A new ridge type estimator and its performance for the linear regression model: Simulation and application

Abstract

Ridge regression is employed to combat the problem of multicollinearity among independent variables. The shrinkage parameter (𝑘) has key role in the bias and variance tradeoff. In this study, we have reviewed some existing promising ride regressions estimators for estimating the ridge or shrinkage parameter k for the Gaussian linear regression model. In addition, we have also proposed a new estimator (CK), which is a function of number of independent variables, sample size and standard error of regression model. The performance of our proposed estimator with OLS and existing shrinkage estimators is compared using extensive Monte Carlo simulations in terms of minimum mean squared error (MSE). The simulation findings show that the proposed CK estimator is an efficient performer in majority of the considered simulation scenarios. A real-life data are analyzed to illustrate the findings of the paper.