Stat 583

Comments to Homework 1

Question 2

For an extension of this model (from a non-Bayesian point of view), see Guttorp and Thompson (1990), JRSS B 52: 157-173.

Question 3

This inequality is sometimes called the Principle of Precise Measurement, and says basically that if the prior is not too peaked, and if the experiment is informative (so the likelihood is peaked), then the posterior is well approximated by the likelihood (normalized to be a density).