FITTING MULTIVARIATE POLYNOMIAL-GROWTH CURVES IN 2-PERIOD CROSSOVER DESIGNS

We discuss the statistical analysis of data from two clinical trials u sing crossover designs. In both studies, response was observed repeate dly over time in each treatment period. The first study involves repea ted measurements of a single response variable whereas the second invo lves bivariate response. Methods are described for fitting polynomial growth curves to achieve data reduction in a two-stage approach to the analysis of crossover designs. Thus, a multivariate parametric analys is frequently can be conducted even when the sample sizes are somewhat small as is the case in many crossover designs. Hypotheses that are u sually of interest in crossover designs can be tested in the second st age of the analysis. Methods for testing the multivariate general line ar hypothesis as a basis for statistical inference in such problems ar e discussed.