• Talk by Prof. Håvard Rue and Dr. Janet van Niekerk are presenting at 15th International Conference of the ERCIM WG on Computational and Methodological Statistics, Dec 17-19, 2022 . Check details

  • Spatial Modeling using R-INLA, International Conference on Mathematics Computational Sciences and Statistics , Oct 2022

Professor Håvard Rue and Dr. Elias Krainski are presenting at IcoMcoS 2022, Indonesia. For more details please visit this site.

  • Talk by Dr. Denis Rustand entitled by "Efficient estimation of joint models for multivariate longitudinal and survival data using INLA", 43rd Annual Conference of the International Society for Clinical Biostatistics, Newcastle, UK, August 2022

  • For more details please visit the site.

  • Sessions presented by Dr. Denis Rustand and Dr. Janet van Niekerk, 31st International Biometric Conference, Riga, Latvia, July 2022

- "Bayesian estimation of joint models for longitudinal markers and a terminal event with R-INLA" (Denis Rustand, contributed session).

- "Complex joint survival models with INLA" (Janet van Niekerk, invited session).

  • INLA BayesComp group members at KAUST have presented their research projects at ISBA Montreal, June 2022

  • "Leave-Group-Out Cross-Validation for Latent Gaussian Models" by Zhedong Liu

  • "Approximate Bayesian Inference for the Interaction Types 1, 2, 3 and 4 with Application in Disease Mapping" by Esmail Abdul Fattah

  • "Correcting the Laplace Method with Variational Bayes: Marginal variance" by Shourya Dutta

  • "Parallel Krylov Approximations for Latent Gaussian Models" by Ablay Zhumekenov

  • INLA course at Bordeaux population health center, June 2022

Some members of the INLA team presented an INLA short course to the Biostatistics group at the Bordeaux population health center of INSERM and the University of Bordeaux. The content was tailored for biostatistics and public health applications, to avail INLA as a tool for fast Bayesian inference of applicable statistical models.

  • Updates on R-INLA, 2021

''There was a regression bug in INLA version 21.12.21 which we discovered on Jan 11, giving wrong expansion of some variables (in some cases), and this version is now removed from the inla site. This is fixed in the new build 22.01.12 starting Jan 12."

  • Winter School December 2021: Details are here.

Title: Contemporary methods in Spatial Statistics in R with applications to life and earth sciences.

  • New testing-version is out, Version_21.07.10-1, July 2021.

This version is built with R-4.1 for Mac, Windows and Linux

- See 'news(package="INLA")' for changes.

Use this command to activate (TRUE) or disactivate (FALSE) this technique:

inla(control.inla=list(use.directions = TRUE))

  • New version of the R-INLA package, June 2021

  • 'A new version of the R-INLA package, 21.06.11, is out with some nice new features.

  • Native build for R-4.1 and Mac M1 processor, included ported PARDISO library.

  • For non-Mac-M1, no R-4.1 support yet, we're still building on R-4.0.

  • Updated PARDISO library to version 7.2 on (Intel based) Mac. The largest change, is a preview of a new option that with 'twostage=TRUE' enables an internal reformulation of the model that will run faster for 'data-rich' models. It is default not enabled. This feature is highly experimental and will likely break in some cases. If it does, please send a reproducible example so it can be fixed.


  • Using ‘rgeneric’ on R-4.1, 28 May 2021

''If you are using 'rgeneric' and R-4.1 on Linux, you need to remove the in the binary build, like rm ~/R/x86_64-redhat-linux-gnu-library/4.1/INLA/bin/linux/64bit/

when we build towards R-4.1, this error will go away.''


  • Updates on R-4.1, 21 May 2021

''R-4.1 is now out, and we have not yet built a new package for 4.1. Hopefully the packages for 4.0 will work, which we have not tested.

We are working to support R-4.1.0-arm64.pkg , Apple silicon arm64, with a specific build but we are not there yet. Hopefully soon.''


  • Multivariate spatial models for lattice data with INLA, 7 May 2020

A paper on fitting several multivariate spatial models for lattice data with INLA has been published in the arXiv by F. Palmí-Perales et al. The associated R package INLAMSM is available from GitHub and CRAN. Models have been implemented using the rgeneric framework and they include a multivariate intrinsic CAR mode, multivariate proper CAR model and the M-model.

  • macOS 10.15 Catalina, 13 Jul 2020

is now used to build the INLA package for Mac. Let us know if this causes any issues.

  • Finally, R-4.0, 9 Jul 2020

The testing version of today, Version_20.07.09, is built on R-4.0.2. This also includes all alternative binary Linux builds, but they are less affected.

There is currently no stable version built with R-4.0, but the testing one can be used instead. If there are any issues, please let us know.

  • Windows ``builder'' wanted

Our knowledge in Windows is limited, meaning that the builds provided in the package works but sometimes with limited support. Like, the PARDISO library, Intel MKL support, jemalloc, is not there for this reason. If someone would like to help us out here, it would be great! The task is essentially to provide a better Windows build of the inla-program for each release, which can be downloaded using similar functionality as `inla.binary.install()', The code is mainly C, with some C++ and Fortran. If you're interested, please email me at .

  • The R-INLA Workshop was a Big Success: An international virtual workshop entitled Introducing R-INLA and its Applications was held by Statistics Study Program on Wednesday, September 30th, 2020 at Airlangga University, Indonesia. Click here.

  • Solving Real-World Problems: A tool developed by Håvard Rue has transformed data analysis, interpretation and communication. It has been applied broadly: from modeling the spread of infectious diseases to mapping fish stocks, 2018.

  • Model 'rgeneric' and R-4.0, 29 Sep 2020

There is a minor change in the R-internals in R-4.0 that may make rgeneric models crash if the statement (used in the rgeneric vignette), is used,

if (is.null(theta)) theta = initial()

This statement is used to ensure that 'theta' is always set as some argument do not depend on 'theta'. In R-4.0 you need to change this statement to

if (!length(theta)) theta = initial()

The vignette and examples are updated and will be available in the next build.

Thanks to GV who reported this error.


  • An INLA-translation of the 2019 homework of Statistical Rethinking: (2nd edition), 23 Sep 2020

Impressive, well done!


  • Stability issues for Ubuntu 18.04 and 20.04, 2 Sep 2020

In the most recent testing version(s) there is a stability issue for Ubuntu when running with nested parallelism and the PARDISO library. It seems like the builds from CentOS 7 & 8 are much more stable.

Is this becomes an issue, please do 'inla.binary.install()' and try the CentOS7 (Ubuntu 18) or CentOS8 (Ubuntu 20) instead. We're looking into the issue.