Research in Health Sciences 2013 OHTH 2160 GOALS FOR THIS TOPIC Understand the essential differences between experimental, quasi-experimental and non-experimental designs, in terms of their design characteristics and the types of questions that each type of design can answer. Understand why experimental research is so important and so highly regarded, particularly in the context of intervention research. Understand the four essential defining characteristics of true experiments Understand why quasi-experiments are not true experiments, and what this means in terms of answering research questions. Understand sources of variability in data and the concept of internal validity, in terms of control groups and research design. Understand the logic underlying null hypothesis significance testing. Define and explain reliability and validity. Created by John eece !age " of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences 2asic esearch Design and 3ypothesis 4esting Hypothetical study: Does exposure to violent videos lead to higher levels of aggressive play in young children?
4otal N 5 *& 6n 5 (* in each group7 Experimental group+ 8atch violent videos Control group+ 8atch neutral videos 9ideotaped playing together after exposure to videos and level of aggressive play recorded. :.9.+ 4ype of video material 4wo levels+ 9iolent and neutral D.9.+ ;evel of post-video aggression Created by John eece !age ( of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences Operationalised as+ <umber of aggressive acts recorded in post-video play session. 4rue experimental design+ ". Control group (. andom allocation of sub.ects to groups ). Clearly defined :.9.=s and D.9.=s 1. >xperimenter manipulation of :9 Compare with quasi-experimental study+ ? comparison of aggression in boys and girls after exposure to violent video material. :.9.+ @ex 6male vs. female7 :.9. groups are pre-existing- cannot randomly allocate. Non-experimental: 4he relationship between television viewing hours and the level of aggressive play in children. Don=t designate variables as :.9.s or D.9.s. hat!s so good a"out experimental research? Cause and effect! Intervention research CONSORT #ose a hypothesis+ Created by John eece !age ) of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences Children who watch violent video material will exhibit higher levels of post-viewing aggressive play than children who watch neutral content videos. $alsifia"ility+ @ir Aarl !opper. :t is possible to prove statements%hypotheses "&&B false, but impossible to prove them "&&B true. C?ll crows are blacDE Fou can falsify statements, but not questions. C8hat colour are crowsGE @o whatG Next Step C!llecti!n !" ra# $ata% Descriptive statistics: ?im to capture the essential features of the results in an easily comprehensible form. Hraphical display <umerical descriptives %easure of central tendency C8hat=s the typical or average scoreGE Iean 6%7 %easure of varia"ility C3ow much do the scores vary around the typical or average scoreGE @tandard deviation 6&D7 Jor our data+ Created by John eece !age 1 of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences 9iolent video group+ % 5 '.1', &D 5 (.&& <eutral video group+ % 5 #.((, &D 5 ".$, H&p!thesis testin' an$ statistical si'ni"icance% 8hen we run our experiment, we are testing a hypothesis' Hypothesis+ a statement about what we expected to find- a prediction of the experimental outcome. ? well- educated guess. Hypothesis+ Children who watch violent video material will exhibit higher levels of post-viewing aggressive play than children who watch neutral content videos. CONTROL GRO(P )*P)RI+)NTAL GRO(P 6<eutral videos7 69iolent videos7 (* @ub.ects 6n 5 (*7 (* @ub.ects 6n 5 (*7 Iean level of Iean level of aggressive play 5 #.(( acts aggressive play 5 '.1' acts K one group has a higher mean than the other, but what does that meanG :s that difference large, small, a chance result, or something that=s unli(ely to occur "y chanceG 4hree important questions+ Created by John eece !age * of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences ". :s there a statistically significant difference between the two groupsG (. 8hat do these results this tell us about our hypothesisG Do these results support, or fail to support our hypothesisG ). ?re these sample results an accurate reflection of what=s going on in the populationG 8hen we test a hypothesis statistically, we want to reach some conclusion about what=s going on in our population of interest, and we use our sample data to reach that conclusion. 4his is called estimation or inference. 8hyG Lur results have greater importance and relevance if we can generalise them to the population at large, and not .ust limit them to our sample. 