On Using the Average Intercorrelation Among Predictor Variables and Eigenvalue Orientation to Choose a Regression Solution

Abstract

The question of the predictive values of the orientation of beta and/or the average absolute intercorrelation among independent variables in guiding an investigator's choice of regression method is interesting and important. The availability of a computer simulation capable of producing data with given R^2 and average absolute intercorrelation
made study of this question possible. The simulation was
designed for the 1979 comparison of shrinkage formuli by Newman, McNeil, Garver, and Seymour.