|
Activity Number:
|
361
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
WNAR
|
| Abstract - #308064 |
|
Title:
|
Temporal Configuration Analysis: Hidden Markov Models for Heterogeneous Multivariate Longitudinal Data
|
|
Author(s):
|
Steven L. Scott*+
|
|
Companies:
|
University of Southern California
|
|
Address:
|
Bridge Hall 401H, Los Angeles, CA, 90089,
|
|
Keywords:
|
hidden Markov model ; longitudinal data ; model based clustering ; panel data
|
|
Abstract:
|
Multivariate longitudinal data can be described using model based clustering. Clusters are defined by a collection of cross sectional models. A subject's path through the data is described by a sequence of cluster membership indicators modeled using a hidden Markov chain. Analysis of such a model proceeds in two steps. First, interpretable labels are mixture components. Second, one uses the posterior distribution of the hidden Markov chain to learn about how individuals under different conditions move through the discrete set of states. This setup allows treatments and interventions to be modeled in terms of the probability of moving subjects to more desirable states. The method can measure temporal inhomogeneity using Bayesian shrinkage methods, and it can describe non-Markov paths by introducing mixtures of latent Markov chains.
|
- 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 |
