Log t for k the dimension of 2 choosing the model m

This preview shows page 6 - 10 out of 11 pages.

  log T for k the dimension of 2 Choosing the model m for which log p y | 2 ± mT   " k m /2   log T is biggest is known as the using the Schwarz Information Criterion (SIC) or Bayesian Information Criterion (BIC)
Image of page 6

Subscribe to view the full document.

7 Since it is asymptotically a Bayesian decision rule, SIC inherits the properties of being asymptotically admissible and consistent However, note that this result required the same regularity conditions needed to get asymptotic Normality of MLE VIII. Model selection A. Marginal likelihood B. Schwarz criterion C. Calculating the marginal likelihood with the Gibbs sampler
Image of page 7
8 Goal: calculate p y   ± ; p y | 2   p 2   d 2 Couldn’t we get this by drawing 2 j   j ± 1,..., J from p 2   and then p ³ y   ± J " 1 ! j ± 1 J p y | 2 j     ? Answer: no, this algorithm is badly behaved numerically. Chib’s idea: think of evaluating at a point with a lot of mass (say the posterior mean 2 ' ). Note that for any 2 ' we have the identity p 2 ' | y   p y   ± p y | 2 '   p 2 '   p y   ± p y | 2 '   p 2 '   / p 2 ' | y   In many applications, we know p y | 2 '   and p 2 '   analytically (evaluating the likelihood and prior at posterior mean, respectively), but couldn’t calculate p 2 ' | y   explicitly
Image of page 8

Subscribe to view the full document.

9 Suppose we’ve generated draws from a two-block Gibbs sampler: p 2 1 | 2 2 , y   and p 2 2 | 2 1 , y   The object of interest is given by p 2 1 ' , 2 2 ' | y   ± p 2 1 ' | 2 2 ' , y   p 2 2 ' | y   where we may know p 2 1 ' | 2 2 ' , y   analytically.
Image of page 9
Image of page 10
You've reached the end of this preview.
  • Winter '09
  • JamesHamilton
  • Gibbs, Akaike information criterion, Bayesian information criterion, y|, Deviance information criterion

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern