A Comparison between Unbiased Ridge and Least Squares Regression Methods Using Simulation Technique

Abstract

The parameters of the multiple linear regression are estimated using least squares ( B̂_LS ) and unbiased ridge regression methods (B̂(KI,J)). Data was created for fourteen independent variables with four different values of correlation between these variables using Monte Carlo techniques. The above methods were compared using the mean squares error criterion. Results show that the unbiased ridge method is preferable to the least squares method.