Mark J. Schervish

I find statistics to be a very broad field that gives me the freedom to work on many different kinds of problems. I have worked on problems ranging from foundational issues to methodology, theory, and applications.

Together with John Lehoczky (Department of Statistics) and Andrzej Strojwas (Department of Electrical and Computer Engineering), I have been working on monitoring systems for multistage manufacturing processes. The development of a good monitoring system requires new statistical theory and methods along with computational and graphical techniques plus an understanding of the physical process being monitored. I find this sort of work rewarding both for the opportunity to work with people in other fields and for the ability to see the value and limitations of the techniques in my own field.

I have done theoretical work in the comparison and combination of expert opinions, the consistency of nonparametric Bayesian procedures, finitely additive probability, Markov Chain Monte Carlo, and other areas. I have also recently compiled a textbook on theoretical statistics for advanced graduate students. My goal, for the book, was to cover those topics in classical and Bayesian inference that would prepare students to be able to read and contribute to the research journals in the field.

I also find the foundations of statistics and decision theory to be a very fruitful area of inquiry. Great insights into the value (or lack thereof) of various statistical techniques and theoretical concepts can be achieved by carefully studying the axiomatic foundations of the field. In particular, Teddy Seidenfeld, Jay Kadane and I have been carefully examining several of the axioms that underly statistical decision theory.

Some Related Publications

Schervish, M. J. (1995) Theory of Statistics. Springer-Verlag, New York.

Seidenfeld, T., Schervish, M. J. and Kadane, J. (1995) "A representation of partially ordered preferences," Annals of Statistics, 23, pp. 2168-2217.

Schervish, M. J. (1996). "P-values: What they are and what they are not," American Statistician, 50, 203-206.