ABSTRACT
This chapter focuses on the limitations of valid statistical hypothesis tests and the testing of the hypotheses other than the traditional omnibus null hypothesis. It is important to recognize that the omnibus hypothesis tests of the traditional mixed-model ANOVA involves pooling multiple sources of variance in order to estimate the appropriate mean square error terms for each of the hypothesis tests. The primary issue for post hoc comparisons in repeated-measures designs is the requirement that the multiple comparison method choose error terms consistent with the error term selected for testing the omnibus null hypothesis. The usual post hoc analysis of multiple comparisons involves calculating a test statistic by a weighted linear combination of cell means and dividing that contrast by an adjusted Mean Square Error, where the adjustment is designed to protect the Type I error rate against multiple hypothesis tests.
