Estimation and selection of survival models for individuals with spatial frailty
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: estimating equation, spatial frailty model, survival analysis, variable selection
Session: IPS 675 - Uncertainty Quantification in Spatial Statistical Inference
Tuesday 7 October 2 p.m. - 3:40 p.m. (Europe/Amsterdam)
Abstract
In simple survival models, the hazard function often overlooks spatial correlation at the individual level, which can be a significant factor in many real-world scenarios. For instance, individuals located nearby may share environmental, social, or demographic factors that affect their survival times. Spatial frailty models address this gap by incorporating spatially structured random effects. In this talk, I introduce an estimating equation method for the Weibull-Cox proportional hazards model with spatial frailty, which handles spatial survival data at the individual level. We also consider a penalized estimating equation to enable group variable selection. We establish consistency and asymptotic normality as well as an oracle property for the estimator. A simulation study with various scenarios supports our theoretical findings. We also apply the proposed approach to real data, focusing on the survival of businesses in developing commercial districts of Seoul, South Korea. Our results suggest that the proposed method is a useful tool for analyzing individual-level, correlated spatial frailty models.