An empirical investigation on the effects of unequal variance-covariance on Bartlett's homoscedasticity criterion (M) and the multivariate analysis of variance tests derived by Lawley and Hotelling (T²₀), Wilks (W) and Roy (R)