A0696
Title: Interval-censored linear quantile regression
Authors: Taehwa Choi - Duke University (United States) [presenting]
Seohyeon Park - Korea University (Korea, South)
Hunyong Cho - Amazon.com (United States)
Sangbum Choi - Korea University (Korea, South)
Abstract: Censored quantile regression has emerged as a prominent alternative to classical Cox's proportional hazards model or accelerated failure time model in both theoretical and applied statistics, as it enables researchers to investigate the complete distribution of survival responses with respect to a set of covariates. While quantile regression has been extensively studied for right-censored survival data, the survival analysis literature methodologies for analyzing interval-censored data remain limited. A novel local weighting approach is proposed for estimating linear censored quantile regression with various types of interval-censored survival data. The regression parameter's estimation equation and the corresponding convex objective function can be constructed as a weighted average of quantile loss contributions at two interval endpoints. The weighting components are nonparametrically estimated using local kernel smoothing or ensemble machine-learning techniques. A modified EM algorithm for nonparametric distribution mass for interval-censored data is employed by introducing subject-specific latent Poisson variables to estimate the nonparametric maximum likelihood estimation. The proposed method's empirical performance is demonstrated through extensive simulation studies and real data analyses of two HIV/AIDS datasets.