nature-statistics

Submit your Methods or Statistical analysis text for a bounded audit that maps each claim to its test, assumptions, sample size, and correction strategy. The review returns severity-labeled issues (P0/P1/P2) plus ready-to-paste revised wording.

Each quantitative panel legend should define what n represents, the summary convention (mean ± s.d., s.e.m., or CI), the test used, and exact p values or defined star thresholds. The figure-statistics reference checks bar plots, box plots, violins, time courses, and heat maps against these requirements.

Pseudoreplication occurs when cells, images, fields, or technical wells are counted as independent n instead of the true experimental unit such as animals, donors, or cultures. The fix is to analyze independent units, aggregate technical subsamples, or use hierarchical mixed-effects models.

It is a reporting and review skill, not a substitute for a statistician. It only performs concrete reanalysis when you supply the raw data and explicitly request computation, and it will not recommend a final test when the unit of analysis or design is unclear.

Provide the reviewer comments and your current text to get a structured response draft covering the concern, the author-side action needed, and revised manuscript wording. It never claims a new analysis was performed unless you supply the results.

The review cannot infer experimental units or independent replicates from figures alone and cannot verify facts not present in the supplied material. Missing sample sizes, tests, or design details are flagged as AUTHOR_INPUT_NEEDED rather than guessed.