The reviewer asked for new analyses. Get them run, checked, and documented.

Is this for you?

• You have a decision letter with specific statistical requests: an added covariate, a subgroup or interaction test, a sensitivity or missing-data check.
• The data already exist. This is reanalysis, not new data collection.
• The requests can be answered with linear models: regression and ANCOVA-style models, robust standard errors, inverse-probability weighting. Other model families are scoped case by case, or declined.

What we agree before any work

One reviewer letter, up to three analyses. For each analysis we write down the reviewer’s request in their words, the exact quantity being estimated, the model, and how missing data are handled. You approve that plan before anything is run.

Missing data are never imputed, weighted, or dropped silently. If the plan does not say how to handle them, we ask. Every row removed from an analysis is counted in the output.

What you get

• Reproducible R code for every analysis, with the package versions recorded.
• Revised tables beside the originals. Your original results are locked first and stay unchanged.
• A change log of every difference between original and revised results, including every change in sample size and why it happened.
• A Word crosswalk: reviewer comment, action taken and what was estimated, original result, revised result, and the exact table and code behind it. The numbers are filled in from the analysis files, not typed.
• An independent check: each model is refit in separate software, and the two sets of results must agree to five decimal places before anything is delivered.

How the work is kept honest

What a crosswalk row looks like

What it is not

• It does not write your response letter. You get plain-language explanations of each result to use in it.
• It does not promise a reviewer will accept the result, and it will tell you plainly if a requested analysis conflicts with your original design.
• More than three analyses, or a second reviewer letter, is quoted separately.

Data, stated plainly

• This is a pilot. Spots are limited, and scope and terms are confirmed with you in the quote before any work starts.
• Unlike a manuscript review, a reanalysis needs your analysis dataset. Send a de-identified file only: no names, contact details, or direct identifiers, and only if your IRB approval or data use agreement lets you share it with an outside analyst. If it does not, we cannot run the sprint on that data.
• Do not put data in the quote form. After you approve the quote, you share files through your own shared link; we work on them on the reviewer’s own computer and delete them within 30 days of delivery. See Confidentiality and the privacy policy. An NDA is available on request.
• Multiple imputation can be quoted, but it cannot yet be cross-checked to the same standard (random draws differ between programs), and we tell you so in the quote.