A Proposed Biased Regression Estimator for Treating the Existence of Autocorrelation and Multicollinearity

Abstract

This study aims at suggesting a biased regression estimator for treating both the autocorrelation between errors and the multicollinearity among the explanatory variables together. The properties of the suggested generalized biased (GB) estimator are stated, and its performance is investigated using the mean square error criterion over the generalized least squares estimator, the generalized ridge estimator, and the generalized Liu estimator. Also, the estimators for the two parameters k and d of the suggested GB estimator and the others are selected. A massive study of simulation is performed, and the results indicate that the suggested GB estimator performed well compared to the other generalized estimators under specific conditions. Finally, two real datasets are used to explain these findings.