Statistical analysis reporting (lite) — v1.0.0
Released: 2026-02-02 · Hash: sha256:7ec7dd661290eeaf716548368772bd287ce959cebe44a1b2a7695af9667de9b5
Analysis scope
| Criterion | Text |
|---|---|
| must STAT-01 | Define the analysis population(s) and unit of analysis, including how repeated measures or clustering are handled. |
| must STAT-02 | Define outcomes, predictors, and derived variables (coding, transformations, categorizations) used in analyses. |
| should STAT-03 | Describe the sample size determination (power/precision) and key design parameters (alpha, effect size, ICC if clustered). |
Descriptive reporting
| Criterion | Text |
|---|---|
| must STAT-10 | Report descriptive statistics with appropriate summaries (e.g., mean±SD vs median[IQR]) and clear denominators. |
| must STAT-11 | Report missing data extent for key variables/outcomes, including counts/percentages by group when relevant. |
| should STAT-12 | If data exclusions, winsorization, or outlier rules were applied, define the rules and report how many observations were affected. |
Inference basics
| Criterion | Text |
|---|---|
| must STAT-20 | For each main analysis, state the estimand/target comparison and the statistical method/test used, including key assumptions. |
| must STAT-21 | Report effect estimates with measures of precision (e.g., confidence/credible intervals), not just p-values. |
| should STAT-22 | Report exact p-values where used and specify two-sided vs one-sided tests; avoid sole reliance on thresholds (e.g., p<0.05). |
| should STAT-23 | Address multiplicity (multiple outcomes, comparisons, interim looks, subgroups) via adjustment or cautious interpretation. |
Model reporting
| Criterion | Text |
|---|---|
| must STAT-30 | For regression/time-to-event models, specify model type, link function, covariates, selection strategy, and any interactions/nonlinear terms. |
| must STAT-31 | Report adjusted and unadjusted estimates where relevant, and list the variables included in adjusted models. |
| may STAT-32 | Report model diagnostics/goodness-of-fit and checks of key assumptions (e.g., residuals, proportional hazards). |
| may STAT-33 | If continuous predictors were categorized, justify cutpoints and consider sensitivity analyses with alternative modeling. |
Robustness & missing data
| Criterion | Text |
|---|---|
| should STAT-40 | Describe sensitivity/robustness analyses, what changed, and what remained stable; separate exploratory analyses. |
| must STAT-41 | Describe missing data handling methods (e.g., complete-case, multiple imputation, weighting) and key assumptions; include sensitivity analyses when assumptions matter. |
| may STAT-42 | If Bayesian methods were used, report priors, computation, convergence diagnostics, and posterior summaries. |
Computation & reproducibility
| Criterion | Text |
|---|---|
| should STAT-50 | Report statistical software (name/version) and key packages, and provide analysis code and random seeds when feasible. |
| may STAT-51 | Document reporting conventions (units, rounding, decimal places) and handling of small cell counts where applicable. |