Opening the book…
The team’s graph should make the evidence easier to inspect. Label the endpoint,
units, comparison groups, and level represented by each point. For a small dataset, individual values with a suitable summary usually show more than a bar displaying only the mean. The reader should be able to see the spread and the number of independent observations. Lines connect each donor’s control and treatment values. The donor-level changes are 9, 12, 10, and 8 percentage points. The pairing is part of the experimental design, not a line added between unrelated values. Original plot.
Use a time-course plot for planned measurements over time, a concentration-response plot for a concentration series, and a paired plot when each value has a designed partner. State whether points represent fields, wells, cultures, runs, or donors. Keep raw and normalized values distinguishable. A normalized control fixed at 100% does not imply that its original measurements had no variation.
An outlier is an observation unusually distant from other values. It may reflect error, a genuine biological difference, or ordinary variation in a small dataset. An unusual value is a reason to investigate, not an automatic reason to delete it.
Investigate an unusual result by checking identity, raw files, instrument warnings, calculations, culture history, and deviations. A documented failure may justify exclusion; use a predefined rule when one applies. If an unforeseen failure requires a new decision, explain that decision openly. Keep the original value and report the reason for exclusion. If its cause is uncertain, show how including or excluding it affects the conclusion rather than quietly removing it.
Data integrity means that the record accurately preserves what was done and observed. Keep original images, instrument exports, plate maps, formulas, and exclusions. Corrections should remain traceable. Do not overwrite an original value with a preferred replacement or present one selected image as though it represented every replicate.
Repeated measurement is appropriate when a documented problem needs resolution. Distinguish a repeated assay on the same sample from a new independent culture experiment. Report what was repeated and why. A second reading does not erase the first, and averaging valid and invalid readings does not create a valid result.
Attribution: Images AI-generated using ChatGPT Modified by Emalee Mackenzie, Irvine Valley College. Licensed under CC BY-NC-SA
This page titled 15.11: Graphs Outliers and Data Integrity was last modified on Wed, 21 Oct 2026 01:35:52 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.