JH Stat Statistical Consulting

Statistical consultancy services

Support across the full research cycle — from framing the question and justifying the sample size through to analysis, publication and training your team.

Service 1

Study design and protocol development

The design decisions made in the first few weeks of a study constrain everything that follows. Getting them right is the cheapest statistical intervention available.

This service covers the work between having a research idea and having a protocol that a funder, ethics committee or journal will accept: sharpening the question into something estimable, choosing a design capable of answering it, defining outcomes precisely, and writing an analysis plan that commits to how the data will be handled.

It also covers methodological review of designs already drafted — a second opinion before submission, when there is still time to change course.

What is included

  • Research question development — turning a broad interest into a specific, answerable question with a defined estimand
  • Study design — randomised, cluster-randomised, stepped-wedge, cohort, case-control, cross-sectional, pilot and feasibility designs
  • Statistical analysis plans — pre-specified, versioned SAPs suitable for trial registries and protocol appendices
  • Outcome selection — primary and secondary outcomes, composite endpoints, and how outcomes will be measured and derived
  • Methodological review — independent appraisal of an existing protocol, design or analysis plan

Service 2

Sample size and power

Funders and ethics committees expect a sample size that is justified, not merely stated. Recruiting too few participants wastes the study; recruiting too many wastes resources and exposes people unnecessarily.

Calculations are documented in full: the assumed effect, where that assumption came from, the variability used, the design effect where relevant, and the consequences of those assumptions being wrong. The result is a section you can paste into a grant application and defend under review.

Where the design is complex — clustering, repeated measures, time-to-event outcomes, multiple comparisons — simulation is used rather than a closed-form approximation that does not fit the design.

What is included

  • Sample size calculations — for continuous, binary, count and time-to-event outcomes
  • Power calculations — including power curves showing how power varies with plausible effect sizes
  • Clustered and longitudinal designs — intracluster correlation, design effects, repeated measures and attrition
  • Feasibility and sensitivity assessments — what is achievable with the recruitment you can realistically expect
  • Simulation-based estimation — where standard formulae do not apply
  • Written justification — a defensible narrative section, not just a number

Service 3

Statistical analysis

Full analysis delivered end to end — data preparation, modelling, tables and figures, and a written interpretation your team can take straight into a manuscript or report.

Regression modelling

Linear and logistic regression as the workhorse of applied health research, with attention to functional form, interaction, collinearity and model diagnostics rather than default output.

Multilevel and mixed-effects models

For data nested within practices, wards, schools, clusters or repeated occasions — where ignoring the hierarchy would understate uncertainty.

Longitudinal analysis

Growth curves, mixed models for repeated measures, GEE, and trajectory or latent growth approaches for change over time.

Survival analysis

Kaplan–Meier estimation, Cox proportional hazards models, flexible parametric survival models, competing risks and time-varying covariates.

Count-data analysis

Poisson and negative binomial models, zero-inflated and hurdle models, and rate models with offsets for exposure time or population.

Propensity score methods

Matching, weighting and stratification for observational comparisons, with balance diagnostics and sensitivity analysis for unmeasured confounding.

Interrupted time series

Evaluating the effect of a policy, guideline or service change, accounting for underlying trend, seasonality and autocorrelation.

Difference-in-differences

Comparing change across intervention and comparison groups, with attention to the parallel trends assumption and to staggered adoption.

Machine learning where appropriate

Regularised regression, tree-based ensembles and prediction modelling — used when prediction is genuinely the goal, with proper validation and calibration rather than as a default.

Also covered as standard: data cleaning and derivation, missing data strategy including multiple imputation, subgroup and sensitivity analyses, publication-ready tables and figures, and documented analysis code handed over at the end of the project.

Service 4

Psychometrics and measurement

If the instrument does not measure what it claims to measure, no amount of subsequent modelling will rescue the study. Measurement quality is a first-order concern, not a methodological footnote.

This service covers the development and evaluation of questionnaires, scales and patient-reported outcome measures: examining structure, establishing reliability, accumulating validity evidence, and testing whether a scale behaves consistently across the groups you intend to compare.

It applies equally to instruments being newly developed, existing instruments being adapted or translated for a new population, and established measures being used in a context where their performance has not previously been demonstrated.

What is included

  • Exploratory factor analysis — dimensionality, extraction and rotation decisions, and interpretation of structure
  • Confirmatory factor analysis — hypothesised structures, model fit, modification decisions and measurement invariance across groups
  • Reliability — internal consistency, test–retest reliability, inter-rater agreement and standard error of measurement
  • Construct validity — convergent, discriminant, known-groups and criterion validity evidence
  • Scale development and validation — item generation, pilot testing, item reduction, scoring and full validation reporting
  • Item response theory — item characteristics, differential item functioning and short-form development where suitable

Service 5

Grant and funding support

Statistical weakness is one of the most common reasons applications are downgraded at review. It is also one of the most preventable.

