jmis-tables-figures
GitHub针对JMIS论文，优化表格与图表以在50页限制内清晰传达核心论点。要求图表自包含、报告经济效应而非仅显著性，并按证据类型规范展示内容，确保符合审稿预期。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill jmis-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "jmis-tables-figures",
"description": "Use when the exhibits are the bottleneck for a Journal of Management Information Systems (JMIS) manuscript — regression\/SEM tables, platform\/value figures, measurement-model exhibits, or artifact-evaluation charts that must carry the argument inside the 50-page budget. Finalizes exhibits and labeling; it does not run the analysis (jmis-data-analysis) or polish prose (jmis-writing-style)."
}
Tables and Figures (jmis-tables-figures)
When to trigger
- A reader cannot get the headline result from the tables/figures without the text
- Exhibits are dense, mislabeled, or report estimates without economic magnitude
- The exhibit count is pushing the manuscript over the 50-page ceiling
- Survey papers lack the measurement-model exhibits reviewers expect (reliability/validity, loadings, HTMT)
Make each exhibit answer one question
A JMIS exhibit should be readable on its own: a self-contained title, defined variables and units, the sample and period, and a note stating the estimator, clustering, and what the stars/SEs mean. The reader should be able to state the finding from the table alone. Order exhibits to track the argument — descriptive/measurement first, main result next, mechanism and robustness after.
Report magnitudes, not just significance
JMIS values managerial relevance, so a coefficient table that shows only significance underdelivers. Report economic magnitude — marginal effects, elasticities, dollar value, or lift — alongside (or instead of) raw coefficients, and put the number that matters in the table the reader will actually look at. Report standard errors (and the clustering level) clearly; a figure with confidence intervals often communicates an effect better than a wall of asterisks.
Exhibits by evidence type
| Evidence type | Exhibits reviewers expect |
|---|---|
| IT-value / platform econometrics | Descriptives + balance; main DiD/IV table; an event-study figure with leads for parallel-trends; robustness table; magnitude in interpretable units |
| Behavioral survey (SEM/PLS) | Measurement model (loadings, CR, AVE), discriminant validity (HTMT / Fornell-Larcker), structural-path diagram with coefficients and significance |
| Experiment | Cell means with CIs, manipulation-check results, treatment-effect figure, mediation/moderation exhibit |
| Design-science / ML | Benchmark table vs. credible baselines with uncertainty; ablation; a figure tying performance to the managerial decision |
| Analytical model | Comparative-static figures showing how the key outcome moves with the parameter that carries the insight |
Respect the page budget and the appendix boundary
The complete manuscript is capped at ≤50 pages (12pt, double-spaced), so exhibits compete with text for space. Keep in the body the exhibits that establish the contribution; move secondary robustness, full measurement batteries, and large parameter tables to the online appendix — but never let a load-bearing result live only in the appendix. Figures should be legible in grayscale and not rely on color alone. (检索于 2026-06；以官网为准.)
Worked example: turning a coefficient table into a JMIS exhibit (illustrative)
A draft Table 3 reports redesign × post = −0.062 (0.018)*** and nothing else. A JMIS reader cannot tell what that means. The fixed exhibit keeps the coefficient but adds a column translating it to the managerial scale (−6.2 pp seller retention; ≈N exits/quarter at the observed base), states the estimator and the clustering level in the note, and is paired with an event-study figure whose flat pre-period leads visibly support parallel trends. The reader now gets identification and magnitude from the exhibits without leaving for the prose — which is exactly what a method-literate, management-minded JMIS referee scans for first.
Sequence the exhibits to track the argument
Order is part of the argument: (1) descriptives/balance establish the sample and comparability; (2) the main effect table/figure delivers the headline with magnitude; (3) the mechanism exhibit shows why (mediation, heterogeneity, the channel); (4) robustness shows the threat-by-threat survival. A reader should be able to reconstruct the paper's logic from the exhibit sequence alone. Resist the temptation to front-load a dramatic figure before the reader can interpret it.
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. JMIS is empirical IS — survey-based SEM and econometric panels; the chain below serves causal / quasi-experimental designs and many-outcome corrections.
- 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
- Every exhibit is self-contained (title, variables, units, sample, period, note)
- The headline finding is readable from the main exhibits without the prose
- Economic magnitude (not just significance) is reported for the key effect
- SEs and clustering level are stated; CIs used where they communicate better
- Survey papers include measurement-model and discriminant-validity exhibits
- Empirical papers include an event-study/leads figure where parallel trends matter
- Body holds the contribution-establishing exhibits; appendix holds only support; ≤50pp respected
- Exhibits numbered consecutively, captioned by content, and called out in order
- No exhibit de-anonymizes the author or proprietary data source in the blind file
Number and reference exhibits for JMIS's house style
JMIS uses numbered bracketed citations [n] with an alphabetized reference list, so any in-text reference to a source inside a table note should use the same numbered form, not author-date. Number tables and figures consecutively, give each a caption that names what it shows (not just "Regression results"), and make sure every exhibit is called out in the text in order. Because the manuscript is double-anonymized, scrub identifying content from figures too — institution logos, internal dashboard screenshots, or proprietary platform names that reveal the author or data source belong out of the blind file or suitably masked. (检索于 2026-06；以官网为准.)
Anti-patterns
- A coefficient table with stars but no magnitude or interpretation
- Exhibits the reader cannot parse without hunting through the text
- A load-bearing result relegated to the online appendix
- A staggered-DID paper with no event-study/leads figure
- A survey paper missing discriminant-validity evidence
- Color-only figures that fail in grayscale
- A "kitchen-sink" main table with no marked preferred specification
- Author-date references in a table note when JMIS uses numbered brackets
- Figures or screenshots that de-anonymize the author or proprietary data source in a blind file
Common exhibit defects JMIS referees flag
- A "kitchen-sink" main table with twenty controls and no marked specification — highlight the preferred column and relegate the rest.
- Stars without a scale — a reader cannot tell whether
0.06***is large; always pair significance with an interpretable magnitude. - A staggered-DID paper with only a single pooled coefficient — show the dynamic event-study so pre-trends and the post-treatment path are visible.
- A survey paper whose constructs overlap — without an HTMT/Fornell-Larcker exhibit, reviewers assume discriminant validity fails.
- Figures that die in print — JMIS appears in print, so a color-coded figure that becomes unreadable in grayscale will draw a production comment; encode with line style or labels too.
- Notes that omit the estimator or clustering — the reader must be able to reconstruct what produced the numbers from the note alone.
Output format
【Headline exhibit】reads alone? magnitude shown? [Y/N]
【Diagnostics shown】event-study/leads | measurement model + HTMT | benchmarks + uncertainty
【Inference labeling】SEs + clustering stated; CIs where clearer [Y/N]
【Body vs. appendix】contribution exhibits in body; ≤50pp respected [Y/N]
【Next step】jmis-writing-style
Version History
- 1839142 Current 2026-07-05 13:46
