|
Activity Number:
|
424
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #300228 |
|
Title:
|
Time Series Analysis with Overcomplete Representations
|
|
Author(s):
|
Guy Nason*+ and Alessandro Cardinali
|
|
Companies:
|
University of Bristol and University of Bristol
|
|
Address:
|
, , BS8 1TW, United Kingdom
|
|
Keywords:
|
time series analysis ; locally stationary series ; wavelets ; wavelet packets ; stationarity
|
|
Abstract:
|
The locally stationary wavelet processes (LSW) were introduced in 2000 as an alternative to locally stationary Fourier processes (LSF). This talk first considers testing for second-order stationarity of time series, first by using tests based on overcomplete wavelet and local Fourier analyses and then by a more general analyses based on overcomplete wavelet packet representations. This leads us to the definition of locally stationary wavelet packet processes, a very general model for locally stationary time series. We discuss methods for fitting such processes and the difficulties that arise. We show these methods for analyses and modeling of both stationary and nonstationary series.
|
