confounders
Third variables (like income or education) that distort an apparent causal relationship between two observed variables.
As defined in Peter Attia on Lifespan, Healthspan, and Outlive
Where it comes up · 2
- Peter Attia on Lifespan, Healthspan, and Outlive2 claims
- Tyler Cowen on the Pandemic, Revisited1 claim
Most important claims · 3 of 3Search all claims for “confounders” →
- Naive cross-country and cross-state COVID comparisons are uninformative because they fail to control for confounders like age, population density, and obesity, and because lockdown is not a 0-1 variable (except perhaps in China) given widespread voluntary self-quarantining—so the data are genuinely messy and resist simple causal inference.Tyler Cowen on the Pandemic, Revisited
- Unlike nutritional epidemiology, exercise epidemiology survives the Bradford Hill criteria: all data point the same direction, there are large consistent hazard ratios with clear dose effects across VO2 max levels in million-person studies, biologic plausibility, and reproducibility in randomized controlled trials where health benefits track the intervention—so confidence in exercise's causal benefit is far higher even though one cannot impute precise lifespan gains.Peter Attia on Lifespan, Healthspan, and Outlive
- The confounder objection to VO2 max is partly answered by the fact that VO2 max is measured in liters per kilogram per minute and thus already normalized by weight; the remaining issue that fitness is a proxy for general health applies equally to the smoking, diabetes, and renal-disease comparisons, so the relative hazard-ratio changes remain informative.Peter Attia on Lifespan, Healthspan, and Outlive