Job search
Biostatistician Jobs: Match Study Design, Analysis Responsibility, and the Required Level
Compare guided analysis with research consultation and study-level ownership, using employer-specific requirements and two completed contribution examples.
Biostatistician jobs differ in how much responsibility they assign for the research question, analysis plan and interpretation. A guided analyst may implement a plan reviewed by a senior statistician. A more independent statistician may help investigators decide the plan and explain its limitations. Compare those responsibilities before treating a job numeral, a degree or proficiency in R as a complete description of the level.
LandOffer publishes this guide for technical candidates evaluating a move into research statistics. We reviewed public employer descriptions on October 22, 2026, without applying, handling patient information or running a clinical analysis. Use LandOffer’s recent roles for discovery, then establish the actual duties and current availability on the employer’s original page.

Distinguish a job classification from an available opening
Two useful reference descriptions are Duke’s Biostatistician I and Biostatistician III classifications. Both carry an April 1, 2022 revision date. They explain the institution’s levels; the pages explicitly direct readers elsewhere to find current openings. A classification remaining online does not mean that a team is recruiting into it.
For an actual vacancy example, Stanford’s Biostatistician 2, job 200568 rendered a description and Apply control on our review date. It identified a hybrid, regular, full-time role. We observed the displayed description and Apply control; we did not submit an application or verify remaining headcount or how long review would continue.
Freshness checks changed which results belonged in this guide. A separately indexed Stanford Biostatistician 1 destination displayed that the job was no longer available. Brown and ICON destinations returned missing-page messages, and an older Cedars-Sinai URL returned 404. We excluded those results from current vacancy examples rather than relying on their search summaries or browser-extension availability badges.
Save the institution, requisition identifier, source date and availability separately from the title. When a lead disappears, its older duties may provide historical context, but they should not remain in your active application list as an open position.
Guided analysis still requires statistical work
Duke’s level-I classification describes basic analysis and programming with support from lead statisticians. It includes working from an analysis plan, documenting results, developing dataset specifications under guidance and checking statistical code and output. Its education description names a relevant bachelor’s degree, with no experience required beyond education. Those are this classification’s requirements, not a universal entry rule or a decision about a particular applicant.
Guidance identifies a working relationship: someone reviews the question, method and interpretation while the analyst develops practical experience. It does not imply that the analyst simply exports a spreadsheet or has no responsibility for detecting an inconsistent result.
Read a junior description for the actual support arrangement. Who reviews the analysis plan? Who checks a new method? How is a data discrepancy raised? “Collaborative” can describe many organizations; a named supervision relationship and concrete review responsibilities tell you more about the work you would learn to perform.
For a technical candidate, maintainability and reproducibility are useful contributions. An interviewer can discuss how you identified the input version, documented a changed variable or traced a discrepancy back to the rows that produced it. State the source of the statistical instructions alongside those engineering actions. Implementing a reviewer’s method accurately is credible work without being described as independent methodological leadership.
More independence does not remove every review boundary
Duke’s level-III classification assigns independent project coordination, preparation and interpretation of analyses, and collaboration with investigators. Yet it still describes sample-size, power and interim-stopping work with guidance. The education alternatives include a relevant doctorate with no relevant experience, a master’s plus two years or a bachelor’s plus four, alongside other relevant-work requirements. The same classification also requires relevant analysis or written research contributions and thorough SAS experience. A degree alternative does not erase those requirements or guarantee appointment.
This mixed description is useful: independence can vary by task within one role. Someone might own routine project delivery and guide less experienced programmers while consulting a senior colleague on a particular design issue. A title alone cannot tell you that every methodological decision is made without review.
Ask for a concrete example of a decision the role makes and one it escalates. Distinguish proposing a method, obtaining agreement, implementing it and approving the final interpretation. An honest account can show substantial ownership while retaining a collaborator’s contribution. It becomes less convincing when every decision in a team project is quietly attributed to one person.
Clinical collaboration also involves explaining assumptions and limitations to people who have different expertise. Duke’s collaborative-careers description places study design, implementation, interpretation and manuscript work within a shared research process. That organizational description is not an individual job offer or a universal allocation of authority.
Compare levels within their own employer context
Stanford’s current level-2 example combines some independence in investigator consultation with collaboration involving senior statisticians. It names a relevant master’s degree and three years of experience, plus proficiency in at least two of R, SAS, SPSS or STATA. Its listed work spans study design, analysis, communication and oversight of lower-level staff on analysis-file quality. Those requirements belong to requisition 200568.
Do not build a cross-employer ladder by putting Duke I, Stanford 2 and Duke III in numerical order. Their duties, qualifications and classification systems are different. Compare the work: which questions you would help define, which decisions you could make, what review remains and which outputs you must defend. Record work location and employment duration alongside the responsibilities, so an otherwise relevant role does not disappear from view when you discover a schedule conflict.
Tool overlap is only one part of the match. A candidate can write reliable R code yet have limited experience discussing study assumptions with investigators. Another may understand a research question but need substantial help organizing analysis code. Identify both strengths and gaps rather than using a software list to stand in for the full role.
Worked example: revise a contribution statement before choosing a level
Mara is fictional. Her stipulated background includes a statistics degree and an R course project using a fully synthetic administrative dataset. A course instructor supplied the question and analysis instructions. Mara cleaned the inputs, implemented the specified analysis, checked discrepancies and wrote a short explanation reviewed by the instructor. No actual health data, employer project or hiring decision is represented.
