DiD с поэтапным внедрением и HonestDiD

research

did-event-study is a Claude Code skill that runs staggered difference-in-differences and event-study analysis to the Sant'Anna practitioner standard, orchestrating the canonical packages R `did`, `DRDID`, `didFF`, `contdid`, and Stata `csdid`/`drdid` without reimplementing any estimator. The eight-phase workflow covers reproducibility setup (fixed seed, `renv::restore`, `here::here` paths), reshaping data to long format with correctly coded `gname` (never-treated = 0), doubly-robust estimation as the default, `aggte` aggregation, a mandatory diagnostic suite, and HonestDiD sensitivity analysis led by relative-magnitudes `Mbar`. Phase 8 delivers a graded credibility verdict rather than a binary pass/fail. The skill targets panel or repeated cross-section data with absorbing binary treatment or a continuous dose — the right choice whenever someone needs Callaway–Sant'Anna ATT estimates, `att_gt`, or a credible event study under staggered adoption.