An independent, post-debate (Jan 2025) Bayesian analysis by a Dartmouth economist, using only spatiotemporal and zoonotic data (no genome-structure/engineering-signal evidence at all). Decomposes the odds ratio into four conditional Bayes factors: outbreak-in-China (2.3:1), outbreak-in-Wuhan-given-China (20:1), spatiotemporal pattern of cases with no known market link (27:1), and spatiotemporal pattern of vendor cases at the market (12:1) — combining to an overall 14,900:1 odds ratio favoring lab leak, which the author calls “overwhelming evidence,” the opposite conclusion from most of the debate’s six analyses despite a similar Bayesian-factorization approach and overlapping data (Worobey-derived case/market data). Data, code and supplementary materials are posted at a linked GitHub repo (andrewtlevin.github.io/bayesian-analysis-of-covid-origins).
relevance_note: shows the debate’s Bayesian-analysis genre continuing well past 2024 and reaching a starkly different conclusion using a similar method on overlapping (largely epidemiological/spatiotemporal) data — a load-bearing data point for how much of the “23 OOM” spread is about inputs vs. inference method.