A Primer for Bayesian Inference and MCMC in R

Speaker:

Dr. David Higdon, Social Decision and Analytics Laboratory, Virginia Bioinformatics Institute

Abstract:

This talk gives a conceptual, and hopefully intuitive overview of Bayesian inference. The Bayesian approach is quite flexible, allowing the modeler to explore non-standard model formulations. The resulting uncertainty is characterized by a (multivariate) probability distribution for the unknown model parameters. This talk will also describe the Metropolis algorithm, a surprisingly general recipe for generating a (dependent) Monte Carlo sample from this posterior distribution. The concepts will be demonstrated using simple R code which will be made available for all to try out. Time permitting, the talk will also show how this basic approach can be used for some "hand crafted" analyses.

About the speaker:
Dr. David Higdon is a professor in the Social Decision Analytics Laboratory at Virginia Tech University. Previously, he spent 10 years as a scientist or group leader of the Statistical Sciences Group at Los Alamos National Laboratory. He is an expert in Bayesian statistical modeling of environmental and physical systems, combining physical observations with computer simulation models for prediction and inference. His research interests include space-time modeling; inverse problems in hydrology and imaging; statistical modeling in ecology, environmental science, and biology; multiscale models; parallel processing in posterior exploration; statistical computing; and Monte Carlo and simulation based methods. Dr. Higdon has served on several advisory groups concerned with statistical modeling and uncertainty quantification and co-chaired the NRC Committee on Mathematical Foundations of Validation, Verification, and Uncertainty Quantification. He is a fellow of the American Statistical Association. Dr. Higdon holds a B.A. and M.A. in mathematics from the University of California, San Diego, and a Ph.D. in statistics from the University of Washington.