Opening the book…
Once the quality checks support analysis, the team must decide how much independent evidence it has. The experimental unit is the smallest unit independently assigned to a treatment. The unit of analysis is the level represented by each value used in a comparison. These levels must fit the design. When measurements share a well, donor, or culture run, the analysis must account for that relationship.
A technical replicate repeats a measurement or a closely related preparation step to assess variation in the method. Biological replicates are independent biological units relevant to the question, such as donors or separate culture preparations. State what was repeated and how treatments were assigned. Wells from one preparation can be experimental units for some questions, but they do not represent independent donors or source preparations.
Pseudoreplication occurs when related measurements are treated as if they were independent biological evidence. Four image fields from one well share that well’s treatment and culture history. Treating the fields as four independent cultures can underestimate uncertainty. Counting thousands of cells within one treated well has the same grouping issue.
Worked example: counting the evidence
A scratch study has three independent culture runs. Within each run, two treated wells are imaged at four positions each. The dataset contains 24 treated image fields, six treated wells, and three independent culture runs. None of these counts is a substitute for the others.
For a run-level comparison, first summarize fields within each well and then use a stated method to summarize the wells within the run. Compare the three independent run-level results with the corresponding control results. A more advanced analysis can model the nested structure directly. In either case, state the level represented by n, the sample size.
Matched controls from the same donor or run can produce paired data. Each treatment value is linked to a particular control by the design. Arbitrary pairing after viewing the values does not create a paired experiment. Independent groups use a different analysis because there is no designed one-to-one relationship.
Attribution: Images AI-generated using ChatGPT Modified by Emalee Mackenzie, Irvine Valley College. Licensed under CC BY-NC-SA
This page titled 15.7: Replication and the Unit of Analysis was last modified on Wed, 21 Oct 2026 01:35:51 GMT and is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by Emalee MacKenzie.
The book could not be loaded just now.