On Tuning Parameter Selection Of Lasso-type Methods

Over the last decade,many regularized procedures have been proposed for regression analysis,and lasso-type method have become popular for variable selection due to their property of shrinking some of the model coefficients to exactly zero.Although the lasso-type methods make great use in variable selection,but their performance depends highly on the tuning parameter selection.In the top half of this paper,we attempted to provide an overview of methods which are available to select the value of the tuning parameter.In the other stability selection methods would be more considered.Here a criterion denoted as MPASS which modified by PASS criterion,one of the stability selection method,was proposed.A simulation study provided a comparison of these methods and assessed their performance confined by various condition.Numerical results of several examples were given to demonstrated the MPASS method outperforms other competitors in variable selection performance.Furthermore,when the dimension of the data increases with the sample size,MPASS performed much better than other methods in terms of having the largest probability of choosing the correct model.