B0240
Title: Competing risks joint model and other complex survival models using R-INLA
Authors: Janet Van Niekerk - King Abdullah University of Science and Technology (Saudi Arabia) [presenting]
Abstract: The Integrated Nested Laplace Approximation (INLA) method is an efficient deterministic approximate Bayesian inference tool that has been used extensively in most fields of statistics and data analysis. This methodology is implemented in the R library INLA (www.r-inla.org) and is continuously reviewed and expanded. Recently, by formulating (complex) survival models as latent Gaussian models, great strides have been made in the efficient Bayesian inference of these models using INLA, instead of sampling-based approaches, which become less efficient as model complexity increases. We will present some complex survival models that we have implemented in INLA - some of which can only be fitted through MCMC besides INLA. Discrete joint models, non-linear joint models, spatial joint models, joint models with competing risks, two-part joint models, illness-death models, and spatial multi-state models are only some examples that can be efficiently estimated using the INLA methodology. The focus will be on complex joint models, even though the methodology presented can be adapted to various survival models quite trivially.
