Backdoor Criterion

The backdoor criterion is a rule, read off a causal DAG, that tells you exactly which set of variables to adjust for in order to estimate a causal effect without bias. A “backdoor path” is a non-causal route from treatment to outcome — typically through a confounder — that creates spurious association.

If you can find a set of variables that blocks every backdoor path while leaving the real causal paths open (and without opening a collider), adjusting for that set lets you recover the causal effect from observational data. It is the formal answer to the everyday question “what should I control for?” — and it warns you that controlling for everything can do harm.