Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› tar-tables-figures
tar-tables-figures
GitHub用于构建和清理《The Accounting Review》稿件中的图表，包括描述性统计、回归表及各类分析图。遵循芝加哥格式与55页限制，确保图表自包含并体现经济显著性，不执行数据分析。
Trigger Scenarios
表格杂乱或不一致
需要标准存档表格集
需呈现DiD或RDD图形
图表数量接近55页上限
审稿人反馈无法阅读
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill tar-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "tar-tables-figures",
"description": "Use when building or cleaning the exhibits for a The Accounting Review (TAR) manuscript — descriptive, correlation, and regression tables, event-study and discontinuity figures, experimental cell means, and analytical-model exhibits in Chicago house style and within the page budget. Finalizes exhibits; it does not run the analysis (tar-data-analysis) or polish prose (tar-writing-style)."
}
Tables & Figures (tar-tables-figures)
When to trigger
- Tables are cluttered, inconsistent, or not self-explanatory without the text
- You need the standard archival table set (descriptives, correlations, main regression)
- You are presenting a DiD/event study or RDD and need the right figure
- The exhibit count is pushing the 55-page initial limit
- A reviewer says "I cannot read the table" or "what is the economic magnitude?"
TAR exhibit conventions
TAR follows The Chicago Manual of Style (16th ed.) for citations and references and the AAA Manuscript Preparation Guide for layout (12-pt Times New Roman, double-spaced, 1-inch margins, serially numbered pages). Remember the 55-page initial limit includes tables, figures, and appendices — so every exhibit must earn its space; move secondary results to an online appendix. Each table must be self-contained: a reader should understand it without the text.
The standard archival table set
- Sample selection — a stepwise screen from the raw database population to the final N, so the sample is reproducible (this also supports the data-authenticity requirement).
- Descriptive statistics — N, mean, median, SD, and key percentiles for all variables; note winsorization. Define every variable in a clearly labeled variable-definitions appendix.
- Correlation matrix — Pearson (and often Spearman) correlations; flag significance.
- Main regression — coefficients with standard errors (or t-stats) below, fixed effects and clustering noted in the table, R-squared/N reported. State the dependent variable in the title.
- Cross-sectional / robustness tables — partitions and alternative specifications.
Report economic magnitude in or alongside the table, not just significance stars; state which standard-error and clustering choice produced the reported inference.
Figures by design
- Event study / DiD: plot event-time coefficients with confidence intervals to show pre-trends and dynamics — often the single most persuasive exhibit for identification.
- RDD: binned scatter with the fitted discontinuity at the threshold.
- Experiment: cell means with error bars for the manipulated factors; show the interaction.
- Analytical model: a figure of the comparative static (how the equilibrium quantity moves in a parameter), and a clearly labeled timeline of the information structure.
House-style hygiene
- Consistent decimal places; aligned columns; explicit units.
- Significance convention stated once (e.g., // with the tail) and used everywhere.
- Variable names in tables match the text and the definitions appendix exactly.
- Notes beneath each table define abbreviations, sample, period, and SE/clustering.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers (the usual source of
body-vs-appendix drift). Full map: execution-with-mcp. TAR is archival accounting — DiD around regulation / standard changes, IV, and earnings-based designs; the corporate-causal chain fits directly.
- Tables:
etable(multi-model columns) 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 effect size in interpretable units.
See a full fitted-result → exhibit chain in the JF execution walkthrough.
Checklist
- Sample-selection table makes the N reproducible
- Descriptives, correlations, and main regression follow the standard archival set
- Fixed effects, clustering, N, and fit statistics reported in every regression table
- Economic magnitude shown, not only significance stars
- DiD/RDD/experiment/model figures match the design
- Variable names consistent across tables, text, and the definitions appendix
- Exhibit set fits the 55-page budget; secondary results moved online
Anti-patterns
- Stars-only tables with no economic magnitude or SE/clustering note.
- Orphan exhibits that cannot be read without hunting through the text.
- Inconsistent variable names between tables and the definitions appendix.
- A DiD with no event-time figure to show pre-trends.
- Page-budget blowout from leaving every robustness table in the main body.
Output format
【Table set】sample-selection / descriptives / correlations / main / robustness — complete? yes/no
【Regression tables】FE, clustering, N, fit reported? magnitude shown? yes/no
【Key figure】event-study / RDD / cell-means / comparative-static — matches design? yes/no
【Variable definitions】appendix present and consistent? yes/no
【Page budget】within 55 pages incl. exhibits? overflow → online appendix
【Next step】tar-writing-style
Version History
- 1839142 Current 2026-07-05 14:29
