Application of robust residuals in matrices of consistent covariance

Abstract

The estimation of the parameters, by the linear regression model, is made by the method of ordinary least squares (OLS). This method provides estimates unbiased, consistent and efficient. However, under heteroscedasticity, the OLS estimators become inefficient and the common estimator of their covariance matrix is not consistent. Heteroskedasticity consistent covariance matrix estimator (HCCME) were proposed to solve the heteroscedastic problem in linear regression. In this work, four types of robust residuals were applied in the HC3, HC4, HC4m and HC5 estimators to evaluate their performances. We also present an empirical application that uses real data.