|
Activity Number:
|
58
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #308501 |
|
Title:
|
Bayesian Mixtures for Modeling Correlation
|
|
Author(s):
|
Lei Qian*+ and Robert E. Weiss
|
|
Companies:
|
University of California, Los Angeles and University of California, Los Angeles
|
|
Address:
|
3110 Sawtelle Blvd, #308, Los Angeles, CA, 90066,
|
|
Keywords:
|
Covariance model ; Longitudinal data ; Correlation model ; MCMC
|
|
Abstract:
|
In longitudinal data analysis, correctly modeling the covariance matrix is important. Simple structured covariance models have advantages in parameter dimensionality and inference, but it is not necessarily right for complex stochastic processes. We propose a Bayesian correlation mixture model which provides a rich family of correlation models. We decompose the covariance matrix into variances and correlations components. The correlation matrix is modeled by a convex combination of two or more simple correlation matrices. Examples and simulation studies illustrate how our model works with hypothetical and real data correlations.
|
