ABSTRACT
The repeated-measures design constitutes the second major building block of experimental design. This chapter presents one method of analyzing data from within-subjects designs. It shows that there are some situations in which the multivariate approach is preferable to any of the mixed-model approaches. The chapter discusses that the restricted model allows an effect due to subjects. It explains testing comparisons, traditional formulas for confidence intervals depend very strongly on the validity of the sphericity assumption. The chapter briefly describes why some intervals are too narrow and others too wide with the pooled variance approach. It emphasizes that the one-way analysis does not model position effects, and thus may always not provide the increased power desired by using a within-subjects design. The chapter introduces some typical situations where a repeated-measures design might arise.
