When Should You Bring in a Statistician, Methodologist, or Other Specialist?

Should You Ask for Specialist Help Now or Wait Until You Actually Need It?

Your research is becoming more complicated than expected. Perhaps you are unsure which design can answer the question, how large the sample should be, whether your planned analysis fits the data, how to develop an instrument, how to manage a complex dataset, or how to integrate qualitative and quantitative evidence.

You know that a statistician, methodologist, data scientist, qualitative researcher, psychometrician, programmer, laboratory specialist, or another expert might help. But when should that person become involved?

Researchers sometimes wait until the difficulty becomes unavoidable. For statistical work, that often means contacting someone after data collection with a spreadsheet and asking which test to run. By then, however, the most important problem may no longer be the analysis. The study may have collected too few observations, omitted a necessary variable, used an unsuitable measurement schedule, or produced data that cannot answer the original question.

The best time to involve a specialist therefore depends less on when you become stuck and more on when that specialist's expertise could materially improve a decision that will later become difficult or impossible to change.

Bring in Specialist Expertise Before the Decisions It Could Change Become Irreversible

How to Recognize When Your Study Needs Specialist Expertise

Do not wait until you have a statistical problem

A statistician's role is broader than choosing a test after the dataset has been collected. Statistical considerations can influence the research question, study design, sampling strategy, randomization, sample size, measurement schedule, data structure, data management, analysis plan, interpretation, and reporting.

Recent guidance on collaboration with biostatisticians emphasizes involvement throughout the research process, from study design through interpretation and dissemination. Early involvement can help researchers develop hypotheses, determine sample-size requirements, choose appropriate data structures, and anticipate analytical requirements before data collection begins.

The same principle applies beyond statistics. A qualitative methodologist may identify problems with sampling or interview design before the first participant is recruited. A psychometrician may influence how a construct should be measured. A data scientist may recognize that the proposed data architecture cannot support the intended analysis. A laboratory specialist may identify procedural requirements that affect the entire protocol.

Specialists are most valuable when they can still influence the research rather than merely diagnose what went wrong.

Consult when you are uncertain whether the design can answer the question

If you are unsure whether the proposed methodology actually addresses the research question, consultation should occur early.

This can arise when choosing between experimental and observational approaches, deciding whether repeated measurements are necessary, determining how comparison groups should be constructed, designing a mixed-methods project, planning a complex survey, or working with hierarchical or longitudinal data.

Study-design consultation services at research institutions commonly support exactly these decisions. Research methodologists may advise on design, sample size, measurement, data collection, and analytical strategy before a project begins.

A methodologist is particularly useful when the uncertainty concerns the logic connecting the research question, design, evidence, and inference rather than one isolated statistical technique.

Consult before calculating a consequential sample size you do not understand

Sample-size planning can be deceptively technical.

Simple designs may permit relatively straightforward calculations. More complex studies can require assumptions about expected effects, variability, event rates, clustering, repeated measurements, attrition, allocation ratios, precision, or other design features.

A sample-size calculator can produce a number even when the inputs or underlying model are inappropriate. The resulting precision is mathematical rather than methodological.

If the sample-size requirement materially affects whether the study is feasible and you are uncertain about the assumptions or calculation, seek statistical or methodological input before committing to recruitment targets.

This is particularly important because an inadequate sample may not be repairable after data collection, while an unnecessarily large target can make an otherwise feasible study needlessly expensive or time-consuming.

Consult when the data structure is more complicated than the planned analysis

Some analytical problems are visible before a single observation is collected.

Students nested within classes, patients within hospitals, repeated observations within individuals, matched participants, multiple outcomes, time-to-event data, complex survey designs, spatial data, and linked administrative records all have structures that can affect analysis.

If your current analytical plan ignores those features because you do not know how to handle them, the appropriate response is not necessarily to simplify the dataset. It is to determine what the design requires and whether specialist expertise is needed.

This connects directly to the question of what to do when the appropriate analysis is beyond your current skills. An unfamiliar method is often manageable when recognized early enough for learning or collaboration.

Consult when the analysis could change what data you need to collect

This is one of the clearest signals that specialist involvement should occur before data collection.

Suppose the intended analysis requires baseline measurements, a particular outcome definition, repeated observations, information about clustering, variables needed to address confounding, dates required for survival analysis, or items needed to calculate a scale. If those elements are absent, the analysis may become impossible or substantially weaker.

A statistician or methodologist reviewing the plan before collection can identify such requirements while they can still be added to the protocol.

After data collection, the conversation changes from "What should we measure?" to "What can we salvage from what we measured?" Those are not equivalent positions.

Consult when measurement itself is specialized

Not every specialist problem concerns analysis.

If your study depends on developing or adapting a questionnaire, measuring a latent construct, translating and validating an instrument, constructing a composite scale, calibrating a laboratory measure, or implementing a specialized assessment, appropriate expertise may be needed during measurement design.

