The DAG Almanac

The DAG Almanac


“Correlation ≠ causation”

We all know that correlation isn’t causation. But in practice, we actually really care about causal effects! Does a social program work? Does a vaccine improve public health? Does a policy have an effect on poverty?

When we can’t do experiments with social phenomena, directed acyclic graphs (DAGs) let us visually map out the causal path between a treatment and an outcome, and do-calculus provides logic for extracting causal estimates from observational data.


Interactive visualizations

Guides

do-calculus

How to convert interventions to observations with the rules of do‑calculus

How to adjust

How to implement backdoor and front-door adjustment with R