Censored quantile regression with partially functional effects

• 成果类型: Article
• 署名作者: Qian, Jing; Peng, Limin
• 署名单位: Harvard University; Harvard T.H. Chan School of Public Health; Emory University
• 刊物名称: BIOMETRIKA
• ISSN/ISSBN: 0006-3444
• DOI: 10.1093/biomet/asq050
• 发表日期: 2010
• 页码: 839850
• 关键词: linear rank-tests; survival analysis

摘要

Quantile regression offers a flexible approach to analyzing survival data, allowing each covariate effect to vary with quantiles. In practice, constancy is often found to be adequate for some covariates. In this paper, we study censored quantile regression tailored to the partially functional effect setting with a mixture of varying and constant effects. Such a model can offer a simpler view regarding covariate-survival association and, moreover, can enable improvement in estimation efficiency. We propose profile estimating equations and present an iterative algorithm that can be readily and stably implemented. Asymptotic properties of the resultant estimators are established. A simple resampling-based inference procedure is developed and justified. Extensive simulation studies demonstrate efficiency gains of the proposed method over a naive two-stage procedure. The proposed method is illustrated via an application to a recent renal dialysis study.