ABSTRACT
Multivariate analysis of variance (MANOVA), with more than one DV, is a natural extension of ANOVA and is used in similar situations. MANOVA makes parametric assumptions that are the multivariate analogues of those made for ANOVA: the DVs are from a multivariate normal distribution for each treatment group, and these distributions have a common covariance matrix. Univariate tests on the angle factor are justified by the significant multivariate effect of angle reported in the first multivariate table. The main problem with missing data is that the reason observations are missing may be associated with the DV. Missing observations are sometimes replaced by the mean or median value for the variable, or else the group mean for the variable. This is not an option for categorical data: the nearest equivalent would be to replace the missing value with the most commonly occurring category.
