Abstract: Screening important features has become one of the important tasks in statistical analysis and correspondingly, various screening procedures have been proposed for various types of studies or data including both complete and incomplete data. However, these methods would be computationally costly or even infeasible when one faces massive health databases with both high dimensionality and huge sample size, which have become increasingly popular for comparative effectiveness and safety studies of medical products. In this paper, we consider such a type of incomplete data, interval-censored failure time data, that have not be discussed before and propose two procedures with the use of distance correlation and orthogonal sampling as well as the the jackknife debiased average technique. The proposed approaches can be easily implemented and their sure screening and rank consistency properties are established. Simulation studies demonstrate that the proposed methods work well for practical situations and they are applied to the SEER breast cancer data.

Key words and phrases: Distance correlation, jackknife debiased average, orthogonal subsampling, rank consistency, sure screening.