Interview with Jim Berger
CHANCE magazine invited Jim Berger, Arts and Sciences Professor of Statistics at Duke University, to talk with editors Sam Behseta and Mine Çetinkaya-Rundel. Berger’s major contributions to statistical sciences, especially to Bayesian statistics, need no introduction. Here, Behseta and Çetinkaya-Rundel talk with Berger about his views on Bayesian statistics, his recent work on reproducibility, and his thoughts about the future of statistics.
Sam Behseta: Jim, you talk about the objective Bayesian approach. Could you tell us more about why you prefer this approach?
Jim Berger: First of all, let me just say that subjective Bayesian analysis is also extremely important. For instance, right now, I’m working on a project with [Eli] Lilly on subgroup analysis, with colleagues Lei Shen, James Scott, and Xiaojing Wang. One of Lilly’s primary interests is in designing experiments where they can state, ahead of time, their subjective odds that a treatment would be effective for certain factors or certain subgroups. They have a lot of scientific knowledge that they want to bring to bear on the problem and, in subgroup analysis, this can be done in a fairly straightforward manner with subjective Bayesian methods. And as long as this is done pre-experimentally, it satisfies all the criteria for objective statistical error control.Some content is only viewable by Chance Subscribers