Argues Pekar et al. 2022’s (S-1) simulation-based hypothesis test applies asymmetric standards to the single- vs. two-introduction models: the single-introduction model’s simulated epidemics are required to match the observed topology under stricter conditions than the two-introduction model’s are, so part of the reported Bayes-factor advantage for two introductions is a byproduct of the test’s construction rather than of the data. Reruns the comparison under matched conditions for both hypotheses and finds the quantitative support for multiple introductions disappears (or reverses). Independent of, but reaches a similar bottom line to, Weissman’s (S-11) separate diagnosis of the same Bayes-factor calculation.
relevance_note: a second, independently-derived methodological critique converging with S-11 on “the published Bayes factor for two introductions is an artifact of the test setup.”