I first met David Cox in 1972. My former advisor and senior colleague Rupert Miller had arranged a year’s sabbatical at Imperial College, where David was the professor and chair of statistics. I did the same, very much wanting to meet the inventor of proportional hazards. The department was located in a beautiful old Victorian building at 53 Princess Gate, almost adjoining the Victoria and Albert Museum. David presided in his gentle but authoritative manner over a dozen or so young researchers, including Michael Godfrey, David Hinkley, Anthony Atkinson, and Agnes Herzberg.

The two Davids, Cox and Hinkley, were writing their influential book on mathematical statistics. I was amazed at how quickly the work went on; whole chapters seemed to emerge in a day or two. It was said that Mozart could write all the parts of a symphony at once, the 17 staves proceeding together across the page. David Cox’s synoptic view of statistics was Mozartian in the same sense. Statistics is a messy field, with its working parts scattered across the scientific world, but David seemed to be able to take it in all of a piece.

Day-to-day life in Princess Gate had a pleasant rhythm. Twice a day, at 10:30 and 3:30, we would all meet for tea and biscuits. David’s formidable secretary Alexandria enforced the rule: one plain biscuit in the morning, one fancy biscuit in the afternoon. David helped guide the discussion, following another rule I learned the hard way of no technical chatter.

In the spring, my colleague Rupert Miller gave a talk on the jackknife, then an intriguing but somewhat mysterious new topic. Rupert’s talk concerned the jackknife’s dependability, a worry since it was seen to fail for the variance of the median. After the talk, David came up to me and said, “Do you think there’s anything to the jackknife?” That was his hint to a young researcher and, given the bootstrap, probably the best one of my life.

The ensuing 50 years didn’t provide a great deal of face-to-face encounters, but David grew in my mind as the commanding voice of statistical sanity. If you look in Section 8 of my JASA article, “Prediction, Estimation, and Attribution” (Efron, 2020), you will see me trying to borrow some of David’s authority on the question of the limitations of machine learning.

In 2016 we met in person for the last time, in Madrid at the Frontiers of Knowledge Award Ceremony hosted by the BBVA Foundation and the Spanish National Research Council, where we were the two statisticians being honored among pairs of scientists from other fields. Each pair was to give a talk on their accomplishments. David gave our talk, which, given the usual mysteries of our field, amounted to saying just what it was that statistics and statisticians do for the world of science. Nobody could have done it better.

Disclosure Statement

Bradley Efron has no financial or non-financial disclosures to share for this article.

Reference

Efron, B. (2020). Prediction, estimation, and attribution. With discussion and a rejoinder by the author. Journal of the American Statistical Association, 115(530), 636–677. https://doi.org/10.1080/01621459.2020.1762613

©2023 Bradley Efron. This article is licensed under a Creative Commons Attribution (CC BY 4.0) International license, except where otherwise indicated with respect to particular material included in the article.