Recognition of Subsets of Informative Variables in Regression

Abstract

A new approach is proposed to address the subset recognition problem in multiple linear regression, where the objective is to recognize a minimal subset of predictor variables without sacrificing any explanatory power. A parameter stability solution of this approach yields a number of informative subsets. To obtain this solution, new parameter stability criteria are repeatedly used. The subsets generated are compared to ones generated by several standard procedures. The results suggest that the new approach finds subsets that compare favorably against the standard procedures in terms of the generally accepted measure: R2.