DoWhy

DoWhy is the Python library for end-to-end causal inference, enforcing a structured, four-step analysis that makes assumptions explicit.

DoWhy is the definitive Python library for end-to-end causal inference, originally developed by Microsoft Research and now part of the PyWhy ecosystem. It mandates a structured, four-step process—Model, Identify, Estimate, Refute—to ensure rigor in your analysis. The library explicitly models causal assumptions via a Causal Graph (DAG) and separates identification from estimation. Crucially, its state-of-the-art Refutation API automatically runs robustness checks (e.g., placebo treatment tests) to validate estimates, a vital step often overlooked in traditional methods. It supports multiple identification methods (backdoor, frontdoor, instrumental variable) and integrates with libraries like EconML for Conditional Average Treatment Effects (CATE).

https://pywhy.org/dowhy/