A psychometrician, measurement specialist, domain expert, survey methodologist, linguist, laboratory scientist, or other specialist may be more relevant than a statistician depending on the problem.

The central question is: What expertise is necessary for this particular methodological decision?

Do not automatically send every research problem to the statistician. They may be delighted to discover that the problem belongs to someone else.

Consult when the sampling or recruitment problem is methodologically unusual

Sampling can become specialized when populations are rare, hidden, clustered, geographically dispersed, institutionally protected, or difficult to enumerate.

A researcher considering respondent-driven sampling, time-location sampling, complex survey sampling, multistage sampling, adaptive sampling, or another specialized strategy may benefit from expertise before recruitment begins.

The specialist may also help determine whether the proposed sampling method supports the type of population inference the research question implies.

If the problem is primarily that the population is rare or difficult to reach, methodological consultation may need to occur alongside engagement with organizations or communities that understand the population itself.

Consult before building a complex data system

Data-management problems can become methodological problems.

A study collecting repeated observations from multiple sites may require consistent identifiers, linkage rules, validation checks, variable definitions, audit trails, access controls, and procedures for handling corrections. Large observational datasets may require programming, database design, or reproducible workflows that exceed the research team's technical expertise.

Research methods cores at institutions such as Duke explicitly combine expertise in biostatistics, epidemiology, data science, qualitative methods, study design, implementation, and interpretation because these functions often interact.

If the integrity of the eventual analysis depends on how data are structured or processed, involve the relevant data specialist before an improvised spreadsheet becomes the study's unofficial database.

Consult when using unfamiliar secondary or administrative data

Existing datasets can contain analytical complications that are not obvious from the variable list.

Complex survey weights, clustering, repeated measures, linkage, censoring, provider-specific coding, derived variables, missingness, and changes across years can affect the analysis. Registry-based research may similarly require understanding both the clinical or substantive context and the statistical structure of the data.

Collaboration between domain researchers and biostatisticians is particularly valuable in these settings because neither substantive nor statistical expertise alone may be sufficient to interpret the data-generating process correctly.

Before committing to the analysis, make sure you have also evaluated whether the existing dataset is suitable for the proposed study.

Consult when qualitative methodology is unfamiliar or consequential

Research consultation is not synonymous with statistical consultation.

A qualitative methodologist may be useful when choosing among qualitative approaches, developing a sampling strategy, designing interview or focus-group protocols, planning iterative data collection and analysis, addressing reflexivity, developing an analytical process, or deciding how methodological claims should be supported.

Similarly, mixed-methods research may require expertise in how qualitative and quantitative components are sequenced, connected, or integrated rather than simply expertise in each component separately.

Institutional methods cores increasingly recognize this breadth. Duke's Biostatistics, Epidemiology, and Research Design Methods Core, for example, connects investigators with quantitative and qualitative methodologists, data scientists, epidemiologists, and other relevant specialists rather than treating methodological support as statistics alone.

Consult when you cannot explain why your planned method is appropriate

A useful diagnostic question is whether you can explain the proposed method without relying on software labels or phrases copied from another paper.

Why is this design appropriate? Why this sampling strategy? Why this model? What assumptions matter? What information does it estimate? What would make its interpretation misleading?

You do not need to know every technical detail before seeking help. That would rather defeat the purpose of consultation. But if the methodological justification is essentially "other studies used it" or "the software can run it," specialist input may be warranted.

Consult when disagreement among advisers reveals genuine methodological uncertainty

Researchers sometimes receive contradictory advice: one supervisor recommends regression, another structural equation modeling; one reviewer wants a larger sample, another says the design is the real issue; one committee member recommends phenomenology, another suggests grounded theory.

Not every disagreement requires another expert. Research methods often permit more than one defensible approach.

Consultation becomes useful when the disagreement reflects expertise outside the team's competence or when choosing incorrectly would materially affect the validity, feasibility, or interpretation of the study.

A specialist should clarify the methodological trade-offs rather than merely serve as a tie-breaking vote.

Bring specialists in before irreversible decisions

The practical principle can be stated simply: consult before the decision becomes expensive to change.

Not every specialist needs to become a long-term collaborator

The appropriate level of involvement depends on the problem.

A short consultation may be enough to confirm a straightforward sample-size calculation, review an analysis plan, or identify appropriate methodological resources. More complex projects may require repeated consultation. Highly specialized or methodologically intensive research may benefit from a specialist becoming a formal collaborator throughout the study.

The boundary is not always sharp, and institutional arrangements vary. What matters is that the level of involvement matches how much the study actually depends on the expertise.

Ask for collaboration when the expertise is central to the study

If a specialist is expected to design substantial parts of the methodology, develop the analysis, write or validate code, interpret complex results, troubleshoot throughout the project, and contribute to reporting, a one-hour consultation may be insufficient.

