bugsParallel is a set of R functions for distributed computing with WinBUGS

bugsParallel:

• Has been successfully installed on MS Windows and Mac OS X platforms.
• Is a modified subset of the R2WinBUGS package.
• Uses the R package Rmpi to implement a network of Windows workstations.
• Uses the rlecuyer package for parallel random number generation.

When used for a Bayesian analysis, Markov chain Monte Carlo (MCMC) simulations generate samples that approximate the joint posterior distribution of the sampled parameters. The sequential nature of MCMC simulation limits the benefits of parallel computation when applied to a single chain. However parallel computation of multiple chains is an “embarassingly parallel” problem that can substantially reduce computation time and is relatively easy to implement using bugsParallel in combination with other freely available software. bugsParallel is an active, open-source project of Metrum Research Group (http://www.metrumrg.com).

Downloads

• R4bugsParallel.zip: R software required for installation on MS Windows platforms
• windowsSW4bugsParallel.zip: Additional software required for installation on MS Windows platforms
• bugsParallelInstall.pdf: bugsParallel-1.0 installation and use documentation
• bugsParallel.zip: bugsParallel-1.0 installation package