ASSESSING THE EFFICIENCY OF CLASSICAL AND BAYESIAN APPROACHES IN ADDRESSING HETEROSCEDASTICITY UNDER KNOWN FUNCTIONAL FORMS.

Abstract

This study provides a robust comparison of the traditional and Hierarchical Bayesian approaches for addressing heteroscedasticity, evaluated under known functional forms where the variance of errors is modeled as a function of exogenous variables. Using simulated data generated through Gibbs Sampling in a Monte Carlo framework, the study examines the performance of hierarchical Bayesian (HB), ordinary least squares (OLS), and generalized least squares (GLS)
approaches across different sample sizes and replications. The findings indicate that the HB demonstrates superior efficiency in addressing heteroscedasticity compared to the traditional approaches, consistently outperforming them across various scenarios. These results underscore the advantage of the HB approach in modeling relationships involving predictor variables and a dependent variable exhibiting heteroscedasticity, offering a robust alternative for researchers and
practitioners.