Nonparametric screening and feature selection for ultrahigh-dimensional Case II interval-censored failure time data

Area of Science:

• Biostatistics
• Data Science
• Survival Analysis

Background:

• Ultrahigh-dimensional data analysis often requires feature selection to reduce dimensionality.
• Existing methods primarily address complete datasets, leaving a gap for incomplete data scenarios.

Purpose of the Study:

• To develop a novel screening procedure for case II interval-censored failure time data.
• To address the lack of existing methods for feature selection in this specific data type.

Main Methods:

• A model-free, nonparametric method based on cumulative residual is proposed.
• The method is designed to possess the sure independent screening property.

Main Results:

• The developed approach effectively ranks active variables higher than inactive ones concerning failure time association.
• Simulation studies confirm the method's utility with general survival models, capturing nonlinear covariates and interactions.

Conclusions:

• The proposed nonparametric screening method is effective for ultrahigh-dimensional, interval-censored failure time data.
• The method demonstrates practical applicability, as shown in a childhood cancer survivor study.