Association vs causation

Two variables can move together without one causing the other. Causal questions ask what happens if you intervene.

Bullet holes on bombers

Planes that returned had more holes in the fuselage than the engines. Naïve conclusion: armour the fuselage. Causal insight: planes hit in the engine often did not return (survivorship bias). You should armour the engines.

Association is what data shows. Causation is about the effect of changing $x$ on $y$. Mixing them up leads to bad decisions.

Key takeaways

Association ≠ causation.
Causal claims need design or assumptions.
Wrong causal reads ⇒ wrong interventions.