|
Activity Number:
|
139
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #302832 |
|
Title:
|
Power Filter for Dynamic Models
|
|
Author(s):
|
Sourish Das*+ and Dipak K. Dey
|
|
Companies:
|
Duke University and University of Connecticut
|
|
Address:
|
Old Chemistry Building, Statistical Science, Durham, NC, 27708-0251,
|
|
Keywords:
|
Dynamic Generalized Linear Models ; Kalman Filter ; Information Processing ; Sequential Monte Carlo
|
|
Abstract:
|
In this paper, we present the method for modeling longitudinal studies as state space dynamic generalized linear models. We accomplish this by introducing the power filter for dynamic generalized linear models, which extends the usual Kalman filter for dynamic linear models. We establish a relationship between the Kalman filter and the power filter as well. An information processing optimality property of the power filter is presented. We present the analysis of a recent longitudinal study, on a new treatment of osteoarthritis of knee, using the power filter.
|
