Engaging a statistical consultant for a doctoral dissertation is legitimate, common, and often sensible. Most doctoral programs train candidates to be competent researchers in their discipline, not professional statisticians, and a study that uses a multilevel model, a structural equation model, or a survival analysis frequently requires expertise beyond what any single candidate has been taught. Bringing in a specialist for that expertise is no different in principle from consulting a research methodologist, a subject librarian, or a university writing center. What matters is not whether you engage a statistician but how you do it: what you ask them to do, when you bring them in, and what you keep firmly as your own work.
That last point is where the whole question turns. There is a line between the statistical support that strengthens a dissertation and the kind that hollows it out, and the line is not vague. A single test defines it: at the end of the process, can you explain and defend every analytic decision in the document as your own? This article maps the scope of legitimate statistical consulting, explains when in your timeline to engage a statistician and why the timing is not negotiable, sets out how the work is priced and what drives the cost, and gives you a red-flag checklist for evaluating whoever you are considering. Throughout, it keeps returning to the defensibility test, because that is the test your examiners will apply.
Quick Answer:
A statistician may legitimately help you design the analysis, run a power analysis before you collect data, clean and check your data, select and run the appropriate models, produce output, and explain that output so you understand it. What a statistician may not do is write your results narrative as though it were your own thinking, or stand in for your understanding when you are examined. Engage a statistician at the proposal or design stage, not after data collection, because a power analysis and several design decisions cannot be repaired retrospectively. Cost is driven by design complexity, data condition, and the number of models and revision rounds. The non-negotiable requirement is that you must be able to defend every analytic choice yourself.
What a Statistical Consultant Actually Does
Statistical consulting for a dissertation covers a wide range of activities, and it helps to see them as a ladder from lightest to most intensive. At the lightest end sits design consultation: advising, before any data exists, on what analysis your research questions imply and what sample size that analysis will require. Above that sits data preparation, the cleaning and assumption-checking that determines whether your planned analysis is even valid for your data. Higher still sits model selection and the running of the analyses themselves. Near the top sits the production of output tables and figures, and the work of explaining that output to you until you understand not just what it says but why.
At the very top of the ladder, above everything a consultant may legitimately do, sits the work that is yours alone: writing the interpretation as your own scholarly argument, and defending the analysis when you are examined. A consultant may bring you right up to that final rung, but cannot climb it for you.
Table 1: The Scope Ladder, From What a Statistician May Do to What Only You Can Do
Activity | What it involves | Side of the line |
|---|---|---|
Design and power consultation | Advising, before data exist, on the analysis your questions imply and the sample size it requires | Legitimate; most valuable at the proposal stage |
Data cleaning and assumption testing | Structuring the dataset, handling missing data, and checking whether the planned analysis is valid for your data | Legitimate |
Model selection and running the analysis | Identifying and running the models your design requires, with the correct specification | Legitimate |
Producing output tables and figures | Generating the tables, figures, and statistics for you to report | Legitimate, provided the reporting decisions are yours |
Explaining the output to you | Working through the results until you understand what they mean and why the model was specified as it was | Legitimate, and the most valuable help of all |
Writing the interpretation as your own argument | Composing the account of what the findings mean and contribute | Yours alone. A consultant may check accuracy, not author the meaning |
Defending the analysis under examination | Answering the examiner's questions about why each choice was made and what it supports | Yours alone. This is what the degree certifies. |
The distinction that matters within this range is between work that produces a result and work that produces your understanding of the result. A statistician who runs your mixed-effects model and hands you the coefficients has produced a result. A statistician who then sits with you until you can explain what the fixed and random effects mean, why the model was specified that way, and what the output licenses you to claim has produced your understanding, which is the thing the doctorate actually certifies. The second kind of help is the kind worth paying for, because it is the kind that survives the viva.
The Line You Cannot Cross
The boundary between legitimate consulting and academic misconduct is not a matter of how much help you receive but of what kind. The clearest way to locate it is through the professional standard that governs statisticians themselves. The American Statistical Association's Ethical Guidelines for Statistical Practice, revised and approved by the ASA Board in February 2022, define ethical practice for anyone who designs, analyzes, interprets, or presents data. The 2022 revision deliberately broadened its scope by replacing the term "statistician" with "statistical practitioner" so that the standard reaches consultants, data scientists, and students alike. A statistician who works on your dissertation is bound by these guidelines whether or not they mention them.
