|
Activity Number:
|
505
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #306002 |
|
Title:
|
Importance Sampling with the Huber and Elliptical Multivariate Logistic Distributions
|
|
Author(s):
|
Stella W. Karuri*+ and Buffy Hudson-Curtis and John Monahan
|
|
Companies:
|
North Carolina State University and North Carolina State University and North Carolina State University
|
|
Address:
|
2501 Founders Drive , Raleigh, NC, 27695-8203,
|
|
Keywords:
|
integration ; importance sampling ; Huber distribution ; multivariate logistic distribution ; radial-spherical distributions
|
|
Abstract:
|
Importance sampling in Monte Carlo integration is one method employed in numerical integration of the posterior. We propose two radial-spherical distributions as importance functions, the Huber and the Elliptical Multivariate Logistic (EMVL) distributions. The motivation of the study was to find good importance functions, functions whose tail behavior matches that of the posterior. The appeal of the Huber and the EMVL as importance functions is that it is possible to generate random variables from them and their densities can be evaluated up to a constant. In addition they have tail behavior that decays exponentially. They can also be constructed in such a way that their tail behavior matches to the 'worst' direction of the posterior in spherical co-ordinates. We outline the construction of these distributions and present several importance sampling applications.
|