8e want to be able to maDe statements about what=s going on in the world at large. $irst (ey concept+ out there in the population at large there are true population means for these variables 6called parameters7. :f : were able to bring every child in the )orld into my lab and sub.ect them to this experiment, : would Dnow the true population mean, the mean for everybody. Created by John eece !age # of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences 8e can never really (no) what the true population mean is, because we can never actually test a )hole population. :nstead, we have to test a small section of that population, which we call the sample. 2ased on the results of that sample, we maDe an Ceducated guessE 6called an estimate or inference7 about the population mean, and our "est guess about the value of the population mean is the sample mean. egardless of what the true population mean is, there is al)ays the li(elihood that : will get a sample mean that has a different value from the population mean, si,pl& $-e t! chance. 4he difference between the mean of my sample and the corresponding mean in the population is called sa,plin' err!r, and it is a chance event. 8e hope that by selecting our sample in a truly random fashion that we minimise the level of sampling error, but it=s always a threat hanging over us. @maller samples have a greater liDelihood of sampling error than larger samples. @o, to understand hypothesis testing, the first concept to understand is the concept of sampling error, and the fact that sampling error happens by chance. Created by John eece !age , of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences &econd (ey concept+ 8hen we carry out a statistical hypothesis test, we have to maDe a choice, based on our sample results, concerning two competing possibilities. Null hypothesis 5 our samples come from populations where there is no difference between the population means. :f : were able to test the )hole population, : would find that the population means for each of these groups would be the same. <ot only does the null hypothesis state that there is no difference in the population means, it states that an& $i""erence that #e see in !-r sa,ple ,eans. is si,pl& $-e t! chance 6i.e. to sampling error7. *lternative hypothesis 5 :f you could drag the whole population into your lab and test them, you would find a difference between the means. 4wo competing hypotheses the null and the alternative. Lur aim in hypothesis testing is to decide which of the two we will accept, and we have to maDe our decision /ase$ !n the res-lts in !-r sa,ple. Lur aim is to decide whether there is a significant difference between these two groups. Created by John eece !age $ of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences 8e do this by testing the null hypothesis only 6the alternative hypothesis doesn=t come into calculations7, and what we try to do is prove that the null hypothesis is )rong. :n statistical terms, we try to re+ect it. :f we decide to re.ect the null hypothesis, then )e are forced to accept the alternative. 4he logic that we follow to do this is as follows+ ". 8e assume that the null hypothesis is true (. >ven when there=s no true difference in the population 6i.e., null hypothesis true7, we accept that our sample means may hint that there=s some difference, simply due to chance 6sampling error7 ). 8e asD the question+ 3ow liDely is it that : could get my observed sample result +ust "y chance, even if there=s really no difference in the populationG 1. :f the liDelihood of your observed sample means occurring by chance alone is less than *B 6.&* in probability terms7, then we conclude that the null hypothesis is wrong, we re+ect it, and we accept the alternative hypothesis Created by John eece !age ' of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences *. 8e conclude that there is a statistically significant difference between the two groups. 4he value that we base our decision on is called the significance level or p level. :f the p level is less than .&*, then we re.ect the null hypothesis. &ignificance level+ 4he probability that your observed results have occurred by chance. *ll statistical hypothesis tests generate a p level, regardless of the type of research you are doing. 4his logic applies to all significance testing, regardless of the area and regardless of the statistical test being used. t-tests 6independent and related%matched samples7 Used to answer the question+ :s there a significant difference between two meansG Correlation, r Used to answer the question+ :s there a significant relationship between two variables?
