Interface 2003
Abstract

A-Family Priors: Smoothing Multinomial Data
Jeff Benedict, (NSA)

Abstract

The A-family is a family of distributions for probability vectors. In this talk, the A-family serves as a prior distribution for probability vectors that generate multinomially distributed count vectors. Posterior inference based on importance sampling is described. The A-family is used to model prior smoothness in frequency histograms and this idea is applied to stamp-thickness counts.


Take me back to the main conference page.