Theory 94/Selection & Bias/Causal inference
Conditioning Creates Associations
The claim
Controlling for a common effect of two variables induces a spurious association between them, so adjusting for the wrong variable can create a relationship that does not exist.
01Mechanism
Conditioning on a collider selects a subpopulation in which the two causes trade off, producing correlation.
02Prediction
Associations should appear after adjustment that are absent unadjusted.
03Refutation
Adjustment never creating associations would remove the collider effect.