A common scenario in statistical consulting is this: a doctoral student spends two years designing a study and collecting data, reaches the analysis stage, and discovers that statistics is exactly the part they handle worst. The supervisor has no time (or doesn't master it either), the funding clock is ticking, and the question appears: should I get help with the statistics of my thesis? And if I do, how do I know I'm not being sold smoke? This guide answers those questions plainly: what a statistician actually does for a thesis, when it makes sense to hire one, whether it's ethical, and how to tell a good professional from someone who just "runs your data through SPSS".
What a statistician actually does for a PhD thesis
Let's clear up what it is not: it is not someone who "does your thesis for you". A statistician or methodological consultant is a specialist who supports the quantitative part of your research so that it is correct, defensible and reproducible. The concrete work depends on your stage, but it usually covers several blocks.
In the design phase, they help define which study type answers your question, calculate the sample size with proper criteria (not "the ones I could get"), and plan in advance which analyses you'll use. This phase is the most undervalued and the one that prevents the most trouble: a poorly planned design cannot be fixed afterwards with sophisticated statistics. If you are still here, this is the best moment to talk to someone, and also to read about how to choose the right statistical test for your design.
With data already collected, the work is cleaning and preparing the dataset, checking the assumptions of each model, running the analysis (from an ANOVA to a SEM, a multilevel model or a meta-analysis) and, above all, interpretation. A good professional doesn't hand you an SPSS table and leave: they explain what each number means in the context of your hypothesis. And they draft the Results section in APA 7 format (or Vancouver, depending on your field), with tables and figures ready to insert.
Finally, there is the phase many forget until it arrives: the defence and the reviewer response. A well-executed analysis must hold up when a committee member asks "why did you do it this way?" or when Reviewer 2 demands an additional analysis. That robustness is built beforehand, not improvised.
Is it ethical to hire a statistician for your thesis?
This is the question that holds doctoral students back the most, and the short answer is: yes, as long as it's done transparently. Receiving methodological and statistical support is recognised in serious university regulations and in the guidelines of international journals. The key is disclosure: the statistician is mentioned explicitly as a methodological adviser, whether in the acknowledgements, the Methods section or the Author Contribution Statement, without appearing as an author if they don't meet authorship criteria.
The ethical line is clear. One thing is a specialist helping you run and understand an analysis that you direct and comprehend; quite another is someone writing your thesis for you or fabricating results. The former is standard, legitimate practice in science (large research groups have in-house methodologists); the latter is fraud. A good consultant makes sure you understand what was done, precisely because you will have to defend it.
The goal isn't to make statistics disappear from your life, but to get you unstuck so you can defend every analytical decision in your thesis with confidence.
When it's worth hiring one (and when it isn't)
There are three moments when seeking help especially makes sense. The first is at the start, in the design phase: here the investment is small and the return huge, because you avoid collecting data that later can't answer your question. The second is once you have the data and don't know how to approach it: the most frequent case, and where a couple of well-used sessions save you weeks of trial and error. The third, and most urgent, is when a reviewer asks for an analysis you don't master and you have a major-revision deadline looming; there every day counts, and it helps to be clear on what to do when a reviewer asks for an analysis you can't run.
When isn't it worth it? If your analysis is simple (a correlation, a t-test) and you have time to learn it, good tutorials and a free tool will probably be enough. An honest consultant will tell you so: not everyone needs consulting, and pushing a service that adds no value is no way to work.
How to choose a good statistician: signals and red flags
This is where most people go wrong, because the market mixes very different profiles under the same label. These are the signals worth watching.
Signs of a good professional
- Real training and publications: someone who researches and publishes in indexed journals knows first-hand what a reviewer demands, not just the textbook theory.
- They ask questions before quoting: without understanding your design and question, no serious professional can tell you which analysis you need.
- They give you the code and syntax: commented R, SPSS or Mplus, so the analysis is reproducible and you can rerun or defend it.
- They are honest about your data's limits: if your N doesn't support a complex SEM, they say so and propose alternatives instead of promising magic.
- They work by project with a closed scope: you know what you receive and when, with no inflated hours.
Red flags
- Promises "significant" or "publishable" results before seeing your data. That can't be promised, and whoever does is willing to force the analyses.
- Hands you tables with no interpretation or explanation of the decisions made.
- Can't explain why they chose one analysis over another with methodological references.
- Charges by the hour with no defined scope, so you never know the final cost.
- Avoids being acknowledged and suggests hiding their involvement.
Before hiring, I also recommend reading about the most common statistical errors in theses: understanding those errors helps you judge whether the person helping you really knows how to avoid them.
How much does it cost? By project, not by the hour
You won't find a fixed rate here, and there's a reason: every project is different. Validating a questionnaire with four subscales doesn't require the same work as a longitudinal multilevel model with missing data. What you can (and should) demand is a closed written quote before starting, calculated by project and not by the hour. The by-project model protects you: the cost is agreed upfront, and however long the professional takes is their concern, not a surprise invoice for you. If you want to see how this kind of collaboration is structured, I explain it in detail in how I work.
What to ask before saying yes
A short initial conversation should be enough to decide. Come prepared with three things: the project type (full thesis, a specific analysis, a reviewer response), the software you've already worked with (or none) and your timeline. With that, a serious professional can advise you on the approach, timelines and an approximate budget. If in that conversation you sense they understand you, explain things clearly and are honest about what your project needs (and doesn't), that's a good sign.
If you're at that point, you can start with a no-obligation assessment of your case in my statistical consulting, or look directly at how I handle the statistical analysis of doctoral theses. And if yours is a revision with a demanding Reviewer 2, there's a specific route for the reviewer response. Statistics shouldn't be the wall that stalls your PhD: with the right support, it's just one more step.