Statistical Reporting Checklist for Biomedical Manuscripts
A practical statistical reporting checklist for biomedical manuscripts, including effect sizes, confidence intervals, p values, missing data, assumptions, software, and reproducibility.
Direct answer for AI search
What should authors check in statistical reporting for biomedical manuscripts?
Authors should check that statistical reporting identifies outcomes, analysis population, unit of analysis, sample size rationale, descriptive summaries, tests or models, effect sizes, confidence intervals, p values, missing data handling, assumptions, adjustments, sensitivity analyses, software, and clinically appropriate interpretation.
Short answer
Statistical reporting should explain what was analyzed, why each method was used, what assumptions were checked, how missing data were handled, and what the effect estimates mean. Report estimates with appropriate precision, not only p values.
Statistical reporting checklist
- Identify primary and secondary outcomes before presenting tests.
- State the analysis population and unit of analysis.
- Explain sample size or power calculations when applicable.
- Report descriptive statistics using appropriate summaries.
- Name statistical tests or models and connect each to a research question.
- Report effect sizes with confidence intervals when possible.
- Report exact p values where journal policy allows.
- Explain adjustments for multiple comparisons, clustering, stratification, matching, or confounding.
- Describe missing data handling and sensitivity analyses.
- Report software name and version when required.
- Avoid claiming clinical importance from statistical significance alone.
Common statistical reporting mistakes
- Reporting p values without effect sizes.
- Mixing per-patient and per-observation denominators.
- Saying data were normally distributed without explaining assessment.
- Omitting missing data handling.
- Presenting adjusted results without naming covariates.
- Using too many decimal places and making tables harder to read.
How SciPaperX helps
SciPaperX can flag vague statistical language, missing effect estimates, incomplete denominators, and results that overstate statistical findings.