- Published
- Updated
Measurements taken before and after an intervention in the same person are not two independent samples. Losing the pairing can obscure uncertainty and which people contribute to the comparison. Preserve the relationship explicitly in the data structure used for analysis.
Define the question and assumptions
Define the pairing unit: it may be a person, an organ or another prespecified connection. Check timing and method comparability between observations. Plan incomplete pairs according to the scientific question. Subtracting two separate means does not always replace analysis of within-person change, because different participants may contribute to those two summaries.
Hypothetical study example
In a hypothetical study, baseline means use everyone but follow-up means use only attendees. Their difference is reported as average individual change. The team examines who contributed to each summary and explicitly defines an appropriate pairing and missing-data approach rather than assuming the populations are interchangeable.
Design review checks
- Verify the pairing key and rules identifying a unique eligible measurement.
- Check whether baseline and follow-up summaries contain the same contributing people.
- Test derivations for missing pairs and multiple candidate paired observations.
Record the decision and its limits
Accepted analysis reproduces the source and selection of each pair. Use a method respecting dependence, and distinguish people from measurement counts. If excluding incomplete pairs changes the population, report that consequence. Do not interchange changes in group means with within-person change measures; show the relevant uncertainty and explain which quantity answers the intended question.
Frequently asked questions
- Is the difference of two means the mean change?
- Assess whether the same people and compatible calculation conditions contribute.
- Do more measurements mean more participants?
- No. Repeated observations remain linked to the same person.