Using the Glm Procedure to Analyze Multivariate Response in Crossover Designs

The crossover design involves observing the same response variate under different experimental conditions for each subject, and as such, is a special case of a repeated measures or multivariate data structure. The general multivariate linear model provides a natural framework for analysis of such data because generalizations to more complex structures are feasible and follow the same rationale. Response to treatment is often multivariate, especially in clinical studies, although multivariate response has received relatively little attention in the literature on analysis of crossover designs. The GLM procedure provides an easy-to-use approach to the analysis of multivariate response data arising in crossover stodies. We demonstrate the methods and outline the steps for using SAS· to carry out a general class of analyses.