ROBUST WHITE’S TEST FOR HETEROSCEDASTICITY DETECTION IN LINEAR REGRESSION

Abstract

The existing diagnostic measures for heteroscedasticity incorrectly detect heteroscedasticity in the presence of outlying observations; usually high leverage points (HLPs). The classical White’s Test (WT) is the most commonly used diagnostic method for heteroscedasticity in linear regression. The WT does not depend on either normality or prior knowledge of the source of heteroscedasticity. The shortcoming of WT is that in the presence of HLPs it incorrectly detects heteroscedasticity in a data set. In this paper, a Robust White’s Test (RWT) has been proposed which is capable of detecting heteroscedasticity in the presence of HLPs. The results based on Monte Carlo simulation study and real data examples show that the proposed RWT correctly detect heteroscedasticity in the presence of HLPs