Statistical Analysis & Study Design
- Study design & analysis planning
- Statistical analysis & interpretation
- Independent statistical review
Connect your research question, study design and analysis
Sound statistical work starts with the question you want to answer and the way your data were collected. We help researchers plan studies, analyse data and interpret results, with methods that reflect the design, the available evidence and the limits of what can be concluded.
Bring us an early study concept, a dataset ready for analysis, an existing analysis that needs checking, or statistical comments from reviewers.
Make the pattern and the uncertainty visible
Kaplan–Meier survival estimates for two groups over 36 months, with 95% confidence bands, censoring marks and the numbers at risk.
From research question to defensible conclusions
The question guides the design and analysis
Research question
What effect, difference or relationship matters?
Study design & data
What was measured, in whom, and how?
Statistical approach
Which methods fit the question and the data structure?
Interpretation
What do the estimates and uncertainty allow you to conclude?
Assumptions, data quality and limitations considered throughout
Support before, during or after analysis
You can commission a defined analysis, seek advice on a particular decision, or involve us throughout the study. We agree the scope around your research question and the stage you have reached.
Plan the study
Clarify the question, outcomes and comparisons before data collection. Develop the design and analysis plan, and assess the sample size or precision needed to address the study aims.
Analyse and interpret
Turn your data into a documented analysis, with appropriate models, diagnostic checks, clear figures and an interpretation of the findings in their scientific context.
Review and resolve
Get an independent assessment of an existing analysis, investigate unexpected results or address statistical questions raised by co-authors, internal reviewers or journal referees.
Some of our clients
What we can help with
The method follows the research problem. We consider the outcome, study design, sampling, dependence between observations and intended interpretation before choosing an analytical approach.
Study design & analysis planning
We help translate research aims into defined outcomes, comparisons and an analysis plan. This includes identifying the unit of analysis, the timing of measurements and design features that the analysis must accommodate.
- Experimental and observational study designs
- Randomisation, controls, pairing and repeated measurements
- Primary and secondary analyses, covariates and missing-data plans
Sample size, power & precision
We help assess the sample size needed for the intended analysis or the precision a feasible sample may provide. Calculations are tied to explicit assumptions about effect sizes, variability, design and anticipated data loss.
- Power calculations for specified effects and designs
- Sample-size planning based on estimation precision
- Sensitivity to assumptions, attrition and clustering
Statistical analysis & modelling
We select and fit methods suited to the scientific question and the structure of the data. We can develop a new analysis or extend an existing one, with attention to model assumptions, complexity and interpretability.
- Regression, generalised linear and mixed-effects models
- Multivariate analysis and relationships among variables
- Time-series methods, trends and temporal dependence
Uncertainty, sensitivity & robustness
We examine uncertainty in the estimates and how much the conclusions depend on analytical choices. Where appropriate, we compare defensible alternatives and explain which findings remain stable and which require caution.
- Confidence intervals and uncertainty in estimates or predictions
- Sensitivity to model specification, exclusions and missing data
- Diagnostics, influential observations and resampling methods
Independent review of existing analyses
We review the design, methods, code and reported results to identify errors, unsupported interpretations or inconsistencies. Where data and code are available, the review can include reproducing key results and checking the computational steps.
- Suitability of methods and treatment of data structure
- Consistency between code, tables, figures and conclusions
- Clear distinction between what was checked and what could not be verified
Reviewer comments & statistical revisions
We help interpret statistical criticism, determine which concerns require clarification or further analysis, and develop a reasoned response. Any additional work is considered in the context of the original design and research aims.
- Methodological explanations and statistical justifications
- Targeted reanalysis and checks requested during review
- Consistent revisions to methods, results and responses
Statistical analysis in practice
Explore how different analyses address scientific questions, from patient outcomes to environmental change. All figures on this page use synthetic data.
Medical research
Estimating how treatment differences evolve over time
Does a treatment change the trajectory of a biomarker, and when do the groups diverge? A mixed-effects model accounts for repeated measurements within participants and estimates a treatment difference that varies over time. The paired panels show both the trajectories and the contrast of direct scientific interest.
Analytical approach: Mixed-effects modelling with time-varying treatment contrasts
Medical research
Interpreting relative and absolute effects
How large is a difference, and what does it mean in practical terms? Presenting risk ratios alongside absolute risk differences helps distinguish relative changes from their consequences at the observed baseline risk. Both views include uncertainty, supporting an interpretation that goes beyond whether a threshold is crossed.
Analytical approach: Risk ratios and risk differences with confidence intervals
Ecology
Connecting multivariate patterns to individual measurements
Which measurements explain the main patterns across sampling sites? Principal component analysis summarises correlated variables, while the accompanying correlation heatmap shows which measurements contribute to each component. Together, the panels connect an overview of the data with an interpretable account of the underlying structure.
Analytical approach: Principal component analysis and variable–component correlations
Medical research
Assessing survival while accounting for patient characteristics
How does survival differ after accounting for patient characteristics? A fitted Cox model connects adjusted survival curves with estimates for treatment and clinical predictors. This example illustrates covariate adjustment and interpretation of relative hazards, with uncertainty shown for the model coefficients.
Analytical approach: Multivariable Cox proportional-hazards regression
Medical research
Testing whether a biomarker improves clinical prediction
Does a biomarker improve prediction beyond existing clinical information? Two models are evaluated on an independent dataset, assessing both their ability to distinguish outcomes and the agreement between predicted and observed risks. This makes model performance more informative than a single accuracy statistic.
Analytical approach: Logistic regression, independent validation, discrimination and calibration
Ecology
Understanding nonlinear relationships and interactions
How does an environmental relationship change when another predictor varies? A nonlinear count model captures curvature and interaction between temperature and nutrient concentration. A response surface and conditional curves communicate the combined relationship and the uncertainty in expected counts.
