Require within‑person causal checks

When longitudinal cohort data include repeated measures, authors should routinely run within‑person (panel) and family fixed‑effects checks before making causal claims; these designs use each person or sibling as their own control and remove many unobserved confounders. High‑profile papers and press releases should report those robustness checks explicitly so policymakers and journalists don't treat conditional between‑person associations as causal. — If adopted as a norm, this would reduce misleading causal claims in science that drive policy and public panic—especially on hot topics like youth social‑media harms.

Sources

Making an Aging Clock

Cremieux 2026.11.03 87% relevant

The article’s core claim is that many aging clocks (exemplified by LinAge2) may be risk scores driven largely by chronological age and cohort structure rather than modulable biological processes; that directly maps to the existing idea that interventions and claims about biological-age measures need within-person (longitudinal/causal) validation before being used as targets — the author cites NHANES-based train/test design, AUC comparisons, and the planned Singapore clinical deployment as concrete reasons why cross‑sectional performance is insufficient.

Social Scientists Are Lazy

Cremieux 2026.03.25 100% relevant

Cremieux's reanalysis of Nagata et al.'s ABCD study (Lancet Regional Health Americas) and his note that authors could have used within‑person and sibling fixed‑effects but instead reported group‑trajectory between‑person estimates.