Name: Felipe Javier Medina
Institution: University of Reading, UK
The reversible-jump MCMC (rjMCMC) algorithm is a popular tool for exploring a posterior distribution of varying dimension. However, rjMCMC commonly performs poorly due to the difficulty of proposing “good” trans-dimensional moves. Recent developments in Bayesian computation methodology provide ways of improving rjMCMC using ideas based on sequential Monte Carlo (SMC) methods. We present an implementation of these ideas for a model of recombination within bacteria. This recombination process is typically modelled using an ancestral recombination graph (ARG), which is the generalisation of a coalescent tree. Inference is then performed via a joint posterior (of no fixed dimension) on the recombination rate, the number of recombination events, and a set of parameters for each of these events.