Studies in the history of probability and statistics, LI: the first conditional logistic regression.

This article focuses on the earliest known application and fitting of conditional logistic regression, a statistical model widely used today in epidemiology and economics, which was previously attributed to works from the 1970s. It reveals that Lionel Penrose and Ronald Fisher applied this model in 1934 to study the relationship between maternal age, birth order, and the occurrence of Down's syndrome in families, using family-based data to control for confounding factors. Fisher's methodological insight involved modeling relative odds within sibships to address outcome-based sampling bias, leading to an iterative maximum likelihood estimation approach that anticipated modern conditional logistic regression techniques. The article also highlights the collaborative peer-review process between Penrose and Fisher and discusses the historical and contemporary significance of this pioneering work.