Figure 5.

Summary of model comparisons. Difference in marginal likelihoods (Bayes factor) estimated by path-sampling between the worst model (linked strict molecular clock, unlinked trees) and all others. Asterisks indicate the best-performing model (unlinked relaxed clocks, linked trees, run 2) for MERS-CoV data. Analyses employing a relaxed molecular clock were run independently 3 times, those with a strict molecular clock 2 times. Marginal likelihoods estimated using stepping stone sampling gave identical results.