Two of the ASA principles bear directly on the boundary. The guidelines require the practitioner to be transparent about their role and to produce valid, interpretable, and reproducible work. They state that the ethical practitioner "seeks guidance, not exceptions" and does not exploit gaps in the rules to justify conduct that the rules would not sanction. Applied to a dissertation, this means a consultant who understands their own professional ethics will insist that their contribution is disclosable, that you understand what was done, and that the analytic reasoning is yours to defend. A consultant who offers to write your results chapter so that no one will know they were involved is offering to violate their own professional standard, and that offer is itself the reddest of flags.
The practical version of the boundary is the disclosure test and the defensibility test taken together. If the help you received could be described openly to your committee without jeopardizing your standing, it is on the legitimate side. If it depends on concealment, it is not. And if you cannot, after the help, explain and defend the analysis as your own, then, regardless of how it was arranged, the work has crossed from support into substitution. This is precisely why ScribeLabWriter's dissertation statistical analysis support is built around the analysis you direct and the interpretation you come to on your own, rather than a hidden results chapter you could not account for under questioning.
When to Engage a Statistician: The Timing Is Not Negotiable
The single most consequential mistake candidates make is engaging a statistician too late, after the data are already collected. By then, the decisions that most benefit from statistical expertise have already been made, often wrongly, and several of them cannot be undone.
The clearest example is the power analysis. An a priori power analysis works by fixing three of the four quantities that govern statistical power, which are the effect size you expect, the alpha level you will use, the power you want, and the sample size, so that the fourth can be solved for. As the open-access methodological review by Kang on power analysis with G*Power sets out, the standard procedure is to establish your hypotheses, choose the statistical test your design implies, select the appropriate power-analysis method, enter the known values, and calculate the sample size you need.
Done before data collection, this tells you how many participants to recruit. Done after, it cannot rescue a study that recruited too few, because the data are already gathered. A retrospective, or post-hoc, power calculation is widely criticized precisely because it is a deterministic function of the result you already obtained and therefore tells you nothing you did not already know. The methodology of getting this right is the subject of a full worked guide on conducting a power analysis in G*Power, and the recurring lesson of that guide is that the calculation belongs at the design stage.
Power is not the only design decision that resists retrospective repair. The choice of measurement instrument, the structure of the sampling, the inclusion of a comparison group, and the timing of measurement occasions are all fixed once data collection begins, and each one constrains what analysis is later possible. A statistician consulted at the proposal stage can catch a design that will not support the analysis your research question requires, at the point when catching it is still free. This is why the most valuable time to bring in statistical expertise is during methodology and design development, well before the first participant is recruited.
Engaging late is not fatal to every study, and a good consultant can often make the best of data already collected. But the candidate who plans the analysis before collecting data, rather than collecting data and then asking what can be done with it, is in a categorically stronger position, and the difference is one of design rather than effort.
Planning your analysis before you collect data? |
|---|
The cheapest statistical problem to fix is the one caught at the design stage, before it is built into the data. A specialist can review your proposed design, run the a priori power analysis, and confirm that the analysis your research question requires is the analysis your design will support. Send us your proposed design, and you will have an itemized quote within 2 to 4 business hours, no obligation. |
What Drives the Cost
Statistical consulting is priced in one of two ways, and understanding which one fits your situation prevents both overpaying and under-scoping. Hourly pricing suits open-ended work: troubleshooting a model that will not converge, coaching you through interpretation, or advising on a design that is still taking shape. Fixed-project or milestone pricing suits a defined deliverable, such as running and interpreting a pre-specified set of models, where the scope is clear enough to quote in advance.
Within either model, the cost is driven by a small number of factors that you can assess yourself before you ask for a quote. The first is the complexity of the analysis. A one-way analysis of variance or a set of t-tests sits at one end of the range; a multilevel model, a structural equation model, a survival analysis, or a Bayesian analysis sits at the other, and the difference in the expertise and time required is substantial. The choice of which test your design actually requires is itself worth settling early, and the reasoning behind it is set out in a decision guide on selecting the right statistical test.
The second cost driver is the condition of your data on arrival. A clean, well-structured dataset with a clear codebook can move straight to analysis; a dataset with inconsistent coding, missing-data problems, and no documentation requires hours of preparation before any model can be run, and that preparation is real work that has to be priced. The third driver is the number of hypotheses and outcome variables, since each additional research question multiplies the analyses required. The fourth, and the one candidates most often underestimate, is the number of revision rounds, because an analysis returned by a committee for reworking generates further consulting work that the original quote may not have covered.
