3.5 Demographic modeling and analysis of gene flow
Using machine learning-based demographic model selection, we identified
divergence without gene flow as the best model for T. blandingiiwith a posterior probability of 0.68, and divergence with gene flow forT. pulverulenta with a posterior probability of 0.63 (Fig. 6).
For both species, models representing no divergence and divergence with
secondary contact received very low support (Tables S2 and S3). The
out-of-bag error rate for T. blandingii was 17.3% and 22.8% forT. pulverulenta . Our values for posterior probability and
out-of-bag error rate are similar to those obtained by Smith & Carstens
(2020). The confusion matrix and number of votes per model can be found
in Tables S2 and S3.