What problem does it solve? Answering "what is the effect of treatment T on outcome Y" requires choosing a defensible identification strategy, cleaning messy uploaded data, running the right estimator, and interpreting results with correct standard errors. This Skill packages 17 causal-inference estimators plus method-selection guidance, data diagnostics, and reporting templates so AI coding agents can run applied econometric analyses end to end. ## Core Features & Use Cases - 17 estimators across five method families: OLS with controls, propensity score methods (PSM, IPW, AIPW, IPW-RA), IV/2SLS with diagnostics, static and staggered DID with event studies, and sharp/fuzzy RDD. - Automatic data diagnostics: analyze_dataset() profiles uploaded CSV/Excel files, flags missingness, duplicates, outliers, and candidate outcome/treatment/unit/time columns, then produces cleaning recommendations. - Research-grade workflow support: references cover method selection, identification memos, diagnostic checklists, robustness grids, heterogeneity plans, and result-table generation. - Use Case: A user uploads a state-year panel dataset and asks whether a minimum-wage reform affected employment. The Skill diagnoses the data, sets up the (entity, time) MultiIndex, runs a staggered DID with cluster-robust standard errors, checks pre-trends via an event study, and reports the ATT with caveats. ## Quick Start Ask the agent to estimate the causal effect of a treatment on an outcome in your uploaded CSV or Excel file, for example: "Use the econometrics skill to run a difference-in-differences analysis of the policy reform on employment using my panel data."
