Rjags

From bwHPC Wiki
Jump to: navigation, search

1 General information

rjags is a R interface to use JAGS, Just another Gibbs Sampler. JAGS is a program for analysis of Bayesian hierarchical models using Markov Chain Monte Carlo (MCMC) simulation

rjags needs a JAGS installation on the side. We recommend to compile via Intel compiler and with the Intel MKL library (Intel Math Kernel Library), which allows JAGS to use various efficient implementations of mathematical computations. These are, as of now, loaded alongside with the module R 4.1.2.


2 Installation

Please enter the following code, presented in the boxes below, directly into your shell/command line on bwUniCluster

#Load R module
module load math/R/4.1.2


# Set up JAGS installation directory 
export JAGS_HOME=$HOME/sw/jags


# Prepare JAGS source directory (if not yet existing) 
mkdir -p ~/src
cd ~/src


# Get JAGS source
wget https://sourceforge.net/projects/mcmc-jags/files/JAGS/4.x/Source/JAGS-4.3.1.tar.gz
tar xf JAGS-4.3.1.tar.gz
cd JAGS-4.3.1
rm JAGS-4.3.1.tar.gz


# Install JAGS 
export CFLAGS="-O3 -axCORE-AVX512,CORE-AVX2,AVX -xSSE4.2 -fp-model strict -qopenmp"
export CXXFLAGS="-O3  -std=c++14 -axCORE-AVX512,CORE-AVX2,AVX -xSSE4.2 -fp-model strict -qopenmp"
./configure  --prefix=$JAGS_HOME --with-blas="-lmkl_rt -lpthread -lm"
make
make install
cd


# Set up environment
export PKG_CONFIG_PATH=$JAGS_HOME/lib/pkgconfig
export LD_RUN_PATH=$JAGS_HOME/lib


# Install rjags package from within R session
R -q
> install.packages("rjags", configure.args="--enable-rpath")
> library(rjags)