Collider bias is a common threat to internal validity in clinical research but is rarely mentioned in informatics education or literature. Conditioning on a collider, which is a variable that is the shared causal descendant of an exposure and outcome, may result in spurious associations between the exposure and outcome. |