|
Activity Number:
|
250
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306156 |
|
Title:
|
Inference for Partially Observed Systems with an Application to Cholera
|
|
Author(s):
|
Edward L. Ionides and Carles Breto*+ and Aaron A. King
|
|
Companies:
|
University of Michigan and University of Michigan and University of Michigan
|
|
Address:
|
439 West Hall, Ann Arbor, MI, 48109,
|
|
Keywords:
|
nonlinear state space model ; maximum likelihood estimator ; cholera
|
|
Abstract:
|
Inference for nonlinear state space models can be a challenging step in developing models for stochastic dynamical systems. The system investigated as a case study is a population model for cholera in Bangladesh. This example motivates the introduction of a new method for likelihood based inference, which we call MAPLE (Maximum A Posteriori Limit Estimation). The MAPLE algorithm computes a maximum likelihood estimator as the limit of an average of Bayesian posterior mean estimators. The new methodology has computational advantages and a theoretical justification.
|
