Fig. 5.

Example of empirical significance tests. Distributions of internal accuracies from 500 null models are shown as histograms. Internal accuracy from the corresponding original model is marked by the blue vertical line. A) Empirical testing results for the AR (+) classifier, which generated an internal accuracy higher than 97.2% of null models, resulting in empirical p-value = 0.028. B) Empirical testing results for the A/B-CHE (−) classifier, which generated an internal accuracy score higher than only 43.8% of null models, resulting in empirical p-value = 0.562