Practical guides · Page 20

Run a Unit-of-Analysis Check Before Pooling Results

Find clusters, repeated observations, paired measurements, and shared control groups before treating results as independent.

A practical pilot checklist for turning a research question into explicit variables, time points, and collection rules.

Work through outcome definitions, follow-up, designs, and overlapping participants before deciding on quantitative synthesis.

A fictional example shows why outcome windows must be defined before analysis and why later changes need transparent documentation.

Read a hypothetical relative-risk row by checking its reference, time window, adjustment, uncertainty, and absolute context.

Use estimates, uncertainty, and design limitations to report an inconclusive finding without chasing a preferred p-value.

Build a defensible adjustment rationale by considering the research question, time order, and the roles of candidate variables.

Document missingness, analysis populations, and the assumptions behind your chosen approach before drafting a response.

Identify the claims and hypothesis family, explain the original analysis plan, and respond transparently without inventing prespecified decisions.

Explain the actual basis for your study size without inventing a prospective calculation after the results are known.

Why significant in one subgroup and not another does not establish a subgroup difference, with a worked hypothetical example.

A practical evidence-overlap check: compare existing reviews, search dates, protocols and available studies before investing in a new project.