A0264
Title: Inverse weighted quantile regression with partially interval-censored data
Authors: Yeji Kim - Korea university (Korea, South) [presenting]
Sangbum Choi - Korea University (Korea, South)
Seohyeon Park - Korea University (Korea, South)
Dipankar Bandyopadhyay - Virginia Commonwealth University (United States)
Taehwa Choi - Duke University-Department of Biostatistics and Bioinformatics (United States)
Abstract: A new inverse-probability censoring weighted (IPCW) estimating procedure for censored quantile regression with partially interval-censored data that include doubly-censored (DC) data and partly interval-censored (PIC) data is proposed. In addition to a certain amount of exact observations, DC data have either left-censored or right-censored data. In contrast, PIC data contain some interval-censored data, frequently occurring in the medical registry or HIV/AIDS clinical studies. Although various complex estimating techniques have been developed for censored quantile regression with DC and PIC data, a more simple and intuitive IPCW adjustment is considered, which can be effectively implemented by assigning a proper inverse-probability weight to each subject with an exact failure time observation. Asymptotic properties, including uniform consistency and weak convergence, are established for the resulting estimators. Further, an augmented-IPCW (AIPCW) estimation approach is discussed to gain more efficiency. Moreover, the proposed method can be readily adapted to handle multivariate partially interval-censored data. Simulation studies show the excellent finite-sample performance of the new inference procedure. The practical utility of the method is illustrated by an analysis of progression-free survival data from a phase III metastatic colorectal cancer clinical trial.
