prof. S.G.Walker Department of Mathematics and Department of Statistics & Data Science, University of Texas at Austin
Date:
Event location: Aula 12 (23rd) and aula 32 (24th) Piazza Scaravilli, Bologna
Type: Cycle 39 - Short courses and seminars
The series of talks takes a new look at the foundations of Bayesian uncertainty.
The idea is based on the notion that statistical uncertainty is present due to what is not observed, which if seen, would render the decision known or, in other words, would leave no uncertainty. On the other hand, the frequentist assumes uncertainty is caused by a finite sample size; the notion being that what is observed is one of many possible sets of data. This contrast leads to fundamental different plans, which can be seen clearly when studying the frequentist and Bayesian bootstraps, both of which start with the empirical distribution function. It is argued that martingales are the key to assessing Bayesian uncertainty and the suitable construction of a probability model for what has not been seen conditional on what has been seen.