In the analysis of ultrahigh-dimensional survival data, identifying key factors associated with patients' survival times from a large set of covariates is essential for optimizing treatment strategies and advancing precision medicine. Feature screening is an effective approach to address this issue, but one key and challenging 问题 is how to choose an appropriate threshold to distinguish important variables from irrelevant ones. In this paper, we propose a data-driven threshold selection method based on a data-splitting strategy and a reassembling technique, enabling the selection of an appropriate threshold in different scenarios to better distinguish between active and inactive covariates. Based on the selected threshold, we introduce a feature screening method for ultrahigh-dimensional right-censored survival data. The significant advantage of this method is that it not only allows for rapid and effective dimensionality reduction while accurately retaining all important variables but also controls the false discovery rate (FDR) within a certain range. Under certain regularization 条件, we establish the theoretical framework for the proposed procedure, including the sure screening 性质, the rank consistency 性质, the finite-sample theory for FDR control, and the asymptotic theories for FDR control. Numerical simulations show that the proposed method performs excellently in both variable selection and FDR control,and outperforms existing methods significantly. Finally, we apply the proposed method to analyze the diffuse large B-cell lymphoma dataset, further validating its effectiveness and practical applicability.
Jing et al. (Thu,) studied this question.