ROC curves for nine simulated data sets obtained with the three different classifiers.

Six PDM data sets (T = 10, 50 and 100 generations and J = 4 and 8 populations) and 3 MDM data sets (FST = 0.05, 0.1 and 0.15 and J = 8 populations) were analyzed. Two ROC curves per analyzed data set were generated (in red for SNPs subjected to positive selection and in blue for SNPs subjected to balancing selection) for each of the three classifiers compared: i) classifier based on BF estimated under model 3 (solid line), ii) classifier based on PPP-values estimated under model 2 (dashed line) and iii) classifier based on PPP-values estimated under model 1 (dotted line).