Performances of Ordinary and Generalized Least Squares Estimators on Multiple Linear Regression Models with Heteroscedasticity

Abstract

This paper focuses on the impact of heteroscedasticity on the estimate and variance of model parameters by studying the performances of Ordinary Least Square (OLS) and General Least Square (GLS) estimators in multiple linear regression models with two independent variables only and error term characterized with different magnitudes of heteroscedasticity related to predictors at different sample sizes. This research explored the patterns of the variance estimates offered by the two estimators when data is front with heteroscedasticity. Also the studies explored the situations where concepts of substitute, complementary or joint demands/supply may reflect in the stochastic characterization of the error term. From Monte Carlo simulation studies using R-package the GLS estimator maintains its superiority over the OLS in multiple linear regression models.