A New Almost Unbiased Estimator for Linear Regression Model in Heteroscedastic or Correlated Errors

This paper presents an almost biased estimator with two parameters (k, d) to improve the estimator efficiency for linear regression model that have heteroscedastics or correlated errors and suffer from the problem of multicollinearity. The properties of the new estimator, the variance matrix, and the mean squared error (MSE), are computed in comparison with the other estimators, using the mean squared error criterion as a measure of fit. The new estimator appears to have better properties than the other estimators because its mean squared error is smaller than the others. Simulations, examples, and graphics are used to illustrate the results.