Seymour Geisser Distinguished Lecture by Jim Hodges

Speaker

Jim Hodges, University of Minnesota

Talk Title

Provocative Observations About the Foundations of Statistics (If Anybody Still Cares)

Abstract

I became aware of the School of Statistics in the late 1970s through my interest in the foundational debate, in which Seymour participated actively. When I was a student in the School, Seymour was my academic advisor and then my thesis advisor. After having some fun with memories about Seymour, I'll defend a statement that Seymour would have enjoyed sparring over: "Bayesian and frequentist methods don't differ as much as their proponents would have you believe." I will use two examples from clinical trial monitoring, in which Bayesians and frequentists supposedly differ most starkly, but in fact (I will argue) don't differ much. This dispute produced much acrimony and ferocious invective, as I will illustrate with examples that Seymour recounted gleefully in his Foundations class. These examples may surprise audience members too young to remember life before the Gibbs sampler but they will, I hope, be nostalgic for the, um, older crowd.

Bio

Jim got his undergrad degree at the University of Waterloo (Canada) in 1979 and entered the School of Statistics in 1981.  He left with a PhD in statistics (1985) and an MA in Public Affairs (1986), then worked as a statistician at the RAND Corporation from 1985 to 1993.  He returned to the University's Division of Biostatistics, in the School of Public Health, in 1993 and has been there since then, first as a Senior Research Associate and now as contract faculty.