Type IV errors and interactions.

Extends L. A. Marascuilo and J. R. Levin's notion of Type IV errors, emphasizing the interpretation of interactions in factorial analysis of variance (ANOVA) designs. To help clarify what an interaction is and what it is not in terms of the ANOVA model, it is emphasized that (a) comparisons designed to explain an interaction are acceptable only if they reduce to comparisons involving interaction parameters exclusively; and (b) such comparisons may be both specified and directional, i.e., they may be defined to test an E's a priori hypotheses. (17 ref.) (PsycINFO Database Record (c) 2010 APA, all rights reserved)