JH Stat contributes the statistical content of funding applications: the methods section, the analysis plan, the sample size justification, and the reasoning that connects the design to the research question. The work is done collaboratively with your team, in your voice, and to the specific requirements of your funder.

Involvement early in the writing process is considerably more useful than a review the week before the deadline — but late reviews are accepted where the timetable allows.

Where a funder requires a named statistician, or where the study warrants ongoing statistical involvement, JH Stat can be included on the application as a collaborator or co-investigator. See research collaboration.

What is included

  • Statistical methods sections — written to the funder's structure and word limit
  • Analysis plans — proportionate to the stage, with detail sufficient for reviewers
  • Sample size justification — fully documented assumptions and sensitivity to them
  • Reviewer-response support — drafting statistical responses to referee and panel comments
  • Collaborative grant development — statistical input from the outline stage onwards
  • Costing input — realistic statistical time and resource for the budget

Service 6

Publications and statistical review

An independent statistical read of a manuscript before submission catches the problems that would otherwise come back from reviewers months later.

Review covers whether the analysis matches the design and the stated question, whether assumptions have been checked, whether the results are described accurately, and whether the conclusions are proportionate to what the data actually show. Overstated claims are flagged as clearly as technical errors.

Where a paper has already been reviewed, JH Stat can help draft statistical responses to referees and carry out any additional analyses requested.

What is included

  • Manuscript statistical review — methods, results and conclusions assessed before submission
  • Results interpretation — what the estimates mean substantively, and what they do not support
  • Reporting guidance — alignment with CONSORT, STROBE, PRISMA, TRIPOD, COSMIN and other EQUATOR guidelines
  • Reviewer-response support — drafting statistical replies and running requested additional analyses
  • Reproducibility checks — re-running the analysis independently to confirm reported figures
  • Tables and figures — clear, journal-ready presentation of results

Service 7

Training and workshops

Applied statistical training for research teams, delivered in software they will actually use and built around examples from their own field.

Sessions are practical rather than theoretical: participants work through analyses during the workshop and leave with annotated code and materials they can reuse. Delivery is available online or in person, as a single session, a short course, or a programme spread across a term.

See full training options

Topics covered

  • R — from first principles through to modelling and reproducible reporting
  • SPSS — applied analysis with syntax for reproducibility
  • Stata — data management, modelling and do-file workflows
  • Regression — linear, logistic and beyond, with diagnostics and interpretation
  • Multilevel modelling — clustered and hierarchical data in practice
  • Survival analysis — time-to-event methods for applied researchers
  • Applied statistics for healthcare researchers — a grounding for clinical and health services teams

A note on academic integrity

JH Stat supports researchers with methodology, analysis and learning. It does not complete assessed academic work dishonestly.

Doctoral and postgraduate researchers are welcome, and statistical advice, methodological training and analytical support are all legitimate forms of research supervision support — the same kind that university statistical services and supervisors provide routinely. What JH Stat will not do is write a thesis, produce work to be submitted as a student's own unaided effort, or provide any support that breaches an institution's academic integrity or research degree regulations.

Students are encouraged to discuss any external statistical support with their supervisor before it begins, and JH Stat is happy to correspond directly with a supervisor or department to confirm the scope of what has been provided.

Working together

How engagements work

What does a typical engagement look like?

Most begin with a short scoping conversation, which can be arranged to establish what you need and whether JH Stat is the right fit. A written proposal follows, setting out scope, deliverables, timescale and fees. The scope, timescale and fees are agreed in writing before any chargeable work begins.

How is work priced?

Pricing depends on the scope and the type of organisation. Short pieces of work are usually quoted as a fixed fee; longer collaborations may be arranged as a day rate, a retained block of time, or costed into a grant. A quotation is always provided in writing before work begins.

How are timescales agreed?

Timescales depend on the scope of the work, the state of the data and the complexity of the design, and on current availability. A realistic timescale is discussed at the scoping stage and set out in the written proposal. If you are working to a grant or submission deadline, please say so in your enquiry so it can be taken into account.

How is data handled?

Wherever possible, work is carried out on de-identified data or within your own secure environment. Data handling follows UK GDPR and the Data Protection Act 2018. Data-sharing agreements, non-disclosure agreements and institutional information governance requirements are welcomed and complied with, and are agreed in writing before any data are transferred.

Will JH Stat be an author on the resulting paper?

That depends on the nature of the contribution and is agreed at the outset. Where the statistical contribution meets recognised authorship criteria, such as those set out by the ICMJE, authorship is expected. Where it does not, acknowledgement is appropriate. Either way, the arrangement is written down before work starts rather than negotiated afterwards.

Can JH Stat join a grant application?

Yes. JH Stat can be named on applications as a collaborator or co-investigator, with statistical time costed into the budget. This works best when the conversation starts at the outline stage rather than close to the submission deadline.

Not sure which service you need?

Describe the study and the problem in a few sentences. If JH Stat is the right fit, you will get a clear proposal; if not, you will get an honest answer and, where possible, a pointer to someone better placed to help.

Discuss Your Project Email [email protected]