Improved ALASSO Methods For Variable Selection In Two Kinds Of Survival Models

In this paper,we study the places that can be improved when selecting variables in the proportional hazards model and proportional odds model with ALASSO method.We give some new methods in the choosing of penalty weightsτin this paper,and also make a discussion and improvement for the estimation of the proportional hazards model with certain simulation data.In addition,in the known ALASSO method,it estimated the proportional odds model by maximizing the marginal likelihood function,but the selection of points in the simulation of the marginal likelihood was based on the uniform distribution which might be instable and slow.We improve the method by using the Number Theory to select points in this paper.We also give the simulation of the variable selection for the two kinds of models with the improved methods in this paper,and the result shows that the effect of improvement is obvious.