A1491
Title: Data-adaptive conditional quantile screening with FDR control in ultra-high-dimensional data
Authors: Saidat Abidemi Sanni - The University of Texas at San Antonio (United States) [presenting]
Yan Yu - University of Cincinnati (United States)
Zhigen Zhao - Temple University (United States)
Abstract: A conditional quantile feature screening procedure for ultra-high-dimensional data with false discovery rate control is proposed. Motivated by genetic association studies, the problem of screening ultra-high-dimensional predictors associated with a specific conditional quantile of the response, given a potentially multivariate set of covariates, is considered. The approach uses quantile partial correlation to assess quantile-specific conditional associations, constructs quantile mirror statistics, and applies a data-adaptive thresholding rule to control the false discovery rate. The method provides a practical tool for quantile-specific feature screening with covariate adjustment, data-adaptive thresholding, and error-rate control in ultra-high-dimensional settings. Simulation studies and a real-data analysis of genetic risk factors for abnormal blood pressure demonstrate the method's performance relative to existing approaches.
