论文数据表格制作与校验
将已有研究数据整理成规范论文表格,统一单位、精度和脚注,并核查数字一致性。
完整提示词
<context>
Tables are often what statistical reviewers read first. Good tables are self-contained: a title that says what, who and when; column headings with units; consistent decimal places by variable type (percentages to one decimal, p values by the style's rules); n per group in the header; abbreviations and the statistics used defined in footnotes; and missing data shown, not hidden. Conventions to know: a baseline characteristics table in a randomised trial should not include p values for between-group differences (CONSORT advises against them, since differences are due to chance); regression tables should say whether coefficients are unadjusted or adjusted and list the adjustment set in a footnote, report the reference category, and give confidence intervals rather than only standard errors or stars; mean (SD) for roughly symmetric variables and median (IQR) for skewed ones. Common failures are copying software output with every column, mixing decimals, and tables that duplicate the text.
</context>
<task>
Build the tables from:
<data>
{{data}}
</data>
{{#style}}
Style: {{style}}
{{/style}}
1. Decide how many tables are needed and what each shows; suggest moving any table that duplicates the text or a figure to supplementary material.
2. For each table, write a numbered title, then the table in Markdown with column headings including units and group sizes.
3. Apply consistent formatting: summary statistic type per variable, decimals, how categories are listed (indented under the variable), reference categories marked, confidence intervals in a consistent format.
4. Add footnotes: abbreviations, statistics used, adjustment variables, missing data per variable, and symbols.
5. Check the numbers: percentages against counts and denominators, group totals against the overall N, CIs containing the estimate.
6. List decisions you made and anything the author must confirm.
</task>
<constraints>
- Use only the numbers given. Do not compute new statistics except percentages from given counts, and label those as computed.
- Do not add p values to a randomised trial's baseline table; explain why if the author asked for them.
- Do not use asterisks for significance unless the style requires them; prefer exact p or CIs.
- If the style is unknown, use a neutral style compatible with APA and medical journals, and name it.
</constraints>
<output_format>
## Tables
Each table with its numbered title, in Markdown.
## Footnotes and abbreviations
Footnotes per table.
## Consistency checks
Each check and its result; problems first.
## Decisions and questions
Bullets.
</output_format>可替换内容
- {{data}}
- {{#style}}
- {{style}}
- {{/style}}
分类与标签
办公效率、论文表格、数据校验、表格排版、学术写作