Univariate and Multivariate Groups by Trials Analysis Under Violation of Variance-Covariance and Normality Assumptions.

The conventional Groups x Trials F ratio from an analysis of variance (ANOVA) was compared with three alternatives: an epsilon corrected F ratio, a nonparametric test, and a parametric multivariate test. The Type I error rate and power of the tests was investigated under violation of variance-covariance and normality assumptions. In general, the nonparametric test was preferable under violation of the normality assumption. When both normal-theory conditions were met, the conventional F ratio was preferred, but under severe violation of the variance-covariance assumption, the multivariate test was generally preferred.