1. Research question and target quantity

The reviewer should identify the population, comparison, outcome, timing, and effect or performance measure the study intends to estimate. Ambiguity here propagates through every later analytical decision.

2. Study design and data-generating structure

Randomization, matching, clustering, repeated measurements, censoring, competing events, sampling strategy, case–control selection, and survey design determine which analyses and standard errors are appropriate.

3. Variable and outcome definitions

Primary and secondary outcomes, exposure timing, covariate measurement, composite endpoints, cut points, transformations, and derived variables should be clinically meaningful and consistently applied.

4. Missing data and exclusions

The amount, pattern, handling, and potential consequences of missingness should be described. Complete-case analysis, imputation, exclusions, and loss to follow-up rely on assumptions that affect validity.

5. Model choice and assumptions

The model should suit the outcome and design. Functional form, proportionality, independence, distributional assumptions, influential observations, multicollinearity, convergence, calibration, and overfitting may require assessment depending on the method.

6. Effect estimates, uncertainty, and multiplicity

Report suitable effect measures with confidence intervals and denominators. The role of hypothesis tests, multiple outcomes, subgroup analyses, repeated testing, and any adjustment strategy should be transparent.

7. Sensitivity and robustness

Alternative defensible assumptions can reveal whether conclusions depend on a particular model, definition, missing-data method, influential subset, or analytic choice. Sensitivity analyses should answer identified sources of uncertainty rather than serve as decoration.

8. Reporting and reproducibility

The manuscript should specify the analysis population, methods, software and versions, variable handling, model-building approach, diagnostics, and deviations from the analysis plan. Code and data availability should be described accurately and subject to ethical, legal, and confidentiality constraints.

Authoritative resources

This educational article provides general guidance and is not a statistical, methodological, clinical, ethical, or legal assessment of any particular study.