dowhy

Why we included this project

DoWhy is the library to reach for when your analysis question is about cause rather than correlation, such as whether a button color change actually lifts purchases or which service in a distributed system is slowing the frontend. It wraps causal inference in a four-step workflow (model, identify, estimate, refute) that covers backdoor, frontdoor, and instrumental-variable settings, and it integrates with EconML for conditional treatment effects. The refutation API is what makes it stand out for teams: it actively challenges the assumptions behind an estimate instead of trusting them, so non-experts can sanity-check results before acting. Beyond effect estimation, the library also supports root-cause analysis, counterfactual what-if questions, and causal structure learning through graphical causal models. Data scientists and analysts who need defensible answers to intervention questions will find it a practical, well-documented starting point.