Restat Identification

Identification & Measurement Strategy (restat-identification)

When to trigger

• A causal claim rests on OLS + controls, or TWFE on staggered timing
• An IV's exclusion restriction or first-stage strength is contestable
• An RD's continuity / manipulation assumptions are not yet defended
• A shift-share / exposure design's exogeneity (shares vs shocks) is unargued
• The outcome or key regressor is measured with error, or you built a new measure/index

The REStat identification-and-measurement bar

REStat is applied econometrics with a measurement tradition, so two things are judged together: the mapping from data to the causal object must be explicit and defended, and the quality of measurement behind every variable must be credible. A clean design on a badly measured construct does not clear the bar; neither does a beautifully measured variable in a hopelessly confounded regression. Report standard errors and modern inference; clustering at the assignment level; address attenuation and other measurement-error biases head-on — REStat referees raise measurement objections sibling journals sometimes wave through.

Branch paths

Branch A: Difference-in-differences / event study

• With staggered adoption, move beyond TWFE (Callaway–Sant'Anna, Sun–Abraham, de Chaisemartin–D'Haultfœuille — the last has REStat-published estimators).
• Show a clean event-study with leads (flat pre-trends) and report a Goodman–Bacon decomposition.
• State the parallel-trends assumption and probe it (pre-trend tests + Rambachan–Roth honest bounds where relevant).

Branch B: Regression discontinuity

• McCrary / Cattaneo–Jansson–Ma density test for manipulation; covariate smoothness at the cutoff.
• Optimal bandwidth + bias-corrected, robust CIs; sensitivity to bandwidth and polynomial order.
• Fuzzy RD: report first stage; defend exclusion of the running variable's other channels.

Branch C: Instrumental variables

• Strong first stage (report effective F / Montiel-Olea–Pflueger); with weak instruments use Anderson–Rubin / weak-IV-robust sets.
• Defend the exclusion restriction in theory, institutions, and falsification tests.
• Shift-share / Bartik: argue exogeneity of shares (Goldsmith-Pinkham–Sorkin–Swift) or of shocks (Borusyak–Hull–Jaravel); report the implied just-identified estimates.

Branch D: Measurement (REStat signature)

• Construct validity: what does the measure actually capture; validate against an external benchmark.
• Measurement error: classical vs non-classical; attenuation correction, validation samples, or bounds.
• New index / data: document construction, sensitivity to choices, and show the applied conclusion is not an artifact of how you measured.

Execution bridge (StatsPAI / Stata MCP)

Estimate and audit the design, don't only describe it. Full map: execution-with-mcp. REStat is applied econometrics/empirical micro — the home of careful identification; DiD/IV/RDD with weak-IV-robust CIs.

• detect_design → recommend → fit with as_handle=true → audit_result.
• Observational causal claims: staggered DiD (callaway_santanna / sun_abraham + bacon_decomposition + honest_did_from_result); IV (effective_f_test + anderson_rubin_ci); RDD (rdrobust + mccrary_test).
• Experiments: randomization-based inference + romano_wolf for many-outcome control.
• Sensitivity: oster_delta / sensemakr for observational claims.

Report the magnitude in interpretable units; route the full battery to the appendix. A run end-to-end (synthetic data, real returns) is in the JF execution walkthrough.