Skip to main content
. 2020 Apr 25;69(6):1163–1179. doi: 10.1093/sysbio/syaa035

Figure A2.

Figure A2.

Binary skyline model selection. We simulate 500 conditionally independent phylogenies from skyline models and test the classification ability of model selection criteria. The null model is a Kingman coalescent with Inline graphic, and the alternative features a sharp fall to Inline graphic at Inline graphic time units. The sampled tips of the phylogeny are introduced at Inline graphic and Inline graphic only. Graph (i) gives the probability of correct classification Inline graphic as a function of data size Inline graphic. The FIA performs best, on average, but the BIC is better at small Inline graphic. Graph (ii) gives the true (TPR) and false positive rates (FPR) of the metrics. The FIA and QK have the best overall rates.