Access the full text.
Sign up today, get DeepDyve free for 14 days.
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
tially indexed time series, in the case of empirical or- thogonal function analysis). The results of singular spectrum analysis consist of a number of orthogonal Addison-Wesley index series, each a linear smoothing of the original Jaco b Way, Rt. 128 time series, with variance decreasing from the maxi- Reading , MA 0186 7 mum possible. Actually, the index series are techni- Telephone : 800-447-2226 cally not linear smoothings of the original time se- ries. The reason is that the weights used to construct Computational Mechanics, Inc. them depend on an estimate of the autocovariance 2 5 Bridge St. function of the original time series, and this estimate Billerica, MA 0182 1 is itself a nonlinear function of the original time Telephone : 508-667-5841 series. As in principal component analysis, once the in- Oxford University Press dex series have been constructed, they can be inter- 19 8 Madison Ave. preted (e.g., as secular trend, a periodic oscillation, ne w York, NY 10016-4314 etc.) and the original time series can be projected onto Telephone : 800-451-7556 a subset of them. An important problem concerns the assessment of the significance of one or more of the Plenum Publishing index series. Eisner and Tsonis discuss this problem 23 3 Spring St. in some detail. They point out that, in assessing sig- Mew York, NY 10013 nificance, it is important to specify the null model. Telephone : 800-221-9369 Too often in this kind of work the null model is white noise, which is unrealistic for many, if not most, geo- University Science Books physical time series. 2 0 Edgehill Rd. The final part of the book describes the application Mill Valley, CA 9494 1 of singular spectrum analysis to signal detection, fil- 415-383-143 0 tering, prediction, and, although it is somewhat out of place, phase space reconstruction. The authors cover Water Resources Publications a lot of ground here and it is clear that some sorting out of methods needs to be done. 720 0 East Dry Creek Rd. Suit e E-104 I am not enthusiastic about singular spectrum analy- Englewood , CO 80112 sis. My main reservation is that I do not believe that it Telephone : 800-736-2405 is possible to get much out of a time series without making some assumptions about its structure. I have a specific worry that the orthogonality constraint ap- plied in singular spectrum analysis can lead to arti- facts to which physical interpretations are attached. I mm \ nou also worry that statistical issues, such as the assess- ment of the significance of a component, have not been The Bulletin relies on its readership for timely infor- thought out carefully. That said, I am enthusiastic about mation on members' accomplishments, chapter this book. It provides a clear, concise description of a happenings, conferences, and science fairs. method that has gained some currency in atmospheric • About Our Members science. Whether this method stands or falls, it is im- • Chapter News portant that it be understood. This book certainly con- • Announcements tributes to that understanding.—Andrew R. Solow. • Science Fairs Andrew R. Solow is a researcher at Woods Submit contributions to Bulletin News Editor, AMS, Hole Oceanographic Institution, Woods Hole, 45 Beacon St., Boston, MA 02108-3693; fax 617- Massachusetts. • 742-8718; e-mail: jburba@ametsoc.org. Bulletin of the American Meteorological Society 2257
Bulletin of the American Meteorological Society – American Meteorological Society
Published: Oct 1, 1997
Access the full text.
Sign up today, get DeepDyve free for 14 days.