#collider bias

Causal Inference From Observational Data: Draw the DAG Before Adding Controls

“Control for as many variables as possible” is bad advice. Some variables must be controlled, others must never be. This guide uses causal diagrams to separate