Red Flags When Evaluating a Statistician
Because dissertation statistical consulting is an unregulated market, the burden of evaluating competence and integrity falls on you. The flags below fall into four groups: credentials, competence for your specific design, process, and ethics.
Table 2: Red Flags When Evaluating a Statistical Consultant
Category | The red flag | What it signals |
|---|---|---|
Credentials | Cannot describe their training, their experience with doctoral work, or the designs they have handled | Asking for trust, they have not evidenced |
Competence for your design | Claims to handle every design equally well and does not ask about the specifics of yours | Inexperience or overselling; the general skill is not fit for your design |
Process | Wants to run the analysis in a black box and return a finished section without ensuring you understand it | Sets you up to fail the viva, whatever the intention |
Ethics | Offers to write your results chapter as undisclosed work, or markets confidentiality as a way to hide involvement from your committee | Proposing something that puts your degree at risk |
Fit test | Cannot immediately name the analysis your design implies and the assumptions it rests on | Not yet demonstrating command of your particular design |
The credential flags are the easiest to check. A consultant who cannot describe their training, their experience with doctoral-level work, or the kinds of designs they have handled is asking for trust they have not evidenced. But credentials alone are insufficient, because a statistician expert in clinical trials may be the wrong fit for a qualitative-dominant mixed-methods design, and one fluent in econometrics may not be the right choice for a psychometric validation study. The competence flag to watch for is generally: a consultant who claims to handle everything equally well, without asking about the specifics of your design, is either inexperienced or overselling.
The process flags concern about how the consultant works with you. A consultant who wants to take your data, run the analysis in a black box, and return a finished results section without ensuring you understand what was done is setting you up to fail your defense, whatever their intentions. The right process leaves you understanding the analysis well enough to explain it. The ethics flags are the most serious of all, and the clearest is the offer discussed above: any consultant who proposes to write your results chapter as undisclosed work, or who markets confidentiality as a way of hiding their involvement from your committee, is offering something that puts your degree at risk.
The Fit-to-Design Test
Before you engage anyone, there is a single question that separates a consultant who fits your project from one who does not. Ask them to name the specific analysis your design implies and the assumptions that analysis rests on. A consultant who fits your project will answer immediately and concretely: they will tell you that your nested data structure requires a mixed-effects model, or that your repeated-measures design calls for a particular form of analysis of variance with specific sphericity assumptions to check. They will name the assumptions without prompting.
A consultant who cannot do this, who answers vaguely or defers the question, is not yet demonstrating command of your particular design, and command of the general field is not the same thing. The assumptions matter here for a reason beyond the fit test itself: they are exactly what your examiners will probe, and a consultant who cannot articulate them cannot prepare you to defend them. The specific assumption checks that an examiner will expect to see reported are set out in a dedicated guide on the statistical assumptions your examiner will ask about, and a consultant worth engaging will treat those checks as routine rather than as an afterthought.
Confidentiality, IRB, and Handing Over a Dataset
The moment you share your dataset with a consultant, you take on obligations that exist whether or not anyone mentions them. If your study involves human participants, your ethics approval and your participants' consent almost certainly govern who may access identifiable data and under what conditions. Handing a spreadsheet of identifiable records to an outside consultant without a plan can breach both your protocol and the terms under which participants agreed to take part.
The safeguards are not complicated, but they must be in place before the data moves. De-identify the dataset before sharing it, removing direct identifiers and, where necessary, the indirect identifiers that could re-identify participants in combination. Put a data-use agreement in place that specifies what the consultant may do with the data and requires its deletion on completion. Use a secure transfer method rather than ordinary email. And check your own ethics approval to confirm that sharing de-identified data with an analyst is permitted, or seek an amendment if it is not. A consultant experienced in doctoral work will expect these safeguards and will often raise them first; one who treats the transfer of identifiable data as casual is showing you how they will treat the rest of your project.
What You Must Still Be Able to Do Yourself
Everything in this article converges on a single requirement: you must be able to defend the analysis as your own. A dissertation is a sole-authored demonstration of your capability as an independent researcher, and the analysis chapter is part of that demonstration. A statistician may help you produce it, but in the viva, you are alone, and the examiner's questions about your analysis are directed at you.
