What counts as an outlier depends on substantive theory, not statistics.

Evidence

A core issue with this approach is that there might be “outliers,” which can significantly extend the scale in one or both directions. The scare quotes on “outlier” suggest a problem: one person’s outlier might be another person’s “good case” or “ideal type.” — Gary Goertz, Social Science Concepts and Measurement, 873

Categorization

• APHORISMS: The scare quotes do the work: one person’s outlier is another’s ideal type — a compact rule for treating data-cleaning decisions as theory-laden.
• DEEPTHINKING: Reframes the whole data-cleaning enterprise: every rule for excluding extremes silently encodes a claim about which cases matter most.