D. R. Cox: aspects of scientific inference

Abstract

Abstract The year 2022 marked 100 years since R. A. Fisher’s landmark paper on the foundations of statistics, and 50 years since D. R. Cox’s seminal work on the proportional hazards model. At a 1-day meeting in recognition of the latter, held at the London School of Hygiene and Tropical Medicine, I attempted to place some of D. R. Cox’s most influential work in its appropriate historical context, reconstructing some of his ideas from unifying principles and providing a different exposition to that of the original papers. The present article is an elaboration of that talk. Additional material includes a geometric explanation of conditional inference in models with binary outcomes, relating the logistic analysis of matched comparisons presented by Cox (1958c, Two further applications of a model for binary regression. Biometrika, 45, 562–565) to the protracted and sometimes polemical discourse over the analysis of 2×2 contingency tables.