|
Activity Number:
|
463
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #309748 |
|
Title:
|
Multisubject Independent Component Analysis Using a Maximum Likelihood Approach
|
|
Author(s):
|
Ying Guo*+
|
|
Companies:
|
Emory University
|
|
Address:
|
1518 Clifton RD NE, Atlanta, GA, 30322,
|
|
Keywords:
|
independent component analysis ; functional magnetic resonance imaging ; Multi-subject ICA ; maximum likelihood approach ; likelihood ratio tests ; test of group differences
|
|
Abstract:
|
Independent component analysis (ICA) has become increasingly popular for analyzing functional magnetic resonance imaging (fMRI) data. The extension of ICA for group inferences is an active research topic. Existing group ICA models assume different underlying structures of group spatio-temporal processes. There is currently no method for assessing the appropriateness of the assumed structure in a data set. Another challenge in group ICA analysis is how to test differences between subject groups. I propose a maximum likelihood approach which offers a unified framework for estimating and comparing group ICA models with different underlying structures. A class of likelihood ratio tests are derived to assess the goodness-of-fit of a group ICA model and to test the homogeneity in ICA between two groups. Simulation studies and application to a fMRI data example would be discussed.
|
