From restat-skills
Designs publication-grade tables and figures for REStat economics manuscripts, ensuring the headline estimate is legible from one exhibit with standard errors, self-contained notes, and consistent house style.
How this skill is triggered — by the user, by Claude, or both
Slash command
/restat-skills:restat-tables-figuresThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- The main result is buried across several dense tables instead of one clear exhibit
A REStat reader should grasp the headline estimate from one main exhibit — a table or, increasingly, an event-study / RD / coefficient-stability figure with a confidence band. REStat follows standard economics house style: significance stars are permitted, but standard errors are expected in parentheses and the exhibit must stand on its own. Every table and figure carries a self-contained note stating the sample, the specification, the unit of observation, the inference (clustering level), and what the reader should take away. The main paper is self-contained; supporting detail goes to the online appendix (≤20 pages; source map refreshed 2026-06-20).
| Element | REStat-compliant practice |
|---|---|
| Headline result | One main table or one figure a reader can read in 30 seconds |
| Standard errors | In parentheses below coefficients; clustering stated in the note |
| Significance | Stars allowed, but never instead of SEs; report the SE always |
| Notes | Self-contained: sample, spec, unit, N, inference, source |
| Figures | Load-bearing (event study with leads, RD plot, coefficient stability) with CIs/bands |
| Decimals | Consistent, sensible precision; no spurious digits |
| Appendix | Robustness and construction detail → online appendix (≤20 pp); main paper self-contained |
| Reproducibility | Every number traces to a script in the Harvard Dataverse package (restat-replication-package) |
A staggered-DID paper had its main result spread over three tables: a TWFE column, a Callaway–Sant'Anna column, and a robustness table. A REStat editor wanted to see the identification at a glance. The fix collapsed the headline into a single event-study figure — event-time on the x-axis, the estimated effect with a 95% confidence band on the y-axis, leads visibly flat and centered on zero, lags showing the dynamic effect. The accompanying table kept only the pooled post estimate with its SE and clustering note. A reader now sees parallel-trends support and the effect in one exhibit; the three old tables moved to the online appendix. Self-contained note states sample, estimator, unit, N, and clustering — nothing to hunt for in the text.
【Headline exhibit】Table/Figure [n]: [what it shows in one line]
【SE + inference】SEs in parentheses; clustering at [level]; stars: [used/not]
【Figure load-bearing?】[event-study / RD plot / validation / coef-stability] with CIs? [Y/N]
【Notes self-contained?】sample/spec/unit/N/inference/source present? [Y/N]
【Main vs appendix】main exhibits: [...]; online appendix (≤20pp): [...]
【Reproducible?】each number ← script in Dataverse package? [Y/N]
【Next step】restat-writing-style
npx claudepluginhub brycewang-stanford/awesome-journal-skills --plugin restat-skillsMakes tables and figures in REStud manuscripts publication-grade: fixes oversized regression tables, unclear figures, and inconsistent formatting. Does not run analysis.
Builds EER-compliant tables reporting standard errors, R², and N; designs causal figures with confidence bands and self-contained captions. Activated when exhibits lack SEs, are cluttered, or need Elsevier economics house style.
Formats AEJ: Applied manuscript exhibits (tables and figures) to highlight main causal estimates with AEA-style compliance, including event-study plots, RD figures, and balance tables.