2 Created by John eece !age "& of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences Used to answer the question+ :s there a significant relationship between the CcountsE associated with two variables. 0e,!nstratin' 0i""erences in 0esi'n Three h&p!thetical in$i1i$-al $i""erence st-$ies ,' Does alcohol change people!s personality? *& people are recruited for a study. 4hey are randomly allocated to two conditions+ alcohol and placebo control. 4hey complete a standardised personality questionnaire before consuming a beverage. 4hey then consume a large amount of either alcohol or an inert placebo control. 4wo hours later, they complete an alternate form of the personality questionnaire. -' *re )omen more .sensation see(ing/ than men? *& men and *& women are administered a standardised sensation-seeDing scale. 4he scores are compared. 0' hat is the relationship "et)een scores on a standardised paper-and-pencil personality questionnaire and the overt "ehavioural expression of personality characteristics? ? large group of people are administered a standardised personality questionnaire, such as the >ysencD >!M. 4hen behavioural correlates of the personality variables are measured. Jor example, for Created by John eece !age "" of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences extraversion, the number of times each person socialises at night is recorded. Jor neuroticism, the number of times a person cries is recorded. 4he study is investigating the nature of the relationship between standardised measures of personality, and the behavioural expression of those measures. 4he " st study is experimental+ 2h&3 :t has an experimental group and a control group. :t involves random allocation of participants to groups. :t has clearly defined independent and dependent variables. 4he levels of the independent variable are under the control of the investigator. 2h& ha1e a c!ntr!l 'r!-p3 4he control group provides a baseline against which we can assess the impact of the treatment on the experimental group. Control groups should be treated the same as the experimental group in every way but the treatment. :f we don=t have a control group, we can=t isolate the effect of the treatment from the effect of other influences 6sources of error that might compromise the internal validity of the design7. o @ources of error are any other factors that might influence a participant=s score on the dependent variable, apart "r!, the treat,ent% Created by John eece !age "( of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences here Does the 1aria"ility Come $rom in an Experimental Design? 8ithin-group variability 2etween-group variability hat Causes Error in a 2esearch Design? :ndividual differences >xperimental error hy randomly allocate? 2ecause it maximises the probability that other influences 6called sources of error7 are randomly spread among the groups. Jor example, what if everyone in the treatment group was over *& years old, and everyone in the control group was under (*G Lr if everyone in the control group was female and everyone in the experimental group was male. Fou wouldn=t be able to separate out the effect of the treatment from the influence of age or sex. 4he influence of the treatment would be said to be confounded by the systematic error associated with age or sex. andom allocation maximises the probability that this doesn=t happen. 3ndependent varia"le+ Hrouping variable. Under the control of the investigator. Dependent varia"le+ @core, outcome, or thing being measured. Created by John eece !age ") of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences Whats so great a!out e"peri#ental designs$ 4hey are the only design that permits you to maDe cause and effect statements about the relationship between the independent and dependent variables. 9ery important in intervention research. Ho) could this "e study have "een done differently? >ach person could have participated in both phases of the study o epeated measures or within-sub.ects design. !articipants could have been matched on personality characteristics before being allocated to either the experimental or control group o Iatched samples design. 2oth of these designs reduce sources of error, and therefore result in any treatment effect being more noticeable 6error clouds treatment effects4 and is something )e try to reduce7. - nd &tudy Muasi-experimental design. Hroups are based on pre-existing characteristic of the participants. ?ll sex difference studies are quasi- experimental. @ome studies can only be done using a quasi-experimental approach. Cannot randomly allocate, because the participants allocate themselves. <ot as powerful as experimental designs, because you cannot maDe cause and effect relationships. Created by John eece !age "1 of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences ?nalyse results in the same way as for an experimental study. 0 rd &tudy Correlational 6non-experimental7 <o independent and dependent variables. <o causal connection between the two variables can be concluded. C!rrelati!n 4 ca-sati!n% ?nalyse using correlation. I#portant %ethodological Issues Relia/ilit& an$ 5ali$it& 2elia"ility+ Does your measuring instrument generate scores that are consistent, stable and dependable, or is it prone to measurement errorG >xample+ Jaulty Ditchen or bathroom scales 1alidity+ :s your personality scale measuring what it claims to measureG Four scale claims to be measuring extraversion. :s itG Created by John eece !age "* of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc Discipline of Psychology School of Health Sciences 2o-2o doll experiment+ http+%%www.youtube.com%watchGv5FclN2hn1&hU Children see, children do http+%%www.youtube.com%watchGv5#Jf32(cruJU !lacebo story on #& minutes http+%%www.youtube.com%watchGv52D?f$HtUr,M Created by John eece !age "# of "# Created on "$%&'%(&") &*+))+&& Contact details+ phone+ ''(*.,*"(- fax+ ''(*.,)&)- e-mail+ .ohn.reece/rmit.edu.au %var%www%apps%conversion%tmp%scratch0(%()##)1'').doc
RA 9994 (Expanded Senior Citizens Act) Law Implementation and Its Relation Among Business Establishments in Maniki, Kapalong, Davao Del Norte, Philippines
Journal of Applied Psychology Volume 79 Issue 3 1994 (Doi 10.1037/0021-9010.79.3.364) Lee, Cynthia Bobko, Philip - Self-Efficacy Beliefs - Comparison of Five Measures. PDF