Confounding and causal inference

A third variable associated with both exposure and outcome creates spurious association. Controlled via stratification, regression adjustment, matching, or causal-DAG-based methods (back-door adjustment, g-methods).