Those questions are predictable, and they are the same ones a good consultant will have made sure you can answer: why did you choose this test rather than an alternative, how did you check its assumptions, what do the results allow you to claim and what do they not, and how would the conclusions change if a key assumption were violated. The candidates who handle these well are the ones who use their consultant to build understanding rather than to avoid it. Preparing for exactly this line of questioning is the subject of a detailed treatment of the questions examiners ask about methodology and analysis, and the standard it describes is the standard a legitimate consulting relationship is designed to help you meet.
Rehearsing those answers aloud, against the specific analysis you ran, is what turns understanding into a fluent defense, and structured defense preparation targets the analytic questions your particular design is most likely to attract.
This is also why the writing of the results chapter is your work, even when the analysis behind it was supported. The chapter is where you demonstrate that you understand your own findings, and the conventions for presenting them are set out in a guide on writing the results chapter in APA style. A consultant can check that your reporting is accurate; the account of what the findings mean has to be yours.
Have you already collected the data, and need the analysis done right? |
|---|
Legitimate statistical support means the analysis is run correctly, and you are able to defend every decision in it. A specialist can clean and check your data, run the models your design requires, and work through the output with you until the interpretation is yours to explain. Tell us about your dataset and design, and you will have an itemized quote within 2 to 4 business hours, no obligation. |
Frequently Asked Questions
Is it ethical to hire a statistician for my dissertation?
Yes, within limits. Engaging a statistician for design consultation, data preparation, running analyses, and interpretation coaching is a legitimate and common form of doctoral support. The line is crossed when the help substitutes for your own understanding, most clearly when a consultant writes your results narrative as undisclosed work or when you emerge unable to defend the analysis yourself. The test is whether the help could be disclosed openly to your committee and whether you can account for every analytic decision as your own.
How much does a dissertation statistician cost?
Cost depends on the pricing model and the complexity of the work. Hourly pricing suits open-ended coaching and troubleshooting; fixed-project pricing suits a defined set of analyses. The main cost drivers are the complexity of the analysis, with multilevel models, structural equation models, and survival analyses costing considerably more than t-tests or a one-way analysis of variance; the condition of your data on arrival; the number of hypotheses and outcome variables; and the number of revision rounds. A specialist can give an itemized quote once they see your design and dataset.
When should I hire a statistician?
At the proposal or design stage, before you collect data. Several of the decisions that most benefit from statistical expertise, above all the a priori power analysis that determines your sample size, cannot be repaired retrospectively. A statistician consulted early can confirm that your design will support the analysis your research question requires; a statistician consulted after data collection can only make the best of decisions already fixed.
Would a statistician be able to write my results chapter?
A statistician can run the analyses, produce the output, and check that your reporting of the numbers is accurate. The interpretation, the account of what the findings mean and what they contribute, has to be your own work, because that is what the dissertation certifies and what you will be examined on. A results chapter written by someone else as undisclosed work is a serious integrity problem, and it also leaves you unable to defend the chapter in the viva.
What if my sample size is smaller than the power analysis required?
This is a common situation, particularly when recruitment falls short, and it is defensible if handled openly rather than concealed. The response is to acknowledge the constraint, report the achieved power, and calibrate your claims to what the sample actually supports, rather than to reverse-engineer a justification. The approach to defending a constrained sample without overstating what it can show is set out in a dedicated guide on justifying a small or constrained sample.
You May Also Find Useful
Hiring Well
Hiring a statistician is not a shortcut and should not be sold as one. Hiring a statistician properly ensures that the analysis at the center of your dissertation is correct, appropriate to your design, and defensible under examination, while the understanding of that analysis remains yours. The candidates who benefit most are the ones who engage early, who scope the work clearly, who choose a consultant whose expertise fits their specific design, and who treat the engagement as a way to build their own command of the analysis rather than to avoid it.
The decisions that determine whether the engagement helps or harms are made before the work begins: when you bring the statistician in, what you ask them to do, and whether you insist on understanding what was done. Get those right, and a statistician is one of the most useful forms of support available to a doctoral candidate. Get them wrong, and no amount of statistical skill will save an analysis you cannot defend.
Where statistical support fits within the wider arc of the dissertation, from the design that determines the analysis to the defense that tests it, is set out across the ScribeLab Writer dissertation service, which works with candidates at every stage of that arc.
If you would like statistical support that keeps the analysis correct and the understanding yours, from design and power analysis through to a results chapter you can defend, tell us where your project stands, and you will have an itemized quote within 2 to 4 business hours, no obligation.
