A0774
Title: Weighted least-squares estimation for semiparametric multivariate accelerated failure time model with regularization
Authors: Sy Han Chiou - Southern Methodist University (United States) [presenting]
Ying Chen - Harvard School of Public Health (United States)
Chuan-Fa Tang - University of Texas at Dallas (United States)
Min Chen - University of Texas at Dallas (United States)
Abstract: In large-scale epidemiology and medical studies, informative sampling and high-dimensional covariates pose significant challenges in statistical inference. These challenges warrant an accurate inference procedure that addresses the sampling bias arising from informative sampling and incorporates an efficient variable selection process within it. A weighted least-squares estimation is considered in the semiparametric multivariate accelerated failure time model framework for right-censored clustered data emerging from informative sampling. By embedding the generalized estimating equation (GEE) techniques, the proposed estimating procedure accommodates situations when observations within a cluster are correlated. The regularization techniques are further incorporated to perform variable selection by penalizing the proposed GEE procedure. The consistency and asymptotic normality of proposed estimators is established, and it is shown that the proposed variable selection method has an oracle property. Extensive simulation results indicate superior performance of the proposed estimators over the existing method that does not account for sampling bias or within-cluster dependence for multivariate responses. The proposed regularization procedure also achieves favorable variable selection performance under moderate sample sizes. Two dental studies are used to illustrate the practical applications of the proposed methods.
