When causes act years before symptoms and through mediators, contemporaneous correlations between an upstream marker and late outcomes can look weak or null, inviting erroneous rejection of true causal links.
— This cautions policymakers and media against dismissing causal models based on cross‑sectional snapshots in health, economics, and climate; it emphasizes modeling of timing, thresholds, and mediation over naive correlations.
Scott Alexander
2025.08.14
100% relevant
Critics note poor cross‑sectional correlation between amyloid load and cognitive impairment; the article explains that amyloid typically accumulates 15+ years earlier and acts via tau and neurodegeneration in the ATN cascade.
← Back to All Ideas