The term repeated measurements refers broadly to data in which the response of each

experimental unit or subject is observed on multiple occasions or under multiple conditions.

Although the response variable could itself be either univariate or multivariate, we restrict
consideration to univariate response variables measured at multiple occasions for each
subject.

In many applications, multiple measurements are made on the same experimental units over a
period of time. Such data are called repeated measures. An example is growth curve data such
as daily weights of chicks on different diets. The design for repeated measures could be one
of the standard design, e.g., a completely randomized design or a randomized complete block
design.
For example, three diet treatments are randomly assigned to the chicks according to a
completely randomized design. The experiment units are the chicks and each chick is
observed weekly for some weeks. The treatment factor is diet and is often referred to as the
between-subjects factor. Time is also regarded as a factor and referred to as within-subject
factor. The experimental units are often called subjects.
In repeated measures experiments, interest centers on
1. How treatment means change over time; and
2. How treatment differences change over time, i.e., is there a treatment by time interaction?
These questions arise in any factorial experiment and there is nothing peculiar about the
objectives of a repeated measures experiment. What makes the repeated measures data
analysis distinctive is the covariance structure of the observed data those data from the same
subject may be correlated and the correlation should be modelled if it exists.
The objectives of repeated measures are (1) to determine whether the optimal
therapeutic effect has been reached, (2) to determine whether a dose titration is
necessary for good clinical practice, (3) to monitor the progress and/or health-
related quality of life of patients with chronic diseases such as cancer, or (4) to
study the behavior of the study drug over time (or to detect whether a potential
pattern or trend in time exists).

Definitions
A repeated measures design is one in which at least one of the factors consists of repeated
measurements on the same subjects or experimental units, under different conditions. Such a
factor is commonly called a within subjects, factor. A between-subjects factor is one in
which each level of the factor contains different experimental units. In this paper, we `will
concentrate on experimental designs with both within-subjects and between-subjects factors.

In practice, different statistical models can be applied to address the above objectives
under different model assumptions. Models include the usual analysis of variance models
for assessment of overall average drug effect across time points, for detection of time
(visit) effect, and for determination of treatmentby- time effect and the method of
generalized estimating equations (GEE) proposed by
Zeger and Liang (1986) and Liang and Zeger (1986).

Many methods have been used for the analysis of repeated measures data. The classical
approach is to treat the experimental units in a repeated measures study as blocks in a blocked
design. Multivariate approaches make fewer assumptions than the classical approach, but in
general are less powerful when the assumptions that the classical approach requires are met.
The newest approach uses mixed models, which may not require as stringent assumptions as
the classical approach, and under some circumstances may be more powerful and flexible
than the multivariate approach.

In a repeated factor, the repeated measurements are not simply replicates of each other, but
there is some sort of qualitative or quantitative relationship among the levels of that factor.
A repeated measures experiment differs from one in which the multiple measurements are
simple replicates; e.g., subjects have blood drawn one occasion, and the sample is divided
into several replicate subsamples, each of which is measured.

Statistical Modelling and Analysis

The modelling and analysis of repeated measures are a complex topic. In this section, we
only highlight some models and analyses by looking at some real data sets.
1. The Univariate Analysis of Variance Approach
In general, we will regard subject effects as random effects. In some analyses, the repeated
measures from the same subject are assumed to be independent. If we take this position, we
will have the univariate analysis of variance approach. The corresponding statistical model
for this experiment is

2. Modelling Covariance Structure

As we said before, repeated measures from the same subject are usually dependent. Consider
the alzheimer experiment again. The measurements from the same subject on 5 occasions
might be correlated. In this scenario, the model will be essentially the same but the error
terms ijk for the same subject are correlated. We should model this correlation structure.
There are threecommonly used covariance structures: compound symmetric, autoregression
of order one (AR(1)) and unstructured
3, Modeling Time As a Regression Variable
Consider the study on body weights of chicks on dierent diets. There are four groups, each
on dierent protein diet. Body weights are measured on alternate days. The body weights for
the fourgroups are plotted in Figure 2.
From the plots, we can see the dierences between the groups. In addition, there are
between- chicks dierences within each group. For each chick, the growth curve can be
reasonably modelled as a quadratic function of time. A reasonable model would be

The traditional approach to repeated measurements in clinical trials is to perform an

analysis of variance at the last visit. This method is usually referred to as endpoint
analysis
or last observation carried forward (LOCF) analysis. The result is compared with those
obtained by the analyses of variance performed at each visit. If the results from the
analysis
at each visit are consistent with the endpoint analysis, then an overall conclusion on the
treatment effect is drawn. However, if there is a signifcant difference among visits and/or
the last visit, then a further analysis is performed to determine whether there is a pattern
and/or trend over time

A commonly encountered problem in clinical trials with repeated measures is missing

data. In general, missing data may have two causes. On one hand, subjects may
withdraw
prematurely from the study at a post-treatment time point. These subjects are referred to
as
dropouts. As a result the data scheduled to be collected since the last visit would be
completely
missing.

On the other hand, subjects might be lost to follow-up at a scheduled visit.

However, they may return either on another unscheduled date or on the next scheduled
visit. These subjects may or may not complete the study. Diggle et al. (1994) refer to
those
missing values due to the missed visits as intermittently missing values. Several methods
have been proposed that assume the missing values to occur at random within treatment
groups.

In clinical trials it is suggested that the dropout rates and missing patterns
across all visits be compared between treatments. If there is a discrepancy in dropout
rates or a missing pattern, it is crucial to investigate whether a response-dependence or
treatment-related problem has occurred