Title: Joint models for longitudinal and survival data with INLA: Applications in credit scoring
Authors: Victor Medina-Olivares - University of Edinburgh (United Kingdom) [presenting]
Raffaella Calabrese - University of Edinburgh (United Kingdom)
Jonathan Crook - University of Edinburgh (United Kingdom)
Abstract: Joint models for longitudinal and survival data are an appealing modelling framework for credit scoring since they allow to jointly model the time to default and the internal time-varying covariates usually seen in credit risk data. However, the estimation procedure is computationally expensive and sometimes unfeasible for the size of the data in the credit context. Although, if the joint model is assumed with a linear bivariate Gaussian association structure, then it can be seen as a latent Gaussian model (LGM) and thus the Bayesian inference can be approximated with the integrated nested Laplace approximation (INLA). We propose a joint model for longitudinal and survival data in a discrete-time setting with applications in credit scoring and estimated with INLA.