In Memory of Herman Rubin and His Contributions
Organizer and Chair: Anirban DasGupta, Professor of Statistics, Department of Statistics, Purdue University
Speakers
- Rodrigo Bañuelos, Professor of Mathematics, Department of Mathematics, Purdue University
- Rick Vitale, Professor, Department of Statistics, University of Connecticut
- Andrew L. Rukhin, Statistical Engineering Division, ITL, National Institute of Standards and Technology, USA - cancelled
Thursday, June 7, 1:30-3:30 p.m. in STEW 202
| Time | Speaker | Title |
|---|---|---|
| 1:30-2:10 p.m. | Rodrigo Bañuelos | A tale of three inequalities, conversations with Herman Rubin |
| Abstract: In 1984, Alice Chang, Michael Wilson and Tom Wolff wrote a seminal paper in harmonic analysis which solved two important problems, one posed by Charles Fefferman and the other by Elias Stein. The key estimate for the solution of Stein’s problem was an inequality for martingales with a beautiful proof (included in their paper) by Herman Rubin. This inequality, often referred to in the literature as “Rubin’s Lemma or Rubin’s estimate" had a tremendous influence on subsequent developments related to many natural problems on subgaussian behavior of many operators in harmonic analysis, probability and PDE. In this talk, I will relate the personal (and curious) story of how the “Rubin Lemma” came about. The other two inequalities referred to in the title are on different topics. Time permitting, I will briefly describe some extremely short conversations with Herman on these two inequalities and how they illustrate his incredible memory and mathematical insight. | ||
| 2:10-2:50 p.m. | Andrew Rukhin |
Heterogeneous data and objective priors |
| Abstract: Challenging meta-analysis issues arise in collaborative studies when within-study uncertainties are unreliable or not reported. A Bayesian model with a non-informative prior for a common mean of several independent but not identically distributed observations is suggested. In the similar setting when the variance estimates cannot be trusted an informative Bayes models are discussed. | ||
| 2:50-3:30 p.m. | Rick Vitale | Two Papers with Herman |
| Abstract: The talk will recall my good fortune to engage with Herman on two topics: one concerning a characterization principle for probability measures, the other an asymptotic analysis of a class of statistics. | ||