|
Activity Number:
|
372
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
IMS
|
| Abstract - #309336 |
|
Title:
|
Collocation Tempering for Differential Equation Models
|
|
Author(s):
|
David A. Campbell*+
|
|
Companies:
|
Simon Fraser University Surrey
|
|
Address:
|
250-13450 102nd Ave, Surrey, BC, V3T 0A3, Canada
|
|
Keywords:
|
MCMC ; Differential Equation Models ; Smoothing ; tempering
|
|
Abstract:
|
Differential equation models often have no analytic solution so statistical models centered on the solution to a differential equation require a numerical solution 'S.' Computational Bayesian methods for differential equation models based on MCMC require S to obtain the likelihood. However, a single differential equation can be used to model unbounded growth, limit cycles or a tendency towards an asymptote simply by changing the parameters. Consequently results of MCMC based on the numeric solution are highly dependent on parameter values used to initialize the algorithm because in crossing the parameter space a mediocre fit to the data must deteriorate substantially before it can improve. I present a collocation based tempering method which overcomes these problems while converging quickly to a stable distribution.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2007 program |
