Marginal likelihood estimation via power posteriors

It is shown how the marginal likelihood can be computed via Markov chain Monte Carlo methods on modified posterior distributions for each model, which then allows Bayes factors or posterior model probabilities to be calculated.