The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.
Online Program Home
Abstract Details
Activity Number:
|
250
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 30, 2012 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Bayesian Statistical Science
|
Abstract - #306016 |
Title:
|
Efficient Multiple MCMC for Bayesian Parameter Estimation in Dynamic Systems
|
Author(s):
|
Samuel Wong*+
|
Companies:
|
Harvard University
|
Address:
|
28 Wendell St. Apt 4, Cambridge, MA, 02138, United States
|
Keywords:
|
Bayesian computation ;
MCMC methods ;
dynamic systems ;
differential equation models ;
parameter estimation
|
Abstract:
|
We propose an efficient MCMC scheme for estimating parameters in dynamic systems governed by a set of ordinary differential equations, which are frequently used to describe behaviors in science. The data observed are usually noisy and collected at discrete time intervals as the system evolves. Bayesian and likelihood-based inference for these systems face two main computational challenges: the unusual shapes of likelihood surfaces encountered, and the time required for numerically solving the differential equations. To address these challenges, our framework introduces a latent variable to control the noise level in the model, producing multiple chains of Monte Carlo samples of parameters to allow the coarser chains to improve convergence of the finer chains. We then improve computational efficiency by only selectively evaluating the numerical solutions, and improve estimation efficiency by combining the multiple chains. Our method is illustrated by applications to biological systems.
|
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 2012 program
|
2012 JSM Online Program Home
For information, contact jsm@amstat.org or phone (888) 231-3473.
If you have questions about the Continuing Education program, please contact the Education Department.