Abstract: | A regression grouping technique is described for reducing the number of predictor variables while retaining predictive information. The data used for illustration are principal responses to the Rokeach dogmatism items as predictors, and school bureaucratization as the criterion. The dogmatism-bureaucratization relationship is analyzed by means of this grouping technique. The results are contrasted with those using the more usual techniques of summing and factor analysis, and are shown to lead to very different conclusions. The merits, assumptions, and potential generalizability of this grouping technique are discussed. |