Established in Memory of Seymour Geisser, Director of the School of Statistics (1971-2001)

Thursday, September 22, 2005
3:00 PM
Moos Tower 2-530

Marvin Zelen
Department of Biostatistics
Harvard University

Reminiscences

James O. Berger
Department of Statistics
Duke University

Something Old and Something New: Bivariate Normal and Computer Models

Abstract

One of the earliest problems considered by Seymour Geisser was objective Bayesian/fiducial inference for a bivariate (multivariate) normal distribution. Forty plus years later, there are still interesting ideas of Seymour's to follow up, both practically and philosophically. Seymour was a forceful advocate for predictive analysis. This is being (belatedly) recognized as the central concept in computer model validation. One of the major activities in science and engineering today is the development of math-based computer models of scientific and engineering processes and, after many false starts, it is being realized that the most basic question in the evaluation of such computer models is "Does the computer model adequately predict reality?" A way to approach this question, with examples, will be discussed.