Abstract

The Integrated Nested Laplace Approximation (INLA) methodology has been shown to produce accurate and efficient approximations for Bayesian inference of Generalized Additive MIxed type models, when compared to sampling-based approaches like MCMC. In this talk we show how INLA can also be used for full Bayesian inference of multi-outcome models through the INLAjoint user interface. INLAjoint facilitates the setup of the model and performs Bayesian inference using INLA. Multiple examples are used to illustrate the impact of INLA and INLAjoint for Bayesian inference of multi-outcome models. In complex models or large data, INLA can perform Bayesian inference with speedups of 100X+ when compared to HMC and MCMC, with the same accuracy. This framework also enables the inclusion of random effects such as frailties, temporal and spatial effects, splines and more.