Abstract

Using a linear regression model often leads to a problem where the error variances are not equal across all observations. This issue is known as heteroscedasticity, and it can result in inefficient ordinary least squares estimators and biased and inconsistent covariance matrix estimators. Consequently, inferences can be misleading, and test statistics may not follow their expected distributions. While a significant amount of literature exists for addressing these problems with classical approaches, heteroscedasticity's negative impact on hypothesis testing has yet to be resolved with Bayesian methods. This article seeks to address this issue by proposing a popular heteroscedasticity consistent covariance matrix estimator that employs the Bayesian techniques to provide more accurate hypothesis testing inferences and precise interval estimation procedures. The Monte Carlo simulations were conducted to evaluate the proposed technique's performance, and a real-life example was presented to illustrate its application.