A FAMILY OF IMPROVED ORDINARY RIDGE ESTIMATORS

This paper studies the Mean Squared Error (MSE) properties of .a proposed family of Ordinary Ridge Estimators (OREs) of the coefficients in the linear regression. We make extensive use of G( ) functions to provide both exact and asymptotic approximations to the MSE. Using these results we propose a new set of OREs whose MSE is smaller than that of the Ordinary least squares (OLS) estimator. These improved estimators can be used when faced with the multicollinearity problem. A simulation study is also done to further analyse the MSE of the proposed estimators compared with some of the existing OREs.