Reversible-Jump MCMC Using Post-Processing
Performs reversible-jump Markov chain Monte Carlo (Green, 1995) <doi:10.2307/2337340>, specifically the restriction introduced by Barker & Link (2013) <doi:10.1080/00031305.2013.791644>. 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) <doi:10.1111/anzs.12263>.