Long-term methodological partnerships are common in complex research settings. Duke's BERD Methods Core, for example, describes methodologists as forming long-term partnerships with investigators and contributing to design, analysis, methodological implementation, and interpretation.

Plan that involvement explicitly. Determine responsibilities, expected effort, timelines, communication, funding where relevant, and appropriate recognition or authorship according to the specialist's actual contribution and applicable authorship standards.

Do not assume specialist time is immediately available

Expertise is a resource, and resources have capacity limits.

A statistician may have a waiting list. A methods core may prioritize funded projects. A laboratory specialist may be available only during certain periods. A consultant may charge fees. Your institution may provide a limited number of consultation hours.

Research on biostatistical service groups has long recognized that specialist time must be allocated among competing projects. Contemporary institutional methods cores similarly use formal intake and collaboration processes to match projects with available expertise.

If your study cannot proceed without a particular specialist, investigate availability early rather than assuming that help will appear during the week you need it.

Bring something concrete to the consultation

A specialist can help more effectively when the research problem is sufficiently developed to discuss.

Depending on the stage, useful materials may include the research question, study aims, proposed design, target population, variables or measures, sampling plan, expected sample size, data structure, relevant protocol sections, dataset documentation, preliminary data, timeline, and a clear description of what you are uncertain about.

You do not need a finished analysis plan before consulting the person who is supposed to help develop it. You do need enough information for the specialist to understand the scientific problem.

Institutional consultation services commonly ask investigators to describe project aims, design, anticipated completion dates, and the kind of methodological support requested before assigning a collaborator.

Ask a scientific question, not just a software question

"How do I run this in SPSS?" may not be the most important question.

Explain what you are trying to learn from the data, how the data were generated, and what decision or inference the analysis is supposed to support. The specialist can then determine whether the method you had in mind is appropriate at all.

A useful consultation often changes the question from "Which button do I press?" to "What analysis corresponds to the structure of this research problem?"

Do not conceal uncertainty from the specialist

Researchers sometimes arrive at consultation wanting confirmation of a methodology already written into a proposal.

That can limit the value of the conversation.

If you are uncertain about the design, say so. If the sample is constrained, explain the constraint. If a variable is an imperfect proxy, disclose it. If the data have already been collected, make clear which decisions can no longer change.

The specialist needs the actual problem, including its inconvenient parts, to provide useful advice.

Consultation does not transfer responsibility for the research

A statistician or methodologist can provide expertise, but the research team still needs to understand the scientific logic of the study and take responsibility for the resulting work.

You should be able to explain why the method addresses the research question, what the important assumptions and limitations are, and what the results do and do not support. The specialist may have deeper technical expertise, but methodological understanding should not disappear at the boundary between collaborators.

This is particularly important for student research, where demonstrating understanding is usually part of the educational purpose of the project.

AI tools are not a substitute for specialist judgment

AI tools can explain statistical concepts, generate code, suggest analytical strategies, critique questionnaires, and help researchers prepare for consultations. They may reduce the number of routine questions that require specialist time.

They can also recommend inappropriate methods, generate incorrect code, invent assumptions, or provide plausible explanations for analyses that do not fit the design.

If the reason you need a specialist is that nobody on the project can verify a consequential methodological decision, asking an AI system does not remove that problem. It merely introduces another source whose output must itself be evaluated.

Specialist involvement must fit the budget and timeline

Some expertise is available through supervisors, departments, institutional methods cores, libraries, research offices, or collaborative networks. Other support may involve consultation fees, staff effort, specialized software, laboratory services, or formal collaboration.

Include these requirements when assessing feasibility. A method may be technically possible but unrealistic if the only available expert cannot join the project until after your deadline or if the necessary service exceeds the available budget.

This is why the broader question of whether the study is actually feasible should include expertise alongside participants, data, time, money, facilities, and access.

The Best Time to Consult a Statistician Is Before the Dataset Exists

Had the student waited until the dataset was complete, the statistician might still have been able to analyze it. But if essential cluster information, measurements, or an adequate sampling structure had been omitted, no advanced model could recreate them afterward. The early consultation protected the study before the analytical problem became irreversible.

Common Mistakes When Seeking Research Specialist Support

Ask for Help at the Point Where Expertise Can Still Change the Study

Review your research workflow and identify decisions that exceed the competence currently available within the project. Then ask when each decision becomes difficult to reverse.

That date, rather than the moment you finally become stuck, is often the useful deadline for consultation.

The purpose of consultation is not to make a study look more sophisticated. It is to obtain expertise at the point where it can protect the validity and feasibility of the research. Sometimes that requires one conversation. Sometimes it requires another researcher at the table from the beginning.

Do You Need a Statistician, Methodologist, or Other Specialist?

Frequently Asked Questions About Research Specialist Consultation

Ask for Expertise Before You Need It to Repair Something

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.