Causal Backdoor Adjustment
Separate observational association from an intervention under an explicit causal model.
Free theory articleSeparate observational association from an intervention under an explicit causal model.
Concept & interpretation
When estimating the causal effect of an action (X) on a response (Y), simple observational association may be biased due to confounding variables (Z) that influence both the action and the response.
Experiment
Explore the interactive module to see how observational and interventional responses differ. Open the Causal Backdoor Adjustment module.
- Adjust the "Propensity at Z0" to 0.1 and "Propensity at Z1" to 0.9. Notice how the "Observed difference" becomes 0.56 due to confounding.
- Adjust the propensities both to 0.5. Notice how the "Observed difference" drops to 0, matching the true "SCM causal difference" of 0.