Department of Statistical Science
Duke University
presents:
Todd Ogden
ogden@math.sc.edu
University of South Carolina
"Wavelets in Bayesian Change-point Analysis"
Abstract: Wavelet methods have become popular tools in statistical function estimation because of their ability to adapt well to discontinuities and other irregular features of the underlying function. This ability makes wavelets an attractive alternative for use in various versions of the change-point problem. A Bayesian analysis based on the empirical wavelet coefficients of a set of data is developed for the standard change-point problem, in which the prior information on the location of the change-point is expressed in terms of the wavelet domain. This general methodology, developed first for the standard single change-point problem, is easily extended to the problem of detecting jumps or other unusual features in functions that are otherwise smooth by tailoring the prior information to the appropriate type of feature and by considering only higher-level coefficients in the computation of the posterior. This procedure is illustrated by simulated data examples.
April 11, 1997
4:00 pm - 5:00 pm
116 Old Chem Building Any questions concerning the seminar may be addressed to Cheryl McGhee @ [919] 684-8029 or e-mail cheryl@stat.duke.edu. Please contact the author(s) directly for reprints etc.