and statements in, 4853 search, 179
Taxonomic analysis, of data, in See also Concepts; Statements in research methodology strategy,
ethnographic perspective, dened, 44, 44b 29
210211 explanations and, 3844 See also Touch, therapeutic, studies of,
Taxonomy, in ethnographic perspec- Explanations 123124, 125t, 126, 404,
tive, 210 grand, 46f, 47 405b406b, 406408,
in research process, nursing levels of, 48 408b410b
process and, 4 middle-range, 46f, 47 Tradition, evidence-based practice
Technical competence, research Orems self care, 5556, 56f, 57b versus, 633
problems concerning, 97 Peplaus interpersonal relations, research utilization and, 633, 638
Telephone surveys, 152154, 155t See 56, 58f, 59b as source of nursing knowledge,
also Phone surveys of planned behavior, 5657, 60f, 61b 89
Terminal cancer, effects on care- practice, 46f, 4748 Training, of interviewers, 156
givers of, as study example, purpose of, 4546 Transciptions, in eld studies, com-
128b129b research and, points of agreement puters and, 249
Terminology, in naturalistic observa- on, 6566 Transferability, in qualitative re-
tional studies, 245 practice and, 46 search, dened, 216
Testing, as threat to internal validity, testing and using of, 5865 test for, 217b
109110 selection of, as quantitative re- Transmural care, as study example,
Testing of theories, 5865 search stage, 13 128b129b
alternative explanations in, 6365 situation-specic, 48 Treatment levels, in experimental re-
middle-range, 6062, 62b testing of See Testing of theories search designs, 108
practice, 62 theoretical frameworks in, 5358 Treatment variables, dened, 107
research topics and, 73 understanding of, 3766 in experimental research designs,
structural equation modeling in, Theory building, in eld study ap- 107
6263, 63f proach, computers and, 249 between-subjects, 118120
theoretical triangulation in, 65 Theory development, in factor analy- Trial and error, as source of nursing
Tests of statistical signicance, sis, 578 knowledge, 9
391419 See also specic tests Therapeutic imperative, research im- Trialability, of innovation, in re-
appropriateness of, 416419 perative versus, ethics and, 332 search utilization, 641
chi-square, 396403 See also Chi- Therapeutic playing, in grounded Triangulation, dened, 30
square test theory perspective, 196 in qualitative research, 216
dened, 391 Therapeutic research, in Helsinki De- methodological, 2931
degrees of freedom in, 396 claration, 330 theoretical, in testing of theories, 65
F test as, 408416 See also F test Therapeutic self-care, dened, 55 Triple-blind study, dened, 305
key elements in, 391396 Therapeutic touch, studies of, in True value, dened, 435, 440
in manual data analysis, 525 study comparison example, measurement error and, 440443,
null hypothesis in, 391392, 392b 123124, 125t, 126 443f
one- and two-tailed, 393394 as t-test sample, 404, 405b406b, Trust, research utilization and, 640
probability levels in, 395 406408, 408b410b Truth, positivist assumptions on, 18
research hypothesis in, 391392 Three-variable causal models, Tuskegee Syphilis Study, ethics and,
sampling in, 392393, 393f394f 87f89f, 8789 331b332b, 337339
signicance levels in, 394395, 395f testing of, 545556 bias in, 344
steps in, 396 cross-tabulation in, 545550, condentiality violations in, 342
t test as, 403408 See also t test 546t, 549t harm done in, 343
type I and II errors in, 395396 intervening variables in, 546552, pressure to participate in, 343
Theme analysis, of data, in ethno- 548t549t, 551t552t Two-phase design, in research stud-
graphic perspective, 212213 Jacksons rule of thirds in, 548, ies, 30
Theoretical coding, in data analysis, 548t Two-tailed test, dened, 393
in grounded theory perspec- overview of, 545 of statistical signicance, 393394
tive, 201 sample data in, 545546, 546f, Type I errors, dened, 395
Theoretical frameworks, 5358 546t in tests of statistical signicance,
conceptual maps in, 5455 sources of spuriousness in, 395396
examples in nursing research of, 552556, 553f, 554t555t Type II errors, dened, 395
55b, 5558, 56f, 57b, 58f, 59b, Time, as action research challenge, in tests of statistical signicance,
60f 268269 395396
nursing and non-nursing, 5354 in experimental research designs,
Theoretical level, in quantitative re- 107 Unfreezing phase, of planned change,
search, 2425 needed for questionnaires, 642, 646b
of research project, 425, 425f 472473, 473t Unilineal model, of comparative stud-
Theoretical memos, in data analysis, research problems concerning, ies, 157
in grounded theory perspec- 9697 Unit of analysis, in content analysis,
tive, 201 research utilization and, 640 255
Theoretical substruction, research Time dimension, in research design, United Kingdom Nursing Association,
project level and, 425426, description of, 9394 code of ethics of, 326, 328b
426f Timing, of questionnaire administra- Units of analysis, individuals versus
Theoretical triangulation, in testing tion, 147 aggregations, 34
of theories, 65 Toronto-Jerusalem girls study, as ex- Univariate statistics, summary tables
Theory See also specic theory types ample of comparative study, of, in research reports, 599t,
bias and, 317 157, 158b, 159160, 163 599600
752 INDEX
Universal themes, in ethnographic control, 51f52f, 5153 evaluation of, 462

