Atlas home ||
Conferences |
Abstracts |
about Atlas
Sequential L1 procedures for change detection
by
Zuzana Prášková
Charles University in Prague, Czech Republic
Abstract
We consider a sequential procedure monitoring a possible change in the shift in a sequence of weakly dependent observations. The procedure is based on cumulative sums of L1 residuals that are more robust with respect to outliers and heavy tail distributions that exhibit e.g. financial time series. We study asymptotic distribution of the detector both under the null hypothesis (no change) and the alternative that the change occurred after a period of stability. We will also prove that the distribution of the stopping time of monitoring process is asymptotically normal under a suitable standardization. Effect of dependency appearing in the standardization factor will be discussed. The proving methods are based on the invariance principle, law of the iterated logarithm, and a Bahadur-type representation of sample quantiles for mixing sequences.
Date received: April 13, 2007
Copyright © 2007 by the author(s).
The author(s) of this document and the organizers of the conference
have granted their consent to include this abstract in
Atlas Conferences Inc.
Document # cauc-62.