Berkson

Part of speech: noun

Definitions

1. A statistical selection bias in which a sample drawn from a restricted group shows a spurious association between two traits that are unrelated in the wider population
2. The misleading correlation that can appear when researchers study only hospital patients or other selected cases, named after the statistician Joseph Berkson
3. An artifact of non-random sampling that makes two independent characteristics seem linked, often discussed in epidemiology and in the analysis of selected data

Etymology: Joseph Berkson was an American physician and statistician at the Mayo Clinic, and the term bearing his name is not a single word but a label for a statistical trap. Berkson noticed that when researchers study only people who have been admitted to hospitals, two conditions can look linked even when they are unrelated in the general population. Because hospital patients are selected for having at least one illness, the sample distorts the picture. Statisticians now call this Berkson's paradox, or Berkson's bias. The name is attached to the effect rather than to any word coined for it. Epidemiologists and statisticians use it to describe selection effects in case-control and hospital-based studies, where a correlation seen in the sample may be an artifact of how people entered the study. The lesson is about sampling: the pattern belongs to the group chosen for study, not necessarily to the wider population.