Send three things before a first call with a statistician: your research question written in plain language (not in the language of a specific statistical test), a data dictionary describing your variables (not the raw dataset itself), and your real timeline, including the deadlines you don't control. Most of what makes a first call unproductive is the absence of one of these three.
"I need someone to run a mixed model" tells a consultant almost nothing useful — it skips straight to an answer before anyone has heard the question. "We're trying to find out whether the tutoring program improved reading scores more for students who started behind" tells them everything: what you're actually trying to learn, what the comparison is, and where the interesting complexity probably lives. Naming a specific method up front isn't wrong, but it shouldn't replace describing the question — a consultant who only hears the method has to reverse-engineer the question anyway, and sometimes the method you had in mind isn't the one the question actually needs.
Before a scope of work exists, raw data usually shouldn't move at all — there's no agreement yet about confidentiality, and a consultant doesn't need to see individual records to give you a useful first read. What's actually useful is a short data dictionary: what each variable measures, how it's coded, roughly how many rows and columns you have, whether the structure is one row per person or something nested (repeated visits, students within classrooms, clusters of any kind), and a rough sense of how much is missing and why. That's usually enough to tell whether the analysis is straightforward or whether the design itself needs a second look, without anyone exchanging a file that shouldn't be exchanged yet.
"As soon as possible" scopes nothing. What scopes a project is the actual chain of dates behind it: an IRB submission window, a grant reporting deadline, a conference abstract due date, a board meeting where results need to be presented. Those external deadlines often matter more than how much time you'd personally like to spend, because they determine what's realistically possible — a robustness check that would take two extra weeks isn't a small ask if the abstract is due in ten days. Say which deadlines are firm and which are aspirational; a consultant scopes those two situations very differently.
If you've already run an analysis, gotten reviewer or committee feedback, or talked to another consultant, say so and share what came of it. This isn't about litigating what went wrong before — it's that a consultant who knows a linear model was already tried and criticized for ignoring clustering can skip straight past that conversation instead of re-discovering it forty minutes into the call.
None of this needs to be formal. A half-page email covering these three things will get you a more useful first call than a data file and a one-line request ever will. If you're not sure how to describe your own question yet, that uncertainty is worth saying out loud too — that's exactly the kind of conversation a first call is for.
One email whenever we publish something new. No spam, unsubscribe anytime.
{{ subscribeErrorMsg }}