You discover that samples were measured repeatedly per donor. How would you revise the analysis?

Why

Correlation reduces the effective information gained from additional observations. Ignoring clustering usually understates uncertainty. The correct approach depends on whether repeated measures are technical repeats, repeated time points or distinct treatments within a donor.

How to work it through

1. Reconstruct the experimental design and who was independently sampled or randomized.
2. Plot donor trajectories and within-donor variability.
3. For paired conditions use paired contrasts where appropriate.
4. For more complex time courses consider a mixed-effects or generalized estimating-equation model with donor grouping and defensible covariance assumptions.
5. If only donor summaries are needed, aggregate within donors and analyse donors.
6. Run sensitivity analyses for influential donors and missing time points.

Where the answer stops holding

With very few donors, random-effect variance estimates can be unstable. A complex model does not invent independent biological replication. If individual wells were randomized independently within donors, donor blocking still matters.