A STUDY OF TWO NON-LINEAR METHODS OF COMBINING PREDICTOR TESTS

摘要

A moderated multiple regression method employing cross-product variables and a configural procedure were compared with the conventional multiple regression procedure, applied to the same combinations of predictors.nIn the first tryout, a build-up procedure vas incorpor-ated into the experimental design but vas abandoned in the first analysis stage, since none of the moderated multiple correlation coefficients vas significantly higher than its respective linear multiple.nIn the second tryout, the selection of variables was based on the extent to which differentia validity could be obtained by dichotomizing a predictor on the basis of a second predictor.