berksons-paradox-is-when-selection-creates-a-correlation-but-isnt-actually-representative

An example: sampling multiple diagnoses in a hospital gives a sampling bias. (Bergstrom and West)