Analysis of a within-subject design with covariates

摘要

This paper deals with a method of analyzing incomplete data, forming a monotone pattern, from a within-subject design with baseline measurements. An assumption of multivariate normality with antedependence structure is made for both the baseline measurements and the response measurements. Adjusted means for the period-by-sequence cells and their dispersion matrix are obtained. Large sample tests are proposed for testing various hypotheses. A numerical example is given to illustrate the methodology.