brycewang-stanford SMR Tables and Figures
Use this to make exhibits carry the methodological argument. At SMR the central exhibits are usually a simulation results grid and a method-comparison table, plus diagnostic figures — not the descriptive-statistics tables of a substantive paper. Each must be readable without the surrounding text and must let a reviewer audit the claim directly.
The simulation results exhibit
This is the most-scrutinized table in a methods paper. Design rules:
- One row per DGP cell, one column block per method — or a small-multiples figure if the grid is large. The reader must compare methods within a cell at a glance.
- Show the metric that matches the claim (bias, RMSE, coverage, size, power) and report Monte Carlo standard error or replication count so the precision of the simulation is visible.
- Mark the contribution cell (where the incumbent breaks and the method holds) and the boundary cell (where the method degrades). Honesty about the boundary reads as rigor.
- Avoid burying the headline in a 40-row table; lead with a compact summary and put the full grid in an appendix if needed.
Method-comparison and notation exhibits
- A comparison table of methods × properties (assumptions required, what each is robust to, cost) orients the reader before the results — a powerful, distinctly-methods-paper exhibit.
- Notation must be consistent between equations, tables, and figures; a symbol that means one thing in the theorem and another in the table is a credibility leak at a methods venue.
- Equations are exhibits too: number them, keep estimands and estimators visually distinct, and place assumptions near the result they support.
Diagnostic figures
- For inference methods, a coverage or size figure across the difficulty parameter often communicates more than a table — show the nominal line and where each method departs from it.
- For estimators, bias/RMSE curves across sample size make the asymptotic story visible.
- Keep figures legible in grayscale where possible and ensure every figure is referenced and interpreted in the text.
ASA and SAGE formatting discipline
- Follow ASA style for table titles, notes, and numbering (检索于 2026-06；以官网为准); put method abbreviations and Monte Carlo details in the table note so the exhibit is self-contained.
- Under double-anonymized review, exhibits must not deanonymize: no author-identifying file paths, software-account names, or "as in our earlier work" captions.
- Confirm figure/table file-format and resolution requirements on the live SAGE author page before final submission.
Checklist
- The simulation grid compares methods within each cell at a glance.
- Metrics match the claim and Monte Carlo precision (SE or replication count) is shown.
- The contribution cell and the boundary cell are both visible.
- A method × property comparison table orients the reader.
- Notation is identical across equations, tables, and figures.
- Inference figures show the nominal line and each method's departure.
- Table notes make every exhibit self-contained; ASA style followed.
- No exhibit deanonymizes the authors.
Anti-patterns
- Wall-of-numbers grid: a giant table with no marked headline cell.
- Metric mismatch: reporting bias when the claim is about coverage.
- Notation drift: symbols that change meaning between theorem and table.
- Hidden Monte Carlo error: results with no replication count or MCSE.
- Deanonymizing captions: "our package" or author file paths in exhibits.
- Decorative descriptive tables: substantive-paper exhibits crowding out the simulation grid.
Output format
[Exhibit status] self-contained / needs repair / not ready
[Core exhibits] <simulation grid / method-comparison table / coverage figure>
[Headline + boundary cells] <which cells make the point and the limit>
[Notation check] consistent / drift found where
[Anonymization check] clean / risk where
[Next SMR skill] smr-writing-style