Fig. 18

Sidewinder/horned rattlesnake (Crotalus cerastes), with scores being the negative log-likelihoods; \(n =\) 1300; Kuiper’s statistic is \(0.1343 / \sigma = 10.47\), Kolmogorov’s and Smirnov’s is \(0.1343 / \sigma = 10.47\). The reliability diagrams whose bins each contain roughly the same number of subpopulation scores are either noisier or missing the severest deviations in comparison with the other reliability diagrams; so there do exist cases in which choosing bins that are nearly equispaced along the scores works better than choosing bins that each contain roughly the same number of subpopulation scores. The scalar summary statistics extremely successfully detect the statistically highly significant deviation of the subpopulation from the full population