The Method of Unweighted Means in Multivariate Analysis of Variance

Abstract

The method of unweighted means in the multivariate analysis of variance with unequal sample sizes is investigated. By approximating the distribution of the hypothesis SSCP with a Wishart distribution, multivariate test statistics are derived that are analogous to the usual ones except the eigenvalues and hypothesis degrees of freedom are adjusted in accordance with the discrepancies in sample size. Monte Carlo methods are used to show the approximate test statistics are accurate over a range of conditions. Conditions are given under which the method of unweighted means yields exact results. A numerical example illustrates the technique.