Repeated Measures Design

Repeated measures experiments are appropriate when multiple measures of a dependent variable are taken on the same subjects or objects or matched subjects or objects under different conditions or over two or more time periods. Repeated measures cover a broad range of research models, from comparison of two treatments on the same subjects to comparisons of multiple treatments on multiple levels of two or more factors on the same subjects and to assessing the differences in means among several related scales that share the same system of measurement.

The terminology used for repeated measures varies among researchers and statisticians. For example, repeated measures designs are also known as repeated measures analysis of variance (ANOVA), repeated measures factors, or within-subjects designs. However, repeated measures may be performed within a multivariate ANOVA design as well. The logic of repeated measures is based on the calculation of difference scores (contrasts) for comparisons of treatment levels for each subject. For example, if there were no treatment effect, the difference between scores before and after the treatment would be close to zero. The number of contrasts is usually 1 less than the number of levels in a factor. A factor with two levels would produce one difference score whereas a factor with three levels would produce two difference scores: one score for the difference between Level 1 and Level 2, and one score for the difference between Level 2 and Level 3, unless other specific comparisons are requested. Contrasts are pooled together to provide the significance test. The mean of contrast scores and variation around the sample mean difference based on the individual difference scores are used to determine whether there is a significant difference in the population.

Repeated measures designs include a broad class of statistical models with fixed effects and/or random effects. Fixed effects are changes in the means of the dependent variable that can be attributed to the independent variables. For example, differences in mean attitude scores about the environment between individuals who live in a rural location and those in an urban location may be attributed to the fixed effect of location. Location is a fixed factor because the researcher made a conscious decision to include only rural and urban locations, although other locations such as suburban could have been included. Random effects are changes in the variance of the dependent variable that can be attributed to the sampling units. For example, participants in a study are usually considered to be a random factor because the researcher selects a sample from a larger population. If another sample were selected, results would likely be different.

Repeated measures designs differ from the traditional between-groups designs that require a separate sample of subjects for each measurement or treatment condition. Observations in a repeated measures design may be from the same sample or experimental unit from one time to the next or from one condition to the next. In addition, repeated measures may be taken on matched pairs of individuals from different groups. Each individual in one of the groups is matched, on the specific variable being investigated, with a similar individual in another group. Measures may be taken over specified or unspecified time periods. Measures taken in a repeated measures study cannot be considered independent of one another because the measures are taken on the same or matched subjects, and one would expect that such measures would be correlated.

...