Use of an all possible combination solution of certain multiple regression problems

Abstract

The advent of electronic computing facilities has made the computation of an R 2 for all possible combinations of a set of predictors practicable. If there are several subsets of predictors with approximately equal predictive efficiency, the choice among subsets may be based upon such additional considerations as face validity, ease of measurement, and dimensionality. In the analysis reported in this paper, a choice could be made from among 6 subsets of predictors with practically no loss of predictive efficiency and from among 68 subsets if some loss were acceptable. (PsycInfo Database Record (c) 2025 APA, all rights reserved)