Some comments on the homoscedasticity criterion M and the multivariate analysis of variance tests T2, W and R

Abstract

Some synthetic sampling results on the power of Bartlett's homoscedasticity criterion M for the multivariate case are given, together with the effects that such departures from equality would have on the supposed significance levels for the analysis of variance tests Hotelling's T2, Wilks' likelihood ratio criterion W and Roy's largest root R.