R-INLA

R-INLA is a package in R that do approximate Bayesian inference for Latent Gaussian Models.

Installation

The installation of the R-INLA package is a bit more complicated than most R-packages, as it downloads other pre-compiled executables that need to be made compatible with our Standard software environment.

The following installation script has been tested in Summer of 2022 with R 4.2.1 and R-INLA. Because R will always install the latest versions of packages, the versions of the modules will likely have to be adjusted in the future.

File : install_INLA.sh

#!/bin/bash

# (1)
module load gcc/9.3.0 r/4.2.1 geos/3.9.1 gdal/3.2.3 udunits/2.2.26 gsl/2.6

LOGFILE=r_INLA_install_${EBVERSIONR}_${CC_CLUSTER}_$(date --iso=sec).log

# (2)
export R_LIBS="$HOME/R/x86_64-pc-linux-gnu-library/${EBVERSIONR:0:3}"
echo  "R_LIBS is $R_LIBS"
mkdir -p $R_LIBS
R -e 'install.packages("INLA", repos=c("https://cran.utstat.utoronto.ca/", INLA="https://inla.r-inla-download.org/R/stable"), dep=TRUE, Ncpus=2)' \
     |& tee  $LOGFILE

# (3)
R -e 'library(INLA); inla.binary.install(os="CentOS Linux-7")' |& tee -a $LOGFILE

# (4)
chmod u+x $R_LIBS/INLA/bin/linux/64bit/{*.so.*,*.so,first/*.so}
sed -i  's/\(^.*export LD_LIBRARY_PATH\)/echo "Skipping LD_LIBRARY_PATH." #\1/g' $R_LIBS/INLA/bin/linux/64bit/*.run
setrpaths.sh --path $R_LIBS/INLA/bin/linux --add_path '$ORIGIN/first:$ORIGIN'

Comments in the script:

• (1) Load required modules. The same modules have to be loaded in the job script as well.
• (2) Install the R-INLA package and its dependencies
• (3) Install the pre-compiled executables that R-INLA needs
• (4) Patch the pre-compiled executables so that they are compatible with our Standard software environment