Berkson's paradox

Berkson's paradox is a phenomenon in statistics and epidemiology where two variables appear to be negatively correlated when observed within a specific population, even though they may be positively correlated in the general population. This happens because the selection criteria for the group being studied inadvertently filters out certain individuals, creating a misleading relationship. For example, if we look at patients with a specific disease who are only hospitalized, we might find that higher levels of a health marker correlate with fewer hospitalizations, despite that in the broader population, higher levels of that marker are associated with a higher likelihood of hospitalization.