Title: Joint model of longitudinal and spatiotemporal survival data
Authors: Victor Medina-Olivares - University of Edinburgh (United Kingdom) [presenting]
Raffaella Calabrese - University of Edinburgh (United Kingdom)
Jonathan Crook - University of Edinburgh (United Kingdom)
Finn Lindgren - University of Edinburgh (United Kingdom)
Abstract: In credit risk analysis, it is generally of interest to model the time-to-event (survival) of a borrower according to two types of covariates: time-fixed and time-varying. When the latter presents possible endogeneity, usually seen in this context, it is preferable to incorporate them in a joint modelling approach that considers the mutual evolution of survival time and the endogenous covariates, rather than treat them separately. Moreover, it is increasingly common to incorporate geographical information about the borrower into the databases, giving way to models that also account for spatial clustering and its variation in time. A Bayesian hierarchical joint model of longitudinal and discrete survival data considering spatiotemporal frailties is proposed. This approach captures the survival effect due to the evolution of the unobserved heterogeneity among subjects located in the same region.