| Abstract: | Used a multivariate discriminant analysis procedure to identify moderator variables in a multipredictable group validation (mpgv) model. Results from mpgv were compared to those of the univariate simple algebraic and absolute difference techniques. 17 variables, determined for 418 undergraduates (assigned equally to an experimental and a cross-validation group), were investigated as potential moderators of the relationship between the predictor and the criterion (gpa). The 3 techniques revealed moderators different from each other. In addition, number and types of moderators varied according to the composition of the prediction groups in mpgv. Results are discussed with respect to problems with moderator variable techniques. (27 ref.) (PsycINFO Database Record (c) 2010 APA, all rights reserved) |
