Heteroscedasticity consistent ridge regression estimators in linear regression model

Abstract

The presence of multicollinearity among predictors and heteroscedastic errors has adverse impact on the performance of ordinary least squares estimates and their covariance matrix. This article introduces modified ridge estimator, namely, the heteroscedastic consistent ridge estimator that efficiently controls the adverse effect of heteroscedastic error term. Our suggestion is based on a novel idea of estimating the error variance using the heteroscedasticity consistent covariance matrix estimator. The suggested approach works well in terms of heteroscedastic as well as homoscedastic scenarios too. We have evaluated the performance of suggested modified estimators with the help of simulations and have illustrated it by applications to real-life datasets. This improvement will help to efficiently handle the estimation problem in these two challenging scenarios.