The labels “biological replicate” and “technical replicate” can hide important details. Start by describing what you treated, what you measured, and what population you hope to understand.

Follow the treatment assignment

The experimental unit is the entity independently assigned to a treatment condition. The observational unit is what you measure. Several measurements can belong to one experimental unit: cells within an animal, for example. Counting those measurements as independent units can create pseudoreplication. Lazic and colleagues explain why the distinction matters in their study of replication in biological experiments.

Repeated instrument readings can improve measurement precision, but they do not create new independent treatment assignments. Neither “one well” nor “one animal” is automatically the right answer for every experiment; assignment and shared structure matter.

A synthetic example: three readings per preparation

Imagine four independent preparations assigned to each treatment. Each preparation is measured three times on the same instrument. There are twelve readings per treatment, but four independently assigned preparations.

For this illustrative design, a prespecified mean of the three technical readings could provide one outcome per preparation. Keep the raw readings and document that summarisation. Do not paste all twelve readings into an independent-groups analysis as though twelve preparations had been assigned.

This example is synthetic. Averaging is not a universal solution: it must suit the scientific question, and cannot remove dependence between treatment groups.

Write down n in words

Before analysis, make a short record:

  • The entity assigned to treatment and how assignment occurred.
  • The measurement taken from each entity, including repeated readings.
  • Any shared donor, litter, cage, plate or experimental run.
  • The population your conclusion is intended to describe.

If a donor contributes measurements to every condition, or treatment is assigned to a whole cage, an ordinary independent-groups analysis of every measured value does not represent that structure. Choose an analysis that accounts for it. The Moleculytics workspace does not fit paired, repeated-measures or nested models.

Report the structure, not just the count

For animal experiments, ARRIVE 2.0 sample-size guidance asks authors to report the exact number of experimental units per group and the total animals used, with exclusions accounted for. In your own analysis notes, also record measurements per unit and any summarisation so readers can follow the path from raw readings to the analysed n.

References