Fig. 5.
Validation of ABC model choice and power analysis. (A) A comparison of the posterior probability as estimated via ABC (pABC) against an empirical estimate of the same probability obtained through simulations (pempirical, see Materials and Methods) reveals that our ABC approach for model choice is well calibrated. For both Models 2 and 3, however, only very few simulations resulted in high pABC, which made the estimation of pempirical problematic. We thus only plotted for bins for which we obtained at least 20 simulations (out of 105 in total). (B) The relationship between number of loci and statistical power to distinguish the three colonization models for D. suzukii. Points denote results of simulations and line is the fitted linear regression. The x-axis (number of loci) is log10 scaled.