Grouping Estimators on Heteroscedastic Data

Abstract

This paper gives numerical comparisons of the efficiency of Ordinary Least Squares (OLS) and Grouping Estimators in simple linear regression. The disturbances are assumed to have unequal variances, and an assumption is made about the form of this heteroscedasticity. It is shown that for some types of heteroscedasticity a Grouping Estimator can be more efficient than Ordinary Least Squares.