A0829
Title: Induced-smoothed quantile regression analysis for competing risks data under case-cohort study
Authors: Dongjae Son - Yonsei University (Korea, South)
Sangbum Choi - Korea University (Korea, South)
Sangwook Kang - Yonsei University (Korea, South) [presenting]
Abstract: Cohort sampling designs offer an economical and efficient way of investigating association between exposure variables and risk of disease outcomes. A case-cohort design is a cohort sampling design in which a disproportionate fractions of failures and censored subjects are sampled. We consider competing risks data arising from case-cohort studies and propose statistical inference procedures for fitting censored quantile regression models for such data. Estimation of regression parameters is based on an induced smoothing approach applied to nonsmooth weighted estimating equations. Two types of weight are considered - The inverse censoring probability and sampling probability weights are included to account for competing risks data and biased feature in case-cohort samplings, respectively. The proposed induced smoothed estimating functions are smooth in regression parameters enabling one to apply the standard numerical algorithms such as the Newton's method. An iterative algorithm is proposed to simultaneously estimate regression parameters and their variances. Asymptotic properties of the proposed estimators are established. Finite sample properties are investigated through extensive simulation studies. The proposed methods are illustrated with two real data sets.
