Quantile Regression with Survival Data Under General Biased Sampling Schemes

• Activity Number: 588
• Type: Topic Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Gongjun Xu* and Tony Sit and Lan Wang and Chiung-Yu Huang
• Companies: University of Minnesota and The Chinese University of Hong Kong and University of Minnesota and The Johns Hopkins University
• Keywords: censored quantile regression ; length-biased data ; case-cohort sampling

Abstract

We propose a unified estimation procedure for quantile regression with survival data collected under general biased sampling schemes. Biased sampling occurs frequently in economics, epidemiology and medical studies by design or due to data collecting mechanism. Important examples include left-truncation, length-biased sampling, case-cohort sampling and its variants. Failing to take into account the fact that data are collected from a biased distribution usually leads to substantial bias in the estimation. While censored quantile regression offers a valuable generalization to the celebrated Cox proportional hazards model and the accelerated failure time (AFT) model, a general approach that can handle various types of biased sampling under one consolidated framework has not been considered in the literature. In this paper, we propose a unified estimation procedure that is built upon martingale-based estimating equations with a flexible weighting scheme to correct for sampling bias. Such a unified framework provides researchers and practitioners a convenient tool for analyzing data collected from a variety of biased sampling schemes. We establish consistency and asymptotic normality