ecj-tables-figures
GitHub用于优化《经济杂志》稿件中的表格与图表，确保其自包含、可读且传达核心经济结论。规范设计标准，强调展示经济幅度而非仅统计显著性，并指导主文与附录的取舍，确保数据与代码复现一致。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill ecj-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "ecj-tables-figures",
"description": "Use when finalizing the main exhibits (tables and figures) of a The Economic Journal (EJ) manuscript so each one carries economic content, reads to a general audience, and is self-contained. Sets exhibit design and notes standards; it does not decide which results to report (see ecj-robustness)."
}
Tables & Figures (ecj-tables-figures)
When to trigger
- Tables are overloaded with columns; the reader cannot find the economic point
- Figures are decorative rather than carrying the argument
- Notes are thin — a reader cannot reconstruct units, sample, or specification from the exhibit alone
- You are deciding what belongs in the main text vs. the online appendix, or working within the short-paper exhibit budget
The EJ exhibit philosophy
An EJ table or figure should be self-contained and legible to a generalist: a referee or non-specialist skimming exhibits should grasp the result, the magnitude, and what it means without hunting through the text. EJ's broad-interest, exposition-first identity makes this acute — an exhibit only your subfield can parse undersells the paper. Every exhibit must also be regenerated by the master script in the Zenodo replication deposit, so the numbers in the paper and the numbers the code produces must match exactly when the EJ Data Editor verifies them before final acceptance (see ecj-replication-package). The headline result usually deserves a figure (an event-study plot, a model-fit plot, a counterfactual) because magnitudes and dynamics read faster than a coefficient.
For a short paper (AER:Insights-style), the exhibit budget is tight (5 exhibits; verified 2026-06-20): every exhibit must earn its place; demote the rest to the online appendix.
Tables
- Report economic magnitudes, not just stars: effect relative to the mean, an elasticity, a standardized effect, or a monetary/welfare figure.
- One table = one idea. The main results table should fit on a page and be readable without the text.
- Columns build an argument (baseline → +controls → preferred → alternative design), not an undifferentiated grid.
- Standard errors in parentheses; state the clustering level in the note. Prefer reporting CIs or exact SEs over star-only inference.
- Self-contained note: sample, period, unit of observation, dependent-variable definition, specification, SE structure, and what each column adds.
Figures
- Use a figure to carry the main result when dynamics/magnitudes matter (event studies, dose-response, model fit, counterfactual paths).
- Plot confidence bands, not just point estimates. Label axes with units and economic meaning.
- Grayscale-safe and legible in print; no 3-D effects, no unnecessary gridlines.
- Each figure's caption states what the reader should conclude economically.
Main text vs. online appendix
- Main text: the identifying result, the mechanism evidence, the key magnitude, the headline counterfactual — typically a small number of exhibits.
- Online appendix: the full robustness battery, descriptive tables, secondary heterogeneity, and supporting derivations.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. The Economic Journal is general-interest economics; the DiD/IV/RDD chain serves its broad applied lane.
- Tables:
etable(multi-model) ordid_summary_to_latexstraight from theresult_id. - Figures:
plot_from_result/enhanced_event_study_plot/event_study_table— axis units and the SE/clustering note baked in. - Every note names the estimator + clustering and states the magnitude in interpretable units.
See a full fitted-result → exhibit chain in the JF execution walkthrough.
Checklist
- Headline result expressed as an economic magnitude (relative to mean / elasticity / welfare), not stars alone
- Main results readable without the surrounding text, by a non-specialist
- Every table note states sample, period, unit, DV definition, spec, and clustering level
- Columns/panels build an argument rather than dumping specifications
- At least one figure carries the main result with confidence bands
- Figures legible in grayscale; axes labeled with units and meaning
- Exhibit count disciplined (short paper: within the 5-exhibit budget); heavy material in the online appendix
- Every exhibit number reproduces from the deposited code; in-text references match exactly
Anti-patterns
- Star-only tables with no economic magnitude — a referee cannot tell if the effect matters
- Numbers in an exhibit that the deposited code does not reproduce — fatal at EJ Data Editor verification
- A 12-column main table that no one outside the subfield can parse
- Figures with no confidence bands, or with chartjunk / 3-D / rainbow palettes
- Notes too thin to reconstruct the specification
- Reporting R² and stars but never the units of the dependent variable
- Blowing the short-paper exhibit budget instead of using the online appendix
Output format
【Main exhibits】tables: N, figures: N (short-paper budget respected? [y/n])
【Magnitude reported】mean-relative / elasticity / welfare / standardized
【Headline figure】what it shows + has CIs? [y/n]
【Notes complete】sample/period/unit/DV/spec/clustering present? [y/n]
【Reproduces from code】numbers match deposit? [y/n]
【Main vs. appendix split】...
【Next】ecj-writing-style
Version History
- 1839142 Current 2026-07-05 14:31
