Empirical comparison of univariate and multivariate analysis of variance procedures

Abstract

Describes a monte carlo study in which univariate and multivariate analysis of variance procedures are used either singly or in combination to form 3 approaches to the analysis of multivariate normal data. Comparisons among the approaches are made with respect to the error rates/comparison, in relation to experiments error rates, and average number of errors in experiments with at least 1 error which were obtained when data generated under 36 combinations of sample size, number of variables, and proportion of variance in common were analysed. The approach consisting of a multivariate analysis of variance followed by univariate analyses of variance is recommended as being the most useful. (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1971 American Psychological Association.