jpam-tables-figures
GitHub为JPAM论文设计自包含、决策可读的表格与图表，展示政策效应、不确定性、有效性及分配结果。不执行估计，侧重呈现标准、可信度验证及可视化规范，确保非专业读者也能理解核心结论。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill jpam-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "jpam-tables-figures",
"description": "Use when building tables and figures for a Journal of Policy Analysis and Management (JPAM) manuscript — self-contained, decision-legible exhibits that show the policy effect, its uncertainty, the design's validity, and cost-benefit \/ distributional results. Designs exhibits; it does not run the estimation."
}
Tables & Figures (jpam-tables-figures)
JPAM exhibits serve a mixed audience — economists, political scientists, public-management scholars, and practitioners — so they must be self-contained and decision-legible: a policymaker should grasp the main result, its uncertainty, and who it affects without reading the methods section. Lead with the exhibit that shows the policy effect and its credibility, not a wall of coefficients.
When to trigger
- Designing the main results table/figure and the design-validity exhibits
- A reviewer found tables unreadable, or the key result hard to locate
- Presenting cost-benefit or distributional results visually
- Preparing exhibits for the (double-blind) submission
What the exhibit set should contain
- The headline effect, clearly. A main results table or a coefficient/effect figure in policy- relevant units, with confidence intervals — not just significance stars.
- Design-validity exhibits. The evidence that the identification holds: an event-study / pre-trends plot for DiD, an RD plot with binned means and the fitted discontinuity, a balance table for an RCT, a synthetic-control fit plot. These often persuade reviewers more than the point estimate.
- Heterogeneity / mechanism. A figure showing effects by the theory-driven subgroups.
- Cost-benefit / distributional. Where central, an exhibit that shows the benefit-cost result and its sensitivity, or the distribution of gains and costs across groups.
Craft standards
- Self-contained captions: define the sample, the estimator, the units, the inference (what the error bars/SEs are and the clustering), and the time window — readable without the text.
- Confidence intervals over stars in figures; report SEs and the clustering level in tables.
- Policy-relevant units on axes and in cells (dollars, percentage points, per-recipient).
- Accessible design: colorblind-safe palettes, legible in grayscale, vector output for print.
- Honest scaling: do not truncate axes to exaggerate an effect; show the zero line where relevant.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. JPAM is policy analysis — program evaluation is the core; DiD/IV/RDD and the policy-relevant magnitude are decisive.
- 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
- Main effect in policy-relevant units with CIs, locatable at a glance
- A design-validity exhibit (event-study / RD plot / balance / SC fit) included
- Heterogeneity exhibit matches the pre-specified subgroups
- Cost-benefit / distributional result shown where it is central
- Captions self-contained: sample, estimator, units, inference, window
- Colorblind-safe, grayscale-legible, vector format
- Every exhibit number/value matches the deposited replication output
Anti-patterns
- A dense regression table with stars and no confidence intervals or units
- Hiding the parallel-trends / RD-validity evidence in an appendix the reviewer must hunt for
- Captions that require the methods section to interpret
- Truncated or rescaled axes that overstate the effect
- A cost-benefit conclusion in prose only, with no exhibit or sensitivity shown
- Exhibits whose numbers drift from the replication package
Calibration anchors (hedged)
- For a DiD or RD paper, the design-validity figure often does more persuasive work than the point estimate — a clean pre-trends or RD plot pre-empts the cross-disciplinary referee's first objection.
- A mixed APPAM audience reads exhibits before prose; if the headline effect and its uncertainty are not legible from the figure alone, the paper feels harder than it is.
- Confidence intervals communicate policy precision better than stars — a wide CI is itself information a decision-maker needs.
Worked micro-example (illustrative)
For a staggered-adoption DiD, the strong exhibit set is: (1) an event-study figure with confidence bands showing flat pre-trends and the post-policy effect; (2) a main table reporting the heterogeneity-robust estimate in dollars with the clustering level named; (3) a subgroup figure for the pre-specified populations; and (4) a benefit-cost panel with sensitivity bars. A reviewer can verify the identification, read the magnitude, and see the policy bottom line without leaving the figures. (Illustrative.)
Output format
【Headline exhibit】main effect + CI in policy units
【Design-validity exhibit】event-study / RD / balance / SC fit
【Heterogeneity / mechanism】subgroup figure
【Cost-benefit / distribution】exhibit + sensitivity (if central)
【Accessibility】colorblind-safe, grayscale, vector? [Y/N]
【Next】jpam-writing-style
Supplementary resources
../../resources/code/— event-study and RD plotting templates../../resources/external_tools.md— figure tooling (coefplot, ggplot2, marginaleffects)
Version History
- 1839142 Current 2026-07-05 13:53
