Writing the Reviewer Response: Handling Statistical Revisions

A major revision decision is not a rejection — it is an invitation. Whether the invitation turns into an acceptance depends as much on your reviewer response letter as on the revision itself. In the second round, editors and reviewers usually read your response letter before they open the revised manuscript; a disorganised, defensive or incomplete letter can overshadow even meticulously executed additional analyses. This guide covers the letter anatomy that works for statistics-heavy revisions, the three legitimate response strategies, and the right way to handle the most frequent statistical requests.

Anatomy of the reviewer response letter

Only one format works: the point-by-point response. Quote each reviewer comment verbatim, then immediately below it give your answer and the exact location of the change in the manuscript (page, line, table number). Do not soften the reviewer's words by paraphrasing them: the verbatim quote is your proof that nothing was skipped and nothing was distorted.

• Numbering: use a fixed scheme such as "Reviewer 1 – Comment 3" so the editor can cross-check in seconds.
• Visual separation: reviewer text in italics or block quotes, your answer in plain text, new manuscript text in quotation marks or a distinct colour.
• Stating the change: "We have corrected this" is not enough; write "The following sentences were added to the Methods section (p. 12, lines 8–14): …".
• Two versions: prepare both a clean copy and a tracked-changes copy, even when the journal does not ask for them.

The three response strategies

There are exactly three legitimate ways to answer any comment; trouble comes from everything outside these three — quietly glossing over a point, or dismissing it without grounds.

1. Comply fully: do exactly what was asked and show where. The large majority of your answers should fall into this category; do not turn easy wins into debates.
2. Comply partially with justification: meet the substance of the request by a different route, and explain why with evidence. Example: when a reviewer proposes a different model, keep your main model and present the proposed one as a robustness check in a supplementary table.
3. Respectfully rebut with evidence: if the request is technically flawed (e.g. a post-hoc power analysis for a significant result), object politely, ground the objection in the methods literature, and always offer a constructive alternative. Rebuttals must remain the exception across the letter.

The most common statistical requests, and how to answer them

Statistical reviewers' requests are remarkably repetitive. "Add effect sizes" is the easiest: attach the appropriate measure (d, η², r) with a 95% confidence interval to every test; our effect size guide maps which measure belongs to which test. "Report assumption tests" likewise calls for straightforward compliance: report normality and homogeneity of variance checks, and where an assumption is violated, also present the results of a robust alternative.

The most delicate request is justifying the sample size after the fact. The dishonest route is computing "post-hoc power" from the observed effect: observed power is a direct function of the p value and carries no new information. The honest route is a sensitivity analysis: in G*Power, take the sample you actually have, set α = 0.05 and power at 0.80, and compute the smallest effect size the study could detect; then discuss that value against typical effects in your field. Reviewers almost always accept this framing, because it is the only technically defensible form of retrospective sample size assessment.

When asked to correct for multiple comparisons, Bonferroni is not the only option; you may propose less conservative procedures such as Holm or Benjamini–Hochberg (FDR), and justify the scope of the correction by clearly separating primary from secondary hypotheses. Robustness checks (alternative estimators, different cut-off points, re-analysis excluding outliers) belong in the supplementary materials so the main text stays lean; if the headline finding survives, these checks strengthen rather than weaken your paper.

Tone rules: thank concretely, never defend

• Never write defensively. If a reviewer misunderstood something, locate the fault in the text's clarity: "We realised we had not expressed this point clearly enough; the paragraph has been rewritten (p. 9, lines 3–10)."
• Make gratitude concrete: instead of the generic "thank you for your valuable comments", say in one sentence how the comment improved the paper.
• Excessive praise and grovelling inspire no confidence; stay professional, calm and data-driven.
• Close every answer with a clear bottom line: what changed, and where.

Contradictory reviewers, deadlines and the cover letter

When Reviewer 1 asks for one change and Reviewer 2 demands the opposite, do not quietly pick a side. Name the contradiction openly in the response letter and place the decision, with your reasoning, in the editor's hands: "Reviewer 1 recommends X while Reviewer 2 recommends Y; we considered both, implemented X for the following reason, and present Y as a supplementary analysis." Where possible, propose a middle path that partially satisfies both requests; if you remain genuinely unsure, a short query email to the editor before resubmission is entirely legitimate.

If the deadline will not hold, request an extension from the editor well before the final day — this is a routine, unremarkable request. At resubmission, keep the cover letter short: a three-to-five-bullet summary of the main changes, thanks to the reviewers, and a note that every comment has been answered point by point. If you re-ran any analyses, verify one final time that every number is identical across the manuscript text, the tables and the response letter — an inconsistent number is the most needless cause of a second-round rejection.

A good response letter is not a defence of your paper; it is the evidence of how your paper got stronger.

Frequently Asked Questions