A1032
Title: High-dimensional variable selection for partially functional Cox regression with interval-censored data
Authors: Tian Tian - University of Missouri (United States)
Yuanyuan Guo - Duke University (United States) [presenting]
Jianguo Sun - University of Missouri (United States)
Abstract: Variable selection is considered for interval-censored data with partially functional covariates and potential nonlinear effects. A flexible additive Cox model is proposed to incorporate the functional principal component analysis for modeling the functional predictors and Bernstein polynomials to approximate the nonlinear effects. We develop a penalized sieve maximum likelihood approach with an efficient group coordinate descent algorithm to allow for both low- and high-dimensional scenarios. The performance of the presented approach is assessed via a simulation study and the analysis of the Alzheimers Disease Neuroimaging Initiative (ADNI) data.
