Correlation and causation
Correlation means two variables move together. It does not prove one causes the other. A third factor (a confounder) may drive both, or the link may be chance.
Last updated
Analogy
Ice cream sales and sunburn rise together. Hot weather causes both. Ice cream does not cause sunburn.
Example
Summer heat raises ice cream salesSummer heat raises sunburn casesSales and sunburn correlateBut neither causes the other
In practice
Only a randomised experiment, where chance assigns who gets the treatment, can support a cause claim.
Practice
Flashcards
- What is a confounder?A third variable that influences both the cause and the effect.
Interview questions
- A study finds that cities with more ice cream shops have more sunburn. Explain why this does not mean ice cream causes sunburn, and how you could find out.
- Correlation does not prove cause
- A confounder such as hot weather drives both
- Randomised experiments can test cause
- Regression can adjust for measured confounders
Questions people ask
How can I tell if a correlation is causal?
Look for a randomised experiment, a plausible mechanism, a consistent result across studies, and a cause that comes before the effect. Rule out confounders. Observational data alone rarely settles causation, however strong the correlation looks, and a fancy model cannot fix a weak design.
What is a confounding variable?
A confounder is a third variable linked to both the supposed cause and the effect, creating a misleading link. Hot weather confounds ice cream and sunburn. Randomised experiments spread confounders evenly across groups, while regression can adjust for those you measured.