E2. To be completed after lesson 6¶
When performing parameter estimation by optimization, after we find the MAP, why do we locally approximate the posterior near the MAP as a (possibly multivariate) Normal?
What do we mean when we say a distribution has a story? How are stories used for setting likelihoods?
What is meant by the term “weakly informative prior?”
What is a credible interval? Why may the approximate credible intervals computed using the MAP be off of the true credible interval?
Write down any questions or points of confusion that you have.