Comparing Bayesian and frequentist prediction techniques
Snigdhansu Chatterjee, School of Statistics,University of Minnesota
Monday,Oct 11, 2004, 4:10 PM
319 Snedecor
This talk is a review of some of the standard techniques used for prediction. We
discuss the predictive likelihood, some of its problems and suggested modifications.
In objective Bayesian prediction, the choice of prior is an important issue.Some of the
standard choices are discussed. Recent developments in the hitherto largely ignored
bootstrap prediction techniques are also studied. It is shown that in regular cases the
bootstrap predictive density is close to the Bayesian predictive density with optimal
chice of prior.