A New Diagnostics Approach for Repeated Measures Designs

The traditional univariate analysis of the repeated measures design is obtained by treating subjects and their associated interactions as random effects. This analysis requires that certain variances and covariances of the dependent variable at various combinations of within-subject factors are equal. Instability of the variance and covariance components may mask significant effects and may compel the researcher to use a less powerful multivariate technique. The number of subjects required in the multivariate approach is a function of the number of dependent variables and levels of the within-subject factors. Thus, the researcher may be forced to use a conservative degree-of-freedom adjustment instead, which may mask significant results if subject availability is limited.