EFFECTS OF NONSPHERICITY TO THE ANALYSIS OF MULTIVARIATE REPEATED MEASURES DESIGNS

Abstract

The analysis of the data obtained from a multivariate repeated measures design, generally, it is accomplished across through of doubly multivariate model (MDM) analysis, or also by means of multivariate mixed model (MMM) analysis. Based on the examination of the Type I error rates and power of the referred procedures, the present investigation puts of relief, on the one hand, the superiority of the MDM approach on the corresponding MMM, except when the dispersion matrix satisfies multivariate sphericity or the sample size is very small and, on the other hand, the poor functioning some of the correction factors used to construct ajusted MMM tests, as for example suggested it in the routine MANOVA of the popular program SPSS.