SIMULATION STUDY ON SOME VARIABLE SELECTION PENALIZED METHODS

This paper focuses on lasso and elastic net penalized methods. These regularization methods are being investigated from two major aspects of regression model analysis, the exploration and the prediction. Exploration can be obtained by using Variable Selection, while the prediction can be obtained by trading off the bras and variance of the estimated parameters. Both lasso and elastic net provide Space solution with which in turn yield more interpretable and more prediction accuracy for the estimated model. Variable selection demonstrates by simulation experiments and the results show that these penalized methods then give sparse solution and performs well compared with some other penalized methods.