perspective, 212213 core, 198, 200, 200t Visual estimation, in correlational
Universe, in sampling terminology, in data analysis, secondary, 165 analysis, 381f, 381385,
497 SPSS in, 535, 535f, 537540, 540t, 383f384f
Unwanted intimacy study, causal 542, 542t Voluntary participation, dened,
model of, 597f dependent, 5051 See also Depen- 334335
Unwarranted conclusions, as expla- dent variables informed consent and, 334335
nation aw, 44 in descriptive statistics See also
Utilitarian perspective, dened, 324 specic variables Warning messages, in SPSS, 540542,
ethical decision making and, individual, 354370 542t
324325 relationships between, 370386 Washout period, dened, 122
Utility to nursing, of research utiliza- in discriminant function analysis, in within-subjects research design,
tion, 636 572, 572t 122
dummy, in multiple regression Web sites, for literature review,
Validity, dened, 427428 analysis, 565567, 566t 80b82b
external, 106, 604 See also External in experimental research designs, for questionnaires, 139
validity 107108 Weber, interpretive perspective and,
in instrument development, independent, 51 See also Indepen- 20
467468, 468t dent variables surveys of, 137
internal, 106 See also Internal valid- intervening, 51, 87 Western Interstate Commission for
ity in causal model testing, 546547, Higher Education model, of re-
measurement error and, 442443, 547f, 550552, 551t552t search utilization, 634b, 643
443f in literature review, 84, 85b Western Journal of Nursing Research,
in qualitative research, project marker, in factor analysis, 579 critiques published by, 620
level and, 433434 in multiple regression analysis, Willowbrook Institution study, ethics
in quantitative research, 2627 560567, 566t and, 337, 339340, 343
project level and, 430433 in multivariate analysis, compar- Within-subjects research design, 120,
in questionnaire development, isons among, 583t 122123
evaluation of, 492t SPSS in, 583, 584t between-subjects design versus,
research project level and, of variance, 573574 123124, 125t, 126
427428, 429t, 430434 nominal, sample size determina- crossover design in, 120, 122f,
of survey designs, 138139, tion and, 511512 122123
139b141b in questionnaire development, Hawthorne effect in, 123
tests for, 429t 480482, 481t, 489 Within-subjects t test, 404, 406408,
types of, 428, 429t ratio, sample size determination 408b
Values, bias and, 299 and, 509b, 509511 Women, feminism and, 281289 See
policy in, 316317 in research design, description of, also Feminist research
in research, critical perspective on, 93 Wording, in questionnaire develop-
2223 source of spuriousness, 52 ment, 482b, 483486
interpretive perspective on, 21 as study limitation, 133 evaluation of, 492t
positivism and, 19 treatment, 51 Writing, of literature review, 79, 83t,
true, 435, 440443, 443f Variance, analysis of See Analysis of 84
Vancouver School of Phenomenol- variance of reports See also Report writing;
ogy, 192, 193f, 194b dened, 360 Research reports
Variability, in questionnaires, rein descriptive statistics, 360, 361t in naturalistic observational
sponse placement in, 491, 491b multivariate analysis of See Multi- studies, 247
wording of, 484 variate analysis of variance in participant observation stud-
Variables, antecedent, 89 Veriable observations, quantitative ies, 232
in causal models, 8489, 87f89f evidence and, 41
in concept continuum, 50f Verstehen, Weber and, 20 Z scores, dened, 365366
as concepts, 4953 Visual analogue scales, in data collec- in descriptive statistics, 365366,
conceptual, 25 tion, 461463 367t
conditional, 52 construction of, 461462, 462f Zyban, in crossover study example,
confounding, 53 Delphi technique in, 462463 120, 122f, 122123

Research For Nurses - Methods and Interpretation

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Research For Nurses - Methods and Interpretation

Diunggah oleh

Hak Cipta:

Format Tersedia

Copyright 2002 F.A.

