Affiliation:
1. University of Texas at Austin
Abstract
A procedure is described for determining trait-treatment interactions when the effects of potentially confounding variables have been controlled in the original design through the analysis of covariance or the calculation of residual gain scores. The procedure determines the homogeneity of group regressions, the regions of covariable values for which group regressions are significantly different, and the percent of error that may be expected from this procedure when subjects are differentially assigned to treatments. Implications for recasting research designs based on the post hoc analysis of trait-treatment interactions are discussed.