Methods Review Request Templates

To a colleague in your institution, your department or your network: someone who already knows who you are. The relationship removes the need to introduce yourself, which means the whole message can be about the question. Keep it short enough to answer from a phone.

Subject: 30 minutes on a clustered-data question? Paper going out [month]

Hi [Name],

Could I ask you to look at one methods question in a paper we are finishing?

The question: our outcome is measured at four time points for each of 312 patients across 18 clinics, and we currently fit an ordinary linear model with a clinic fixed effect. I am not confident that is right, and I would like to know whether a mixed model with random intercepts for patient and clinic is what this design needs, and what that would change in the estimates we report.

What I would send you: the analysis section (about 900 words), the two results tables, the R script, and a synthetic version of the dataset with the same structure. Everything is ready to go now.

Time: I think 30 to 45 minutes if you read the script, and I am happy to do it as a call instead if that is easier.

Timing: we would like to submit by [date], so any time in the next three weeks would work. If that is too tight, tell me and I will ask [alternative] instead, with no awkwardness at all.

On credit: acknowledgement by default, and if this turns into you reworking the analysis with us then obviously co-authorship, your call.

Thanks either way,
[Your name]

Put the time cost in the subject line. It is the fact that determines whether the message is opened now or later, and later usually means never.

State the question in technical terms in the first paragraph, with the actual numbers: four time points, 312 patients, 18 clinics. A methodologist can size the problem from the design description alone and knows immediately whether they are the right person.

List what you will send and confirm it already exists. Nothing kills momentum like an agreement followed by a week's silence while you prepare files.

Name the alternative person you will ask if they decline, in the same sentence as the deadline. It makes the no genuinely costless, and it is the line that most reliably prevents an obligated yes that never happens.

To a specialist you have not met, chosen because they published on the method. This message has to do three extra jobs: establish who you are, establish why them specifically, and make clear that you are not asking for an open-ended review of a paper in a field they do not work in.

Subject: One question on staggered-adoption difference in differences, from your 2024 paper

Dear Dr [Name],

I am [name], [position] at [institution], working on [one clause: your area]. I am writing because of your 2024 paper on estimator bias under staggered treatment timing, which we read while trying to work out whether our own design is affected.

The specific question. We evaluate a policy adopted by 23 regions between 2018 and 2023, with never-treated regions as controls, and we currently estimate a two-way fixed effects model. Our reading of your paper is that the negative-weighting problem applies to us, and that we should be reporting a heterogeneity-robust estimator instead. I would like to know whether you agree that this design falls in the affected class, and if so which of the alternatives you would consider defensible here.

What this would involve. Reading roughly two pages of design description plus one table, or a 30 minute call, whichever you prefer. I am not asking you to review the paper: the substantive area is [field] and I know that is not yours. The estimation strategy is the whole of my question.

What I can send: the design and estimation section, the treatment timing table, the Stata code, and the panel structure. Data are administrative and restricted, but I can supply a synthetic panel with identical structure.

If you are willing, I would acknowledge you by name in the paper, or handle it as a paid consultancy through our institution if that is how you prefer to do this kind of work. If your contribution went beyond a review and into the design of the estimation, we would want to discuss co-authorship.

If this is not something you have time for, I completely understand, and I would be grateful for the name of anyone you think would be better placed.

With thanks for your time,
[Your name]
[Position, unit, institution]
[ORCID or institutional page]

Make the subject line name their work and your question. It is the only thing distinguishing you from the many people who email specialists asking them to read papers.

Cite the specific paper of theirs that led you to them, and say what you took from it. This proves the approach is targeted, and it usually establishes that you have already done the reading.

Identify yourself in one line with an institution and a checkable link. A stranger deciding whether to spend an hour on you will look, and giving them the link is faster than making them search.

Say explicitly that you are not asking for a review of the paper and that you know their field is not yours. This is the sentence that prevents the most common decline, which is that the topic is outside their area.

Offer both recognition routes, acknowledgement and paid consultancy, and let them pick. Many methodologists do this work formally and will find an unpaid favour request from a stranger easier to decline than a costed one.

Close with the invitation to refer someone else. A decline with a name attached is worth almost as much as an acceptance, and asking for it costs one sentence.

The quality of the answer is decided mostly by what arrives with the question. Prepare all of it before the first message, so that a yes turns into work the same day.

The design and analysis section as a standalone document, not the whole manuscript. Two to four pages that describe the data structure, the estimation and the assumptions, extracted so they can be read without the introduction.

The code, cleaned enough to be read. Comments at the top saying what each block does are worth more than perfect style, and a reviewer who can run the code finds things a reviewer who reads prose never will.

Data, or a synthetic version with the same structure. Where the data are restricted, say so plainly and offer the synthetic panel or a dummy dataset: an accurate structure with fabricated values answers most specification questions.

The reporting checklist for your study type, if your field has one. Sending it filled in shows what you have already checked yourself and stops the reviewer spending their time on it.

Leave out

the introduction and discussion, the cover letter, the journal target, the reviewer reports unless they are the reason for the request, and anything about how much this matters to your career. None of it helps the analysis and all of it adds reading.

The last part is what turns a helpful reply into a better paper, and it is also what determines whether the same person answers you next time.

Come back with what you did. A short message saying which of their points you acted on, which you did not and why, takes ten minutes and is the single strongest predictor of whether they will do it again.

Do not attribute conclusions to them in the paper. Acknowledge the reading, not the endorsement, and never write that the analysis was verified by a named person unless they have agreed to that exact sentence.

If their answer means the paper needs a different analysis, say so to your co-authors before you decide anything. A methods review that changes the result is doing exactly what you asked it to do, and burying it is the outcome that eventually costs most.