|
Activity Number:
|
288
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #306693 |
|
Title:
|
Applications of Copulas To Improve Covariance Estimation for PLS
|
|
Author(s):
|
Gina D'Angelo*+ and Lisa Weissfeld and Scott Ziolko and Chester Mathis and William Klunk and Steven DeKosky and Julie Price
|
|
Companies:
|
University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh
|
|
Address:
|
Department of Biostatistics, 130 DeSoto Street, Pittsburgh, PA, 15216,
|
|
Keywords:
|
covariance ; PLS ; brain imaging ; PET imaging ; Alzheimer's disease
|
|
Abstract:
|
Dimension reduction techniques such as partial least squares (PLS) are currently being applied to classification problems in the area of genetics. These methods have also been applied to PET imaging data with the goal of creating summary measures and examining relationships between voxel-level data and covariates of interest. Previously, we have examined the use of standard PLS techniques for the analysis of amyloid deposition in AD and control subjects using PIB PET imaging techniques (Ziolko et al., 2005). The present work extends PLS to accommodate the unique correlation structure of this data set, for which the distribution of PIB voxel intensity values is a mixture of normal distributions while that of FDG PET is a single normal distribution. This extension is implemented by using a copula to estimate the covariance structure and illustrated in the PIB/FDG data set.
|
