Simpson's Paradox

Definition

A problem in statistics where trends appear in different groups of data but disappear (or even reverse) when these groups are combined.

Origin

First described by Edward Simpson in 1951, the term was popularized from a 1976 account in Scientific American by Martin Gardner, where in a famous study from University of California, Berkeley admissions, it appeared that men were more likely to be admitted than women, but when examining individual departments, there was in fact a slight bias in favor of women.

Everyday Use

More than just a boring artifact of statistics, it's an excellent reminder to always zoom out — look beyond the data we see right in front of us and wonder how it might look in a broader context. So many of the decisions we make or the judgements we pass are in such a narrow band, that we might not only be missing the bigger picture, but telling the exact opposite story, due to our narrow vision.