BUGS: a Program to Perform Bayesian Inference using Gibbs Sampling
Andrew C. Thomas, David J. Spiegelhalter, Wally R Gilks · 1992
Abstract We describe a preliminary version of a general program for Bayesian inference using Gibbs sampling. A crucial feature of the program is the exploitation of conditional independence assumptions in the original statistical model to allow a smooth transition between model specification in a simple language, internal representation of the model in an object-oriented environment, and automatic derivation of the necessary sampling distributions. A useful starting point is the graphical representation of conditional independence assumptions. The eventual aim is to produce a program that can handle arbitrarily complex problems by decomposing them into modular components which then can be “intelligently” handled by the software.