Gene identification using independent screening procedure

In a survival analysis, identifying influential genes is a crucial task in ultra-high dimension. In this paper, we particularly analyze Cardiomyopathy microarray data in a ultra-high dimensional setting and propose a screening method to identify most important influential genes. In this respect, we use an additive model and approximate the nonparametric additive components using a B-spline basis. We first apply a fast screening procedure to reduces the number of covariates to a moderate order and then apply the adaptive group Lasso to perform the final variable selection and parameter estimation simultaneously