COMPARING THE MOST COMMONLY USED CLASSICAL METHODS FOR DETERMINING THE RIDGE PARAMETER IN RIDGE REGRESSION

Abstract

Multicollinearity is a most common problem in multiple regression models. Many different methods are developed to solve the problem of multicollinearity. Ridge regression estimation is a popular one of these methods. And the mostly investigated matter of this method is determining the ridge parameter; therefore, various procedures are developed for determining the ridge parameter in ridge regression. In this study, we compared the commonly used methods for choosing the ridge parameter in an actual data set taken from General Directorate of Turkish Highways and Turkish Statistical Institute. We calculated the ridge parameter by the methods which are known as ridge trace, ordinary ridge estimator and an iterative method for ordinary ridge estimator. We indicated the good and worst aspects of these methods in terms of mean square error (MSE), variance inflation factors (VIF) and the other multicollinearity statistics. According the results of this study, we found that, the ridge trace method is more useful then others when it is evaluated together with the statistics of model.