Abstract
We study the variable selection problem in Cox proportional hazards model for prevalent survival data. In this study, we face some challenges. Firstly, from many potential predictors, we would like to select a small number of key risk factors, including continuous or discrete variables. Secondly, data were collected from a prevalent sampling which is exactly a biased sampling scheme. The proposed method not only can select and estimate variables simultaneously but also can correct the sampling bias. Further, the proposed method can allow for different penalty functions, including continuous or discrete variables. The results of simulation study show that the proposed procedure is stable and more accurate to select the true model. We also apply the proposed method to a real data.