Tip: Click any non-T/Y node to toggle Conditioning (Adjustment). Drag nodes to reposition.
T (Treatment)
Y (Outcome)
Conditioned {Z}
True Causal ATE (β)
1.50
Naive Correl. (Observed)
3.48
Adjusted ATE Estimate
1.52
Identifiability Status
IDENTIFIED
⚠️ Collider Stratification Bias Active: Conditioning on a collider or its descendant opens a spurious non-causal path, inducing artificial association (Berkson's Fallacy).
Backdoor Paths & D-Separation
Minimal Sufficient Adjustment Sets
{ U }
Live SCM Monte Carlo (N=1000)
Observational: E[Y | T=1] - E[Y | T=0] = +3.48
Interventional: E[Y | do(T=1)] - E[Y | do(T=0)] = +1.50
Adjusted: E_Z[ E[Y | T=1, Z] - E[Y | T=0, Z] ] = +1.52