Robert L. Wolpert

Robert L. Wolpert

Professor of Statistical Science

External address: 
214 Old Chemistry, Durham, NC 27708-0251
Internal office address: 
Box 90251, Durham, NC 27708-0251
Phone: 
(919) 684-3275
Email: 

Overview

I'm a stochastic modeler-- I build computer-resident mathematical models for complex systems, and invent and program numerical algorithms for making inference from the models. Usually this involves predicting things that haven't been measured (yet). Always it involves managing uncertainty and making good decisions when some of the information we'd need to be fully comfortable in our decision-making is unknown.

Originally trained as a mathematician specializing in probability theory and stochastic processes, I was drawn to statistics by the interplay between theoretical and applied research- with new applications suggesting what statistical areas need theoretical development, and advances in theory and methodology suggesting what applications were becoming practical and so interesting. Through all of my statistical interests (theoretical, applied, and methodological) runs the unifying theme of the Likelihood Principle, a constant aid in the search for sensible methods of inference in complex statistical problems where commonly-used methods seem unsuitable.

Three specific examples of such areas are: 

  • Computer modeling, the construction and analysis of fast small Bayesian statistical emulators for big slow simulation models
  • Meta-analysis, of how we can synthesize evidence of different sorts about a statistical problem
  • Nonparametric Bayesian analysis, for applications in which common parametric families of distributions seem unsuitable.

Many of the methods in common use in each of these areas are hard or impossible to justify, and can lead to very odd inferences that seem to misrepresent the statistical evidence. Many of the newer approaches abandon the ``iid'' paradigm in order to reflect patterns of regional variation, and abandon familiar (e.g. Gaussian) distributions in order to reflect the heavier tails observed in realistic data, and nearly all of them depend on recent advances in the power of computer hardware and algorithms, leading to three other areas of interest:

  • Spatial Statistics,
  • Statistical Extremes 
  • Statistical computation

I have a special interest in developing statistical methods for application to problems in Environmental Science, where traditional methods often fail. Recent examples include developing new and better ways to estimate the mortality to birds and bats from encounters with wind turbines; the development of nonexchangeable hierarchical Bayesian models for synthesizing evidence about the health effects of environmental pollutants;  and the use of high-dimensional Bayesian models to reflect uncertainty in mechanistic environmental simulation models.

My current (2015-2016) research involves modelling and Bayesian inference of dependent time series and (continuous-time) stochastic processes with jumps (examples include work loads on networks of digital devices; peak heights in mass spectrometry experiments; or multiple pollutant levels at spatially and temporally distributed sites), problems arising in astrophysics (Gamma ray bursts) and high-energy physics (heavy ion collisions), and the statistical modelling of risk from, e.g., volcanic eruption. 
 

Education & Training

  • Ph.D., Princeton University 1976

  • B.A., Cornell University 1972

Selected Grants

Collaborative Research: Using Precursor Information to Update Probabilistic Hazard Maps awarded by National Science Foundation (Principal Investigator). 2018 to 2020

EMSW21-RTG: Geometric, Topological awarded by National Science Foundation (Key Faculty). 2011 to 2018

Collaborative Research: Statistical And Computational Models and Methods for Extracting Knowledge from Massive Disparate awarded by National Science Foundation (Principal Investigator). 2012 to 2015

FRG: Collaborative Research: Prediction and Risk of Extreme Events Utilizing Mathematical Computer Models of Geophysical awarded by National Science Foundation (Principal Investigator). 2008 to 2012

SCREMS: Distributed Environments for Stochastic Computation awarded by National Science Foundation (Co-Principal Investigator). 2004 to 2007

Spatial-temporal Models for Environmental Health Effects awarded by Environmental Protection Agency (Co-Principal Investigator). 2001 to 2006

Pages

Berger, James, et al. “Statistische und Probabilistische Methoden der Modellwahl.” Oberwolfach Reports, European Mathematical Publishing House, 2005, pp. 2611–704. Crossref, doi:10.4171/owr/2005/47. Full Text Open Access Copy