An observed value that is very different from the rest of the data. This can occur through data collection errors, or because the experimental unit is different from the others through disease or other environmental insult.

These values should be identified during statistical analysis, confirmed with scientific rationale, and deleted. If undiscovered, they can increase variance which lowers power, and can change treatment means. Equivalently, they should be deleted because they are not representative of the experimental population of interest.