An Examination of Statistical Interactions in the Special Education Literature

Abstract

This study investigated the prevalence in the special education literature of the Type IV error, an error encountered when the finding of a significant interaction in an analysis of variance model is followed by discussion that does not correspond to the correct conceptualization of an interaction. One hundred ninety-three studies were identified that included significant interaction effects based on a two-way ANOVA design. It was found that only 9% of the studies reviewed interpreted the significant interaction consistent with the accurate explanation of a differential or joint effect. These studies did not, however, include a post hoc analysis to determine the cause of the interaction. Seventy-seven percent of the studies with a significant interaction were explained by using the cell means comparison procedure, which does not correspond to the ANOVA model with interaction effects.