INTERACTIONS, PARTIAL INTERACTIONS, AND INTERACTION CONTRASTS IN THE ANALYSIS OF VARIANCE

Proposes a 2-step framework for the interpretation of significant 2-treatment interactions. First, a contrast between the means of 1 treatment is estimated separately at each level of the 2nd treatment. A partial interaction test tests the hypothesis that the variability among these contrasts is zero. The 2nd step consists of computing a difference between the separately estimated contrasts. An interaction contrast test tests the hypothesis that this difference is zero. The familywise Type I error rate can be controlled at alpha by employing K. R. Gabriel's (1969) simultaneous test procedure for partial interaction tests and H. Scheffé's (1953) method for interaction contrast tests. (11 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1979 American Psychological Association.