red-tables-figures
GitHub专为《经济动态评论》(RED) 论文设计，生成脉冲响应、校准表、矩匹配表及政策实验图等定量展示内容。强调机制清晰、模型与数据拟合可见，确保图表自包含且符合印刷分辨率要求。
触发场景
安装
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill red-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "red-tables-figures",
"description": "Use when designing the exhibits of a Review of Economic Dynamics (RED) manuscript — impulse-response functions, calibration tables, moment-matching (model-vs-data) tables, policy-experiment figures, and accuracy diagnostics — so a quantitative dynamic paper communicates its mechanism and fit clearly to the SED readership."
}
Tables & Figures for RED (red-tables-figures)
When to trigger
- Building the exhibits that carry a quantitative dynamic paper
- Deciding how to show model-vs-data fit and the mechanism behind a result
- Making IRFs and policy experiments legible at print resolution
Exhibits that fit RED
RED papers live or die on whether the mechanism and the quantitative fit are visible. Favor:
- Impulse-response functions (IRFs) — the workhorse figure for dynamic models; plot model responses to the key shocks, overlay alternative parameterizations to isolate the mechanism, and (where relevant) overlay empirical responses.
- Calibration table — every parameter, value, source/target, and status (calibrated/estimated/assumed), so a reader can audit discipline at a glance.
- Moment-matching table — targeted and untargeted moments side by side, model vs data; untargeted fit is a credibility highlight, so give it prominence.
- Policy / counterfactual experiment figures — show the magnitude of the dynamic effect and the transition path, not just steady-state comparisons.
- Accuracy diagnostics — Euler-equation errors or convergence plots where the computation is non-trivial.
Make each exhibit self-contained: numbered, called out in order, with notes stating the model variant, shock, units, and data source. Author-year citations in notes match RED's reference system.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. RED is quantitative macro — mostly structural/calibration, which is outside this causal-inference toolchain; apply the chain to its empirical/reduced-form papers.
- 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
- IRFs / transition paths show the mechanism, not just outcomes
- Calibration and moment tables let a reader audit discipline and fit
- Untargeted moments are reported, not hidden
- Every exhibit is self-contained, numbered, and legible at print resolution
Anti-patterns
- Steady-state-only comparisons that hide the dynamics
- Moment tables showing only targeted moments
- Figures with no notes on model variant, shock, or units
Exhibit sequence
For most RED papers, the exhibit order should mirror the argument:
- Mechanism figure: state variables or transition paths that show the dynamic force.
- Discipline table: parameters, calibration targets, estimation targets, and sources.
- Fit table: targeted and untargeted moments, model vs data.
- Counterfactual/experiment: the quantitative result or policy path.
- Accuracy/reproducibility note: solver error, convergence, seed, or runtime when computation matters.
If the first exhibit is a static regression table, ask whether the paper is really using RED's dynamic lens or drifting toward a field journal.
Mock moment table (model vs data)
The fit table referees scan first should look like this (all numbers illustrative):
Table X: Targeted and untargeted moments
Data Model Source
Targeted
Wealth-to-income ratio 2.90 2.90 NIPA / Flow of Funds
Share with negative net worth 0.135 0.132 SCF
Untargeted
Wealth Gini 0.78 0.74 SCF
Top-10% wealth share 0.71 0.66 SCF
Average quarterly MPC 0.16 0.19 literature estimates (author-year)
Notes: model moments from the stationary distribution; simulation of 100,000
households; seed recorded in the archive readme.
The Targeted/Untargeted split must be typographically explicit — a single undifferentiated column invites the circularity objection from quantitative-macro referees.
Exhibit-budget anchor
Hedged anchor (verify against recent RED issues rather than treating it as policy): mainline quantitative papers commonly carry on the order of 4–8 main-text exhibits — typically one mechanism/IRF figure, a calibration table, a fit table, and one or two counterfactual exhibits — with accuracy diagnostics and extended sensitivity in an appendix. Treat that as a budget: each additional main-text exhibit must change what the reader believes about the mechanism or the magnitude.
Figure-craft pushback
| Exhibit complaint | Fix for a dynamics paper |
|---|---|
| IRFs without units or horizon labels | percent vs pp deviation, quarters vs years, stated on axes and in notes |
| Counterfactual shown only at the new steady state | add the transition path — adjustment dynamics are often the contribution |
| Mechanism overlays illegible in grayscale | vary line style, not only color; print-test the PDF |
| Distributional result collapsed to a mean | plot the policy function or the cross-sectional shift; heterogeneity is usually the point |
Supplementary resources
../../resources/external_tools.md— plotting and solver outputsred-data-analysis— the analysis behind the exhibits
版本历史
- 1839142 当前 2026-07-05 14:21
