APPLIED MULTIVARIATE STATISTICS: WITH SAS® SOFTWARE

6.4. Statistical Tests for Covariance Structures

In Chapter 5, we encountered the spherical or compound symmetric covariance structures as a required prerequisite for the univariate analysis of variance methods to be valid for the repeated measures data. It was suggested that one should first test if such a covariance structure can be assumed using the likelihood ratio tests. In many real world applications the hypothesis of spherical or compound symmetric covariance structure will be rejected. While the univariate analysis of variance approach is then invalidated, one may still pursue the (univariate-like) analysis using the asymptotic results based on the maximum likelihood theory. However, in such problems, first one should also determine ...