PERFORMANCE OF A WEIGHTED RIDGE ESTIMATOR

Ridge regression (RR) is used as an alternative method to OLS in the presence of multicollinearity to obtain stable estimates of parameters of a model when standard multiple regression methods fail. Here, a new ordinary ridge estimator is suggested based on weights, where weights being the first two largest eigen values. The performance of the suggested estimator is evaluated empirically, for a wide range of degree of multicollinearity between any two predictors with some of the well-known existing estimators. Results showed that the MSE of the proposed estimator is least and it yields more stable estimates as compared to MSE of the other estimators considered in this study.