StatsPAI_skill

Perform end-to-end causal inference by importing a unified API that takes an analysis-ready DataFrame and returns structured result objects with diagnostics and robustness outputs, covering everything from balance checks to estimation.

The best way to estimate treatment effects is using an estimand-first decisioning approach that formalizes your research question and guides estimator selection. You can apply methods like DID, RD, or IV through a structured workflow that includes diagnostics and robustness checks.

Yes, ML-based causal inference is supported alongside traditional econometrics. You can apply Double Machine Learning, causal forests, meta-learners, and TMLE while using LLM-assisted DAG discovery to inform identification reasoning.

Yes, Python supports staggered adoption designs and synthetic control methods for panel data. The workflow includes staggerated-DID workflows, Propensity Score Matching, and Synthetic Control Method estimation with structured diagnostics and robustness outputs.

You validate causal DAGs and specification choices through pre-flight diagnostics and LLM-assisted DAG discovery. The workflow proposes, validates, and constrains causal DAGs to inform identification reasoning before running the final estimations and robustness checks.

An estimand-first research question DSL is a domain-specific language used to formalize population, treatment, outcome, estimand, and design choices. It guides estimator selection, such as deciding between DID vs RD vs IV, before running the actual estimation.