Model selection via two-step Bayes factor criterion.

Based on 2,000 simulations for each true model assuming n = 25 individuals with 100 loci with 50 mutations each on average. For each simulation, the coalescent parameter is fixed and the growth parameter g and the allele misorientation rate e are randomly chosen (g ∈ [0, 11.25], e ∈ [0, 0.1]). The second column shows whether the parameters used for simulation were included in the inference grid. Fractions are rounded to two digits. The maximum of each row is marked in bold. MMC refers to cases in which neither the Psi- nor Beta-coalescent is preferred. An expanded version with enhanced sample size is provided in Table B in S1 Appendix. For details on simulations and inference parameters see A.7 in S1 Appendix.