Causal Inference

Beyond Correlation

Correlation is the floor, not the ceiling. These tests probe direction, dose-response, lagged effects, and what would have happened under cleaner air.

Dose-Response Curve

Avg respiratory cases by PM2.5 quantile. A monotonic rise = the kind of curve epidemiologists expect for a real exposure-effect.

The exposure-effect signature

A clean monotonic rise is exactly what you'd expect if PM2.5 actually causes excess respiratory cases. A flat or jagged curve would have raised concerns about confounders.