A1163
Title: Penalized variable selection for joint AFT random-effect model with clustered competing-risks data
Authors: Lin Hao - Weifang University of Science and Technology (China) [presenting]
Il Do Ha - Pukyong National University (Korea, South)
Abstract: Clustered competing-risks data frequently arise in clinical studies, such as multi-center clinical trials, where the occurrence of an event within a cluster may hinder the observation of other types of events. The correlation induced by clustering can be modeled through random effects. These competing-risks data are typically analyzed using hazard-based models rather than direct modeling of survival times. A cause-specific joint accelerated failure time (AFT) random-effects modeling approach for analyzing clustered competing-risks data has been proposed, which offers straightforward interpretability. However, there are no prior studies on variable selection for this approach, which is an interesting work. To address this, a penalized h-likelihood (HL) procedure for variable selection of fixed effects in the joint AFT competing-risks model was developed. Simulation studies were conducted to evaluate the performance of the proposed variable selection method, with results indicating that the penalized SCAD and HL methods demonstrate superior appropriateness compared to LASSO. Finally, the proposed approach was further illustrated with two real clinical datasets.
