|
Activity Number:
|
141
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #302879 |
|
Title:
|
Adaptive Mixture Modeling Metropolis Methods for Bayesian Analysis of Nonlinear State-Space Models
|
|
Author(s):
|
Jarad Niemi*+ and Mike West
|
|
Companies:
|
Duke University and Duke University
|
|
Address:
|
, , ,
|
|
Keywords:
|
Bayesian computation ; forward filtering, backward sampling ; non-linear state space model ; regenerating mixture procedure ; Smoothing in state-space models ; systems biology
|
|
Abstract:
|
We describe a strategy for Markov chain Monte Carlo analysis of non-linear state space models involving batch analysis for inference on dynamic, latent state variables and fixed model parameters. The key innovation is a Metropolis-Hastings method for the time series of state variables based on sequential approximation of filtering and smoothing densities using normal mixtures. These mixtures are propagated through the nonlinearities using a local mixture approximation method and we use a regenerating procedure to deal with potential degeneracy of mixture components. This provides accurate approximations to sequential filtering and retrospective smoothing distributions, and hence a useful construction of global Metropolis proposal distributions for simulation of posteriors for the set of states. This analysis is embedded within a Gibbs sampler to include uncertain fixed parameters.
|