Analytical approach: Poisson regression with curvature and interaction terms
Energy & environmental research
Distinguishing an intervention-associated change from an existing trend
Did energy use change after an efficiency upgrade, beyond the existing trend and seasonal cycle? Segmented regression estimates changes in level and slope while accounting for serial correlation. The second panel isolates the difference from the projected baseline and shows how its uncertainty evolves.
Analytical approach: Segmented seasonal regression with autocorrelated errors
Medical research
Updating event-risk estimates as biomarker measurements accumulate
How does a patient’s biomarker history inform their subsequent event risk? A joint model links repeated measurements to time-to-event outcomes while accounting for measurement error and differences between patients. The example updates the same patient’s conditional survival prediction at months 12 and 24, using only the measurements available at each prediction time.
Analytical approach: Joint longitudinal and time-to-event modelling with a shared random intercept
Medical research
Comparing skewed biomarker concentrations across treatment groups
How large are the differences between treatment groups when biomarker concentrations are strongly right-skewed? A linear model of log-transformed concentrations estimates geometric-mean ratios for three planned comparisons with the control group. Individual observations and boxplots show the underlying distributions, while simultaneous confidence intervals and adjusted p-values account for the multiple comparisons.
Analytical approach: Log-transformed linear modelling with Bonferroni-adjusted contrasts
Ecology
Separating nonlinear environmental relationships from regional variation
How does vegetation productivity vary with elevation when sampling sites are grouped within regions? An additive mixed model fits separate nonlinear relationships for forest and grassland habitats while accounting for regional differences. The paired panels distinguish the habitat-specific response curves from estimated regional deviations, with uncertainty shown for both.
Analytical approach: Gaussian additive mixed modelling with penalised splines and regional random intercepts
Hydrology & environmental research
Forecasting seasonal dynamics with explicit uncertainty
How can future river discharge be predicted from a noisy seasonal record? A Bayesian state-space model uses six years of monthly observations to estimate annual seasonality and a changing underlying level. The forecast combines uncertainty in the model and latent state with future process and observation variability, providing predictive intervals for discharge over the following year.
Analytical approach: Bayesian seasonal state-space modelling with posterior predictive forecasting
Have a similar question about your data?
Discuss your analysisResults you can understand, check and use
We agree the outputs before work begins. Depending on the assignment, these can range from written advice on a specific question to a complete analysis with code, figures and a scientific interpretation.
An analysis or study-design plan
A written account of the research questions, outcomes, methods and assumptions, with sample-size or precision calculations where included in the scope.
Documented analyses and code
Analysis scripts documenting the agreed data preparation, modelling and diagnostic steps, so that the work can be inspected, repeated and updated.
Clear tables, figures and interpretation
Outputs that communicate estimates, uncertainty and relevant comparisons, together with an explanation of the findings, their limitations and their scientific meaning.
A review report or revision support
Prioritised findings from an independent review, recommendations for further work, or help explaining methods and addressing statistical comments in a manuscript or reviewer response.
How we work with you
Share the question and context
Tell us what you want to establish, how the study is designed and what data are available. Include relevant protocols, variable descriptions, existing analyses or reviewer comments, and your deadline.
Agree the approach and scope
We clarify what can be addressed with the available information and agree the analysis or review, deliverables, fee and timetable. Any important limitations are discussed at this stage.
Analyse, check and discuss
We carry out the agreed work, examine relevant assumptions and diagnostics, and discuss questions that arise from the data or design. Changes to the scope are agreed with you.
Deliver and explain
We provide the agreed outputs and explain how to interpret them. Where included, we help incorporate the findings into your report, manuscript or response to reviewers.
Practical questions
Can you help before data collection begins?
Yes. Early involvement allows us to examine the research question, design, outcomes and analysis plan while changes are still possible. We can also help explore sample-size requirements and the trade-offs between feasible designs.
Can you work with data that have already been collected?
Yes. We assess which analyses the design and available data can support, identify data-quality issues and explain any limits on interpretation. Some design limitations cannot be resolved through analysis alone; we make those limitations explicit.
What should I send for an initial assessment?
Start with a short description of the research question, study design, available data and the output you need. A variable list or data dictionary, the protocol, existing methods or results, and any reviewer comments are useful. We can then agree which files are needed and how they should be shared.
Can you review an analysis without rerunning it?
Yes. A methods or reporting review can identify conceptual problems and inconsistencies from the study description and outputs. Reproducing numerical results or checking implementation requires the relevant data and code. We state clearly what the review covers.
Can you help with a non-significant or unexpected result?
Yes. We can check the analysis, examine uncertainty and investigate whether the result is sensitive to justified analytical choices. The aim is a defensible interpretation of the evidence. A particular result or statistical significance cannot be guaranteed.
Does the service include manuscript editing?
We can help describe the statistical methods, present the results and address reviewer comments within the agreed consulting scope. Full English language editing can be arranged as a separate service and is performed entirely by human editors.
How are fees and confidentiality handled?
Fees depend on the questions to be addressed, the condition and complexity of the data, the analyses required and the agreed deliverables. We confirm the scope, price and confidentiality arrangements before work begins, including an appropriate way to share any necessary data.
Tell us about your study or analysis
Send us a brief description of your research question, study design and available data, together with the support you need and your deadline. We will assess the scope and provide a tailored quote.
Not ready to request a quote? Contact us to discuss your study and the support you may need.
Trusted by Publishers, Societies, and Researchers Worldwide
Selected Publishers & Societies
Selected Client Affiliations
Make a Payment
Buy Credit