Her initial profile says, “Independently designed and led a biomedical study.” The completed responsibility ledger shows why that sentence exceeds the example:
| Responsibility | What happened in the fictional project | Accurate contribution |
|---|---|---|
| Define the research question | Instructor supplied it | Worked from a supplied question |
| Choose the analysis approach | Course instructions specified it | Implemented a specified approach |
| Prepare the analytic input | Mara documented her cleaning changes | Owned the described preparation work |
| Investigate discrepancies | Mara identified inconsistent counts and asked for review | Investigated and escalated the discrepancy |
| Interpret the result | Instructor reviewed her draft explanation | Wrote an explanation with review |
| Design a clinical study | No clinical study occurred | No clinical-design experience claimed |
| Employer selection | No application or interview occurred | No eligibility or hiring conclusion |
Her revised paragraph reads: “In a statistics course project using synthetic administrative data, I prepared the analysis inputs and implemented the approach supplied in the course instructions in R. I documented transformations, investigated inconsistent counts and revised my explanation after instructor review. The work was a learning project, not an independently designed biomedical study.”
Her completed route decision reads: “I will compare guided analysis roles with the work I can explain and examine each vacancy’s full qualifications. For roles asking for independent study consultation, I need stronger examples of helping define the question and justify an analysis choice. I will not add that responsibility to my history because I implemented code.”
This decision organizes her search without determining admission or employability. It also identifies a useful development question: how to gain appropriately supervised research collaboration. It does not imply that a course, certificate or rewritten paragraph guarantees a position.
Separate statistical programming from deciding the analysis
A programming-oriented role may emphasize implementing specifications, reviewing code and preparing reproducible outputs. A biostatistician description may also ask why the question is posed that way and how the proposed analysis answers it. The boundaries can overlap; read who owns the method rather than inventing a universal division between titles.
Duke’s analysis-planning resource connects planning with the research question, collection and methods. For career evaluation, that suggests a practical distinction: an implementation story explains what your program did; a design story also explains why that analysis was chosen and what limitations affected the conclusion.
Keep the reviewer visible in both stories. If an investigator supplied a definition, say so. If you proposed a change and another statistician approved it, distinguish the proposal from approval. When a discrepancy remained unresolved, explain the handoff instead of manufacturing a clean final result. These are preparation principles, not instructions for analyzing an actual clinical study.

Worked example: a denominator question becomes a bounded handoff
Idris is a fictional analyst practicing how to explain a discrepancy. All counts concern an invented administrative follow-up exercise, not patients, clinical outcomes or a treatment comparison. No statistical software, study protocol or missing-data procedure was executed. His task is to identify the unresolved question and show the limits of his own responsibility.
The exercise contains eighty synthetic records. Sixty have a recorded yes-or-no response, including forty-eight yes responses; twenty have no response recorded. A draft description calls the result an eighty-percent rate across the whole file. Idris completes this record before accepting that wording:
| Item | Completed fictional finding | Handoff consequence |
|---|---|---|
| Total records | 80 | Keep the file count visible |
| Recorded responses | 60 | Do not silently use this subset for a whole-file claim |
| Recorded yes responses | 48 | Preserve the same numerator in the comparison |
| Missing responses | 20 | Missing does not automatically mean no |
| Fraction among recorded responses | 48 divided by 60 equals 80% | Describes the observed subset only |
| Recorded-yes fraction of the full file | 48 divided by 80 equals 60% | Different denominator; not a substitute clinical estimate |
His completed handoff reads: “The draft’s eighty-percent wording uses the sixty recorded responses, while its sentence refers to all eighty records. Twenty responses are missing. The two displayed fractions answer different descriptive questions; neither establishes an effect or resolves how missing responses should be handled. Please confirm the intended question and approved denominator before the whole-file wording is released. I have not classified missing responses as no.”
His revised ownership statement reads: “I identified a mismatch between a description and its denominator, documented the counts and requested clarification. I did not determine a study endpoint, choose a missing-data method or approve the final interpretation.”
The completed result is an unresolved-question record with a release hold on that sentence. It is not a validated analysis plan or evidence of clinical statistical leadership. For an interview, Idris can explain his detection and communication clearly while naming the methodological decision that remained with the responsible reviewer. If a role assigns analysis decisions to the applicant, Idris would need separate examples showing that responsibility; this discrepancy record does not supply them.
Choose one responsibility gap before adding another credential
Take one current employer description and pair each major responsibility with an example you can explain. Mark whether you implemented, proposed, reviewed or approved the work. Keep the actual education and experience wording attached, and have the employer resolve uncertainties about its requirements. Do not turn this exercise into a personal hiring determination.
If your examples mainly show dependable implementation, compare guided analysis and adjacent programming work while developing research collaboration. If they show analysis-plan decisions, investigator discussion and interpretation under appropriate review, explain those decisions directly. Browse recent roles on LandOffer.ai, save one original posting, write a paragraph identifying your contribution and reviewer, and note one unanswered question about the role’s supervision arrangement.
Sources and Further Reading
- Duke Biostatistician I classification — revised April 1, 2022; role framework, not an opening.
- Duke Biostatistician III classification — revised April 1, 2022; task-specific guidance and qualification alternatives retained.
- Stanford Biostatistician 2, job 200568 — actual employer page checked October 22, 2026; displayed state may change.
- Duke collaborative biostatistics careers — organizational description of research collaboration.
- Duke analysis-planning resource — context for the relationship between a question and an analysis plan.


