Reversible-Jump MCMC Using Post-Processing

Performs reversible-jump Markov chain Monte Carlo (Green, 1995) , specifically the restriction introduced by Barker & Link (2013) . By utilising a 'universal parameter' space, RJMCMC is treated as a Gibbs sampling problem. Previously-calculated posterior distributions are used to quickly estimate posterior model probabilities. Jacobian matrices are found using automatic differentiation. For a detailed description of the package, see Gelling, Schofield & Barker (2019) .

Performs reversible-jump MCMC, a Bayesian multimodel inference method. The process is simpler than a manual implementation; for instance, all Jacobian matrices are automatically calculated using the madness package. The effort required to find Bayes factors and posterior model probabilities is reduced.


For each model considered, the user requires a posterior distribution obtained via MCMC or the like. They then define a bijection between its parameter space and the universal parameter space; the likelihood model on the data; and the priors on the parameters. The rjmcmcpost function uses a post-processing algorithm to estimate posterior model probabilities. See ?rjmcmcpost for a simple example using binomial likelihoods.


install.packages("rjmcmc") library(rjmcmc)


Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


0.4.5 by Nick Gelling, 8 months ago

Browse source code at

Authors: Nick Gelling [aut, cre] , Matthew R. Schofield [aut] , Richard J. Barker [aut]

Documentation:   PDF Manual  

GPL-3 license

Imports utils, coda, mvtnorm

Depends on madness

Suggests FSAdata

See at CRAN