Audit the data before choosing a test
Provide a data dictionary with variable names, coding, units, missing-value conventions and participant or observation identifiers that do not expose personal information. Resolve impossible values, duplicate records and inconsistent categories with you. Keep an unchanged original and a documented analysis copy.
Match methods to the design
Descriptive summaries, group comparisons, associations, regression and longitudinal models address different questions. The choice depends on design, measurement, distribution, dependency between observations and assumptions. An analytical plan should make primary and exploratory work distinguishable.
Make results interpretable
Agree tables and graphics that show the study population, denominators and estimates clearly. Where appropriate, explain uncertainty, effect size and model diagnostics alongside p-values. Statistical significance alone does not establish practical importance or a causal relationship.
- A documented cleaning or query log.
- Analysis outputs with variable definitions and method notes.
- Tables and figures consistent with the supplied data.
- An explanation of results and limitations for author review.
Software and reproducibility
State whether your institution requires SPSS, R, Python, Stata or another format. We confirm software-specific work and whether syntax, scripts or only reviewed outputs are included. New datasets, additional models and changed hypotheses should be agreed as new analytical tasks.
Scope, files and delivery
Start with a short project brief. We assess the work, discuss the expertise required and confirm the deliverables, fee, schedule and included revisions in writing. Share the current version and the instructions that govern it. Additional data, a new target journal or a substantially changed brief may require a revised scope. You review the proposed changes and retain responsibility for the research.