|
Activity Number:
|
361
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #301189 |
|
Title:
|
Analyzing Bivariate Time Series via Nonparametric Likelihood
|
|
Author(s):
|
Suddhasatta Acharyya*+ and Hernando Ombao
|
|
Companies:
|
Brown University and Brown University
|
|
Address:
|
Center for Statistical Sciences, Providence, RI, 02912,
|
|
Keywords:
|
nonparametric-likelihood ; cross-spectrum ; coherence ; time-series ; brain-signal
|
|
Abstract:
|
Cross-spectrum and coherence are widely used measures for characterizing linear dependence between two time series. Inference on these quantities relies on the assumption of asymptotic Gaussian distributions. Such an assumption could be problematic when the length of the time series is not sufficiently large and when the noise innovations come from distributions with heavy tails. Nonparametric likelihood-based methods have provided competitive alternatives to standard parametric approaches for many problems. The application of such methods to the spectral domain has not been well-studied. In this paper, we shall assess the performance of nonparametric likelihood based estimation for spectral time series through simulation. We shall also apply the method to multi-channel brain signal data, and discuss possible implications.
|
