|
Activity Number:
|
210
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #301497 |
|
Title:
|
Smoothing Splines ANOVA with Heterogeneous Variances and Application to EEG Data
|
|
Author(s):
|
Ming Dai*+ and Wensheng Guo
|
|
Companies:
|
University of Pennsylvania and University of Pennsylvania
|
|
Address:
|
79 Drexelbrook Dr., Apt. 2, Drexel Hill, PA, 19026,
|
|
Keywords:
|
Cholesky decomposition ; heterogeneous variances ; multivariate time series ; smoothing spline ANOVA ; spectral analysis
|
|
Abstract:
|
EEG is an important way to diagonize the epilepsy. It records the brain electrical activity using electrodes placed on the scalp or implanted in the brain. In clinical practice, the majority of EEG are analyzed visually and the important information about the seizure initialization and propagation is often missed. Guo and Dai (2002) propose to fit the EEG data by locally stationary time series models, transform the problem of estimating the time-varying spectrum into estimating a bivariate surface which can be done by smoothing spline ANOVA (SS-ANOVA). To overcome the difficulty that the spectral matrix is positive definite while the elements may have different smoothness, they propose to estimate the Cholesky decomposition of the spectrum using homogeneous SS-ANOVA and then reconstruct the final spectral estimate. But in fact, the variances of the Cholesky decompositions are not identical. In this paper, we consider the SS-ANOVA with heterogeneous variances to smooth the Cholesky decomposition. The method improves the performance of the estimate as well as the bootstrap inference. And the SS-ANOVA with heterogeneous variances can be extended to more general settings.
|
