Registration and call for papers

Bayesian computing with R-INLA

VIBASS4 - Invited Course

This course (12 hours) is provided by Håvard Rue, founder of the INLA development team, Professor of Statistics and Principal Investigator of the Bayesian Computational Statistics and Modeling (BAYESCOMP) group, King Abdullah University of Science and Technology (KAUST) and Janet van Niekerk, Post-Doctoral fellow In Statistics at KAUST.


  • Audience

    Statisticians and applied researchers with strong interest in quantitative analysis

  • Abstract

Integrated nested Laplace approximation (INLA) facilitates the fitting of a large range of complex hierarchical statistical models, by dramatically reducing computation time. During two days at VIBASS4, we aim to give an introduction to the general methodology and the R-INLA package. We will start with a general overview of the possibilities of INLA for applied research and for model development, and it proceeds with examples of generalised linear models with several random effects. We will discuss the innovative ideas making INLA fast; why most likelihoods are near-Gaussian in the posterior, how to represent random effects with sparse matrices, and more.

The main part of these lectures will be practically oriented. Using examples we will explain how to fit different models in practice using R-INLA, and how to interpret the results. The models we fit include, for example, Gaussian, Binomial and Poisson likelihoods, auto-regressive, random walk, varying-intercept models, models for survival data and joint survival-longitudinal models. We will discuss the choice of priors, and how to compare models.

We will also present what we have not discussed, but could have done, like spatial (geographical) modelling and how to construct spatial models with the stochastic partial differential equation (SPDE) approach, and its use in point process models like log-Gaussian Cox processes. In ecology, these models result in habitat maps and abundance maps, and in disease mapping they result in continuous risk maps and in risk classification maps.

For a background on R-INLA, one of the books listed at can be consulted, as well as the two recent published review papers, which arxiv versions are and