A NEW METHOD FOR COVARIATE SELECTION IN COX MODEL
In a wide spectrum of natural and social sciences, very often one encounters a large number of predictors for time to event data. An important task is to select right ones, and thereafter carry out the analysis. The l1 penalized regression, known as “least absolute shrinkage and selection operator" (LASSO) became a popular approach for predictor selection in last two decades. The LASSO regression involves a penalizing parameter (commonly denoted by l) which controls the extent of penalty and hence plays a crucial role in identifying the right covariates. In this paper we propose an information theory-based method to determine the value of l in association with the Cox proportional hazards model. Furthermore, an efficient algorithm is discussed in the same context. We demonstrate the usefulness of our method through an extensive simulation study. We compare the performance of our proposal with existing methods. Finally, the proposed method and the algorithm are illustrated using a real data set.
More...
