Background to BUGS
The BUGS (Bayesian inference Using Gibbs Sampling) project is concerned with flexible software for the Bayesian analysis of complex statistical models using Markov chain Monte Carlo (MCMC) methods. The project began in 1989 in the MRC Biostatistics Unit, Cambridge, and led initially to the `Classic’ BUGS program, and then onto the WinBUGS software developed jointly with the Imperial College School of Medicine at St Mary’s, London. Developments were later focused on OpenBUGS, an open source equivalent of WinBUGS.
This site at the MRC Biostatistics Unit hosts the stand-alone WinBUGS 1.4.3 package.
- Features a graphical user interface and on-line monitoring and convergence diagnostics.
- Over 30000 downloads, and a huge number of applications and links.
- WinBUGS development site includes facilities to add distributions, functions, and includes add-ons for pharmacokinetic modelling, differential equations, and reversible jump MCMC.
- Can be called from R with R2WinBUGS.
WinBUGS 1.4.3 is stable and recommended for standard use. However OpenBUGS is now also stable and has been tested on a wide range of examples.
Note: The preferred reference for citing WinBUGS in scientific papers is: Lunn, D.J., Thomas, A., Best, N., and Spiegelhalter, D. (2000) WinBUGS — a Bayesian modelling framework: concepts, structure, and extensibility. Statistics and Computing, 10:325–337.
The OpenBUGS project can be found here. This is an open-source program, developed from WinBUGS, with a variety of interfaces.
- Runs under Windows with a very similar graphical interface to WinBUGS.
- Runs on Linux with a plain-text interface.
- Can be embedded in R as BRugs.
MultiBUGS is a new package that builds on the existing algorithms and tools in OpenBUGS and WinBUGS, and automatically parallelises the MCMC algorithm to dramatically speed up computation. See the MultiBUGS site for more information.
JAGS (Just Another Gibbs Sampler) by Martyn Plummer is an open source program which was developed independently of the BUGS project. JAGS uses essentially the same model description language, but it has been completely re-written. This runs natively on Windows, Mac, Linux and several other varieties of Unix.
Stan is another program for general Bayesian analysis, developed even more recently at Columbia University. It uses a modelling language inspired by BUGS and superficially similar, but it is conceptually different in many ways.
NIMBLE can also be used to fit general models written in the BUGS language, and includes the ability to write novel sampling algorithms.
The programs are reasonably easy to use and come with a wide range of examples. There is, however, a need for caution. A knowledge of Bayesian statistics is assumed, including recognition of the potential importance of prior distributions, and MCMC is inherently less robust than analytic statistical methods. There is no in-built protection against misuse.