Feature Screening Via Distance Correlation For High-dimensional Interval-censored Data

Feature screening methods have an important role in identifying the main factors affecting survival time in high-dimensional survival data.However,failure-time data are often accompanied by deletions in survival analysis,which further complicates the study of feature screening methods.The existing methods are mainly applied in right censored data,with few studies on feature screening methods of interval censored data.Based on the feature screening method,this paper mainly studies the feature screening problem under two interval deletion data.The full text is divided into the following two parts:The first part considers the feature screening problem under t Case Ⅰ interval-censored data.In particular,we propose a model-free feature screening method based on the distance correlations between vectors of interest,and the numerical simulation results show that the proposed algorithm achieves high feature selection accuracy and stable performance in the presence of random deletion of observation time,nonrandom deletion,and outliers of covariates.Finally,the method was applied to non-Hodgkin lymphoma data to verify the effectiveness of the method.The second part considers the feature screening problem under the Case Ⅱ interval-censored data.The frequency of high-dimensional interval censored data is becoming more and more frequent.How to reduce the dimension of high-dimensional interval censored data is a problem worth studying.Based on the idea of distance correlation,this chapter proposes an independent feature screening method for dimension reduction of the Case Ⅱ interval-censored data in a high-dimensional interval,which does not rely on various models.The numerical simulations consider three cases under different censoring ratios,and a large number of numerical simulations verify the effectiveness of this method for finite samples.The proposed method was applied to a longitudinal dental data to illustrate the feasibility and utility of the proposed method.