Best Subset Selection Criteria for Multiple Linear Regression

• In Event: Multiple Linear Regression: The General Linear Model SIG Special Topics Paper Session

Abstract

This study presents comparisons of subset selection criteria used to help determine the "best" regression model in multiple linear regression. No such criteria can replace a researcher's knowledge of theory to help choose useful models, but such criteria may help exploratory research. Relationships among variables may be uncertain a priori, may be more complicated than expected, or may appear different than theory implied. Knowing how well various subset selection methods perform under varying conditions can be useful in such cases. Monte Carlo simulations were performed to compare a number of well-known criteria (e.g., AIC, BIC, PRESS, adjusted R2) to some less well-known criteria (e.g., AICc, AICu, GCV, RIC, cross-validity R2).

Authors

• Gordon P. Brooks, Ohio University - Athens

• Pornchanok - Ruengvirayudh, Ohio University - Athens