ABSTRACT
This chapter lays out the repeated measures model and discusses some initial approaches to analysis. It focuses on a design with repeated measurements over time on subjects. Analyzing single measurements per subject may be rather conservative, but it side-steps problems of correlated measurements over time. Experiments in a wide range of disciplines involve selecting a sample of 'subjects', assigning them to treatment groups and over some period of time taking repeated measurements to observe the response to treatment. Repeated measures designs are essentially split plot designs allowing for correlation within each random effect. Experiments which involve sequentially assigning two or more treatments to the same subject and then measuring some response are known as crossover designs. The chapter considers the situations where a split plot analysis is more or less appropriate for repeated measures designs. The presence of two random errors has the same flavor as for split plots, with an added concern about the covariance structure.
