Fall Seminar Series - December 8, 2005
University of Minnesota
School of Statistics
College of Liberal Arts
Spectral
and Wavelet-based Methods for Nonstationary Processes
Peter Craigmile
Department of Statistics
The Ohio State University
Thursday, December 8, 2005
3:30 PM, 115
Ford Hall
Minneapolis, East Bank Campus
Social at 3:00 PM, 300 Ford Hall
Abstract
It
is a standard practice in time series analysis to transform a time
series to stationarity using, for example, detrending or differencing.
This is because weakly stationary processes are characterized by only
the mean and autocovariance function (or equivalently the spectral
density function). This is an over-simplification, especially when the
time-varying dependence is of key interest to the application. There is
a growing need to be able to investigate, discriminate, and model
multivariate time-varying structures, especially in areas such as
hearing science, neurosciences, and atmospheric science. This talk will
discuss the use of spectral and wavelet-based methods for examining
nonstationary phenomena. We will apply our methodology to the analysis
of distortion products, used in non-behavioral tests of hearing.
This
work is based on collaborations with Drs. Brandon Whitcher, Peter
Brown, and Wayne King.