Investigating the structure of the error terms and biased estimators in the linear regression models with heteroscedastic and autocorrelated errors

When the assumptions on the errors and explanatory variables in the linear regression model are not satisfied ordinary least squares estimator is not a suitable estimator. In this study heteroskedasticity and autocorrelation which violate these assumptions about the errors are introduced, test are given for determining them and the error structure are examined. Multicollinearity problem see when the explanatory variables are not independent is considered under the heteroskedasticity and autocorrelation. When both heteroscedasticity /autocorrelation and multicollinearity come true together, alternative estimators to generalized least squares estimator are introduced and the performance of these estimators are considered under the mean squared error and the bias criteria. These theoritical results are illustrated in a numerical example and a Monte Carlo simulations study by using Matlab program.