A Comparative Study of Bayesian and Frequentist Testing for Model Comparison in Nested Model

Abstract

Comparison of the two modelling approaches; frequentist and Bayesian had been carried out by various researchers in the area of estimation of parameters in regression model. Modelers are also interested in comparing different models to know the correct one. However, the use of relevant prior information about the data in the concept of model comparison for Bayesian can help to select the right model. This work facilitates comparison between frequentist and Bayesian methods in a nested model when all the parameters considered are evaluated at zero through a simulation study. The results from both simulation and real life data suggest that Bayesian approach with the use of Savage-Dickey density ratio (SDDR) provide a reasonable decision for the nested model than frequentist approach. It was also demonstrated that the SDDR is a true representation of Bayes factor.