Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› joe-tables-figures
joe-tables-figures
GitHub用于为《计量经济学杂志》稿件生成蒙特卡洛模拟表格和理论图示。规范大小/功效表结构，确保图表清晰展示渐近与有限样本行为，符合期刊单PDF提交及排版要求。
Trigger Scenarios
蒙特卡洛结果存在但表格密集、无标签或难以比较方法时
大小和功效混合导致失真不可见时
图表未显示定理预测的行为（如速率、覆盖率）时
格式化仅支持PDF的初始投稿展品以确保打印可读时
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill joe-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "joe-tables-figures",
"description": "Use when building the Monte Carlo tables and theory-illustrating figures for a Journal of Econometrics (JoE) manuscript. Covers size\/power table conventions, estimator-comparison layout, and figures that make asymptotic and finite-sample behavior legible."
}
Tables & Figures (joe-tables-figures)
When to trigger
- Monte Carlo results exist but the tables are dense, unlabeled, or hard to compare across methods
- Size and power are mixed into one block so distortion is invisible
- Figures show output but not the behavior the theorems predict (rates, coverage, distributions)
- You are formatting exhibits for a PDF-only initial submission and want them print-legible
What JoE exhibits must do
At the Journal of Econometrics the exhibits are the empirical backbone of a methods paper: they have to let a referee verify, at a glance, that the estimator is well-behaved and the test controls size. Initial submission is a single PDF (~40 pages, ≥1.5 spacing, 11pt), so tables and figures must be readable inline at that density; structured formatting and source files are handled at revision/acceptance. Build self-contained exhibits whose notes state the DGP, sample sizes, replication count, and nominal level.
Monte Carlo table conventions
- Separate size from power. Report empirical size at nominal 5%/10% in its own panel; report size-adjusted power separately so a size-distorted test cannot masquerade as powerful.
- One method per column (or panel), one DGP/$n$ per row block. The reader should compare your method to the nearest competitor down a column without hunting.
- Report bias, RMSE, and CI coverage for estimators; mark the best in each row if it aids reading.
- Notes are mandatory: DGP description, error distribution, dependence structure, $n$, replication count, and any tuning values. A referee should reproduce the table's meaning from the note alone.
- Keep decimal places consistent; do not over-report precision that the replication count cannot support.
Figures that illustrate theory
- Size/power curves across the parameter; coverage vs. $n$ to show asymptotics kicking in; QQ plots or densities of the standardized statistic against its limiting distribution.
- Sensitivity plots over tuning parameters (bandwidth, lag length, penalty) to show robustness.
- Plot confidence bands / Monte Carlo error; avoid chartjunk (no 3D, minimal color, legible at print size).
- Use vector output (PDF/EPS) so exhibits stay crisp; ensure they survive grayscale printing.
Formatting notes
- Number tables and figures, call them out in order, and give each a self-contained caption.
- Math in captions/notes in LaTeX via the elsarticle class for consistency with the manuscript.
- At submission everything lives in the single PDF; keep source and high-res files staged for acceptance.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. Journal of Econometrics is a methods venue — estimator validity + simulation evidence are the contribution; pair estimates with diagnostics and Monte-Carlo where relevant.
- 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.
Anti-patterns
- A 12-column table with no row/column grouping and no notes
- Power reported without empirical size (size-distorted power is meaningless)
- Tuning-parameter sensitivity omitted, hiding fragility
- Figures with default software styling, illegible in grayscale, or missing Monte Carlo error
- Over-precise decimals implying accuracy the replication count cannot deliver
Output format
【Size table】separate panel, nominal 5%/10%? [Y/N]
【Power table】size-adjusted, vs. nearest method? [Y/N]
【Estimator table】bias / RMSE / coverage? [Y/N]
【Notes】DGP / n / reps / tuning stated? [Y/N]
【Theory figures】coverage-vs-n / power curves / limiting-dist check? [list]
【Print quality】vector, grayscale-safe, legible in single PDF? [Y/N]
【Next step】joe-writing-style
Version History
- 1839142 Current 2026-07-05 13:32
