Umacs: A Universal Markov Chain Sampler

Umacs (Universal Markov chain sampler) is an R software package that facilitates the construction of the Gibbs sampler and Metropolis algorithm for Bayesian inference. Umacs is a practical tool to write samplers in R. This is sometimes necessary for large problems that cannot be fit using programs like BUGS. The user supplies the data, parameter names, updating functions, and a procedure for generating starting points. The updating functions can be some mix of Gibbs samplers and Metropolis jumps, with the latter determined by specifying a log-posterior density function. Using these inputs, Umacs writes a customized R function that automatically updates, keeps track of Metropolis acceptances (and uses acceptance probabilities to tune the jumping kernels), monitors convergence, returns a summary of the results that can be coerced into random variable objects for further processing in R. Umacs is extendable so that users can also write their own updater-generating classes which may be used in this framework along with the existing Gibbs and Metropolis samplergenerating classes.