An analysis of repeated measurements on experimental units in a two-way classification

In experiments with repeated measurements made on the same subjects, the repeated observations in time may be correlated. Therefore, the assumption of independent observations cannot be made in general. This thesis considers the experimental design with treatments in a two-way classification with a disproportionate number of subjects allocated to each treatment combination and repeated measurements made on the subjects.

A procedure is shown to be applicable for computing an analysis under somewhat restrictive assumptions. It is assumed that the variances are equal for all times and the correlations in time are equal. The tests obtained are for the three-factor interaction, the two-factor interactions assuming the three-factor interaction zero, and the main effects assuming all interactions zero. The procedure requires the inverse of one matrix, some matrix multiplication, and the calculation of some standard sums of squares.