Incorporating Substantive Knowledge Into Regression Via A Bayesian Approach To Modeling

Abstract

A multiple regression example is used to illustrate advantages of a Bayesian approach that incorporates situation-specific substantive information over frequentist and Bayesian approaches that ignore such information. Frequentist and Bayesian analyses of a traditional regression model produced nearly identical results. A Bayesian analysis of a modified model yielded preferred estimates of parameters and quality of prediction.