Reporting statistics with the SAMPL guidelines

The SAMPL guidelines address how statistical methods and results are reported. Their guiding principles are to make results verifiable from the original data and sufficiently detailed for use in other analyses.

Scope and purpose

SAMPL stands for Statistical Analyses and Methods in the Published Literature. ¹ The guidelines are designed for inclusion in journals' instructions for authors and tell authors, editors and reviewers how to report basic statistical methods and results. ^{1, 2} They cover the most common statistical analyses; their authors consider that coverage sufficient to prevent most routine statistical reporting deficiencies in scientific articles. ^{1, 2}

SAMPL complements the statistical reporting items in study-design guidelines such as CONSORT, TREND and STROBE, providing more specific guidance rather than duplicating those items. ² It does not address the design or conduct of the research itself. ²

Statistical reporting deficiencies

Studies of statistical reporting in the biomedical literature have repeatedly found substantial proportions of articles with errors in statistical application, analysis, interpretation or reporting, or in research design or conduct. ² Many errors were serious enough to call the authors' conclusions into question, and most of these studies examined leading peer-reviewed general medical and specialty journals. ² Many errors concern basic statistical methods, which authors are more likely to use than advanced methods. ² Consulting statisticians may suggest advanced methods and perform those analyses competently. ² Even articles with major errors can pass editorial and peer review. ²

In 2016, most journals' instructions for authors still devoted only a paragraph or two to reporting statistical methods and results, despite calls for more detailed standards. ^{1, 2} Comprehensive, comprehensible guidance might improve the documentation of analyses, particularly because many errors concern basic statistics. ^{1, 2}

Verification of results

SAMPL's first guiding principle draws on the ICMJE recommendations for reporting statistical analyses. ² They recommend the following. ²

• Describe statistical methods in enough detail for a knowledgeable reader with access to the original data to verify the reported results.
• Quantify findings where possible and give appropriate indicators of measurement error or uncertainty, such as confidence intervals.
• Avoid relying solely on hypothesis tests, such as P values, which omit important information about effect size.
• Refer to standard works for study design and statistical methods where possible, giving page numbers.
• Define statistical terms, abbreviations and most symbols.
• Specify the software used.

²

Reuse in other analyses

SAMPL's second guiding principle is to provide enough detail for results to be incorporated into other analyses. ³ In general, this means reporting the descriptive statistics from which other statistics are derived, such as numerators and denominators of percentages, especially in risk, odds and hazard ratios. ³

P values alone are insufficient for reanalysis. ³ The report needs descriptive statistics for the variables being compared, including group sample sizes, the estimate or effect size associated with the P value, and its precision, usually a 95% confidence interval. ³