RESEARCH FOR NURSES:

This page intentionally left blank

Angela Gillis, BSN, PhD

Winston Jackson, PhD

F.A. DAVIS COMPANY PHILADELPHIA

lyzing research data and in reading A chapter devoted to the problem of

E. Joyce Black, RN, EdD Mary Taylor Martof, RN, EdD

Dr. Janet Brookman, DSN, RN Martha MacLeod, RN, PhD

Elizabeth S. Carlson, PhD, RNC Sharon Moore, RN, PhD

Karen Fernengel, PhD, FNP, CS Carla Mueller, PhD, RN

Rebecca L. Hartman, EdD, RN Kathleen Poindexter, RN, DNSc

Teresa Shellenbarger, DNSc, MSN, BS Susan Urbanski, MSN, RN

PART 1 GETTING STARTED 1 Predictable Setbacks and Challenges 96

Introduction to Nursing Research 3

Scientific Adequacy of Qualitative PART 4 A STATISTICS PRIMER 349

Fundamental Sampling Techniques 498 Organization of Quantitative Reports 595

Chapter 19 GLOSSARY 699

The Research Report 589 BIBLIOGRAPHY 717

This page intentionally left blank

This page intentionally left blank

A. Research and Nursing: Parallel Processes E. Key Methodological Perspectives in

Aesthetics Explanatory research Positivist perspective

INTRODUCTION TO NURSING RESEARCH 5

INTRODUCTION TO NURSING RESEARCH 7

BOX 1.1 Definitions of Research munity highlights the differences in nursing

3. What Are the Goals of practice empowers the nurse. At the

INTRODUCTION TO NURSING RESEARCH 9

transfer learning from one situation to a to develop an argument to support their

2. Four Ways of Knowing

INTRODUCTION TO NURSING RESEARCH 11

A seminal piece of work by Carper (1978) c. Personal Knowledge as a Pattern

INTRODUCTION TO NURSING RESEARCH 13

in professional and scientic journals, making related to prophylactic use of con-

INTRODUCTION TO NURSING RESEARCH 15

BOX 1.2 Nurse Researchers at Work

SEXUAL HARASSMENT: EVERYDAY VIOLENCE IN THE LIVES

INTRODUCTION TO NURSING RESEARCH 17

Most people have some familiarity with a. Positivist Assumptions about

their research. Again, following Lastrucci, Positivists are interested in understand-

INTRODUCTION TO NURSING RESEARCH 19

INTRODUCTION TO NURSING RESEARCH 21

underpins practice is consistent with described in detail because there is a de-

3. The Critical Perspective of practice (Stevens and Hall, 1992). The

INTRODUCTION TO NURSING RESEARCH 23

Table 1.2 Summary of Three Key Methodological Perspectives in Nursing Research

INTRODUCTION TO NURSING RESEARCH 25

Table 1.3 Summary of Key Distinctions in Types of Research

BOX 1.3 Nurse Researchers at Work

THE EFFECTS OF RELAXATION THERAPY ON PRETERM

(iii) Operational Level (iv) Linkages between Levels:

INTRODUCTION TO NURSING RESEARCH 27

BOX 1.4 Nurse Researchers at Work

A NURSE PRACTITIONER MODEL OF PRACTICE

INTRODUCTION TO NURSING RESEARCH 29

INTRODUCTION TO NURSING RESEARCH 31

INTRODUCTION TO NURSING RESEARCH 33

4. Units of Analysis: Individuals is easy to compute average scores, but it

INTRODUCTION TO NURSING RESEARCH 35

and expertise. However, as we have tice and usually an in-depth perspec-

A. Explanations D. Theoretical Frameworks

1. Purpose of Theory E. Testing and Using Theory in Research

Alternative False dilemma Selected evidence