outliers

Terms from Statistics for HCI: Making Sense of Quantitative Data

Unexpectedly extreme values in data, which may be caused by a fault in equipment, or other irrelevant effect, or may occur by chance. While they may be valid data points within the random nature of the phenomenon being studied, even a few outliers can have a disproportionate effect on statistics such as arithmetic mean and variance. For some purposes, better results are obtained by removing the most extreme outliers, but this can introduce the potential for cherry picking, so, where possible, a strategy for dealing with outliers should be determined prior to any analysis.

Defined on page 81

Used on pages 81, 82

Also known as removing outliers