BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:David Cox's 1972 proportional hazards paper - Phil Dawid\, Univers ity of Cambridge DTSTART:20081029T163000Z DTEND:20081029T173000Z UID:TALK14653@talks.cam.ac.uk CONTACT:Richard Samworth DESCRIPTION:In March 1972\, David Cox presented his paper "Regression Mode ls and Life Tables" to a meeting of the Royal Statistical Society. The pa per and the ensuing lively discussion were published in the Journal of the Royal Statistical Society\, Series B (Methodological)\, Vol. 34\, No. 2 ( 1972)\, pp. 187-220 . According to t he Web of Science\, this paper has had over 23300 citations\, which is alm ost certainly a gross underestimate. Most of those citations are from the medical literature\, since what Cox did was to provide a simple yet extre mely flexible solution\, based on his formulation of a "proportional hazar ds model"\, to the problem of comparing survival across different individu als while taking full account of differing treatments\, baseline and time- varying personal characteristics\, withdrawal from follow-up\, etc. In 19 90\, Cox won the prestigious Kettering Prize and Gold Medal for Cancer Res earch for "the development of the Proportional Hazard Regression Model." The importance of the paper as a contribution to Science is thus clear. B ut its contributions to statistical methodology were of equal originality and importance\, and it is those I propose to discuss. \n\n\n\nCox's model was perhaps one of the earliest examples of a "semiparametric model"\, an d his method of eliminating the nonparametric part through the formation o f a "partial likelihood" was certainly imaginative. But his argument for the validity of the method was heuristic in the extreme\, and it was not u ntil much later that it was set on a firm mathematical basis through the a pplication of martingale theory. It could be argued that its philosophica l basis\, which has much in common with the prequential approach to statis tical inference\, is less secure. I hope others will join me in a thoroug h-going discussion of the contents and ramifications of this pathbreaking paper. LOCATION:MR11\, CMS END:VEVENT END:VCALENDAR