Joint Research Conference

June 24-26, 2014

Bayesian Inference of Paired Comparison of Relative Importance of Predictors in Linear Regression

Abstract:

Regression analysis is perhaps the most frequently used statistical tools for the analysis of data in practice. The purpose of determining predictor importance is not model selection, but rather exposing the individual contribution of the predictor in the presence of other predictors within a selected model. The purpose of this article is to expand the current research practice by developing a statistical paired comparison model with different link functions in the Bayesian framework to evaluate the relative importance of each predictor in a multiple regression model. Results from simulation studies demonstrate that the proposed weighted paired comparison model with to wsided power (TSP) link function provides the most effective and reliable measure of the relative importance of predictors.