Spatial or proximity‑based epidemiological claims should routinely include placebo or 'landmark' falsification tests (e.g., test whether living near supermarkets, universities, or retailers shows similar associations) and report those results alongside primary estimates to reveal method artefacts. The failure to run or report such checks allows noisy methods to be dressed up as policy‑relevant causation and fuels media panic.
— If adopted as a reporting standard, these checks would sharply reduce false causal narratives in public health and make studies less likely to drive misguided policy or public fear.
Alex Tabarrok
2026.10.04
100% relevant
Tilson & Stein's demonstration that the Harvard papers' method would (spuriously) attribute over 2.2 million cancer deaths to Costco is the concrete example that motivates this proposal.
← Back to all ideas