Wavelet-Based Monitoring for Disease Outbreaks and Bioterrorism: Methods and Challenges
by Bernard L. Dillard and Galit Shmueli.
We introduce a set of flexible statistical methods for monitoring univariate or multivariate streams of syndromic data for the purpose of quickly detecting anomalous behavior associated with a large-scale but localized bioterrorist attack or disease outbreak. These wavelet-based methods were developed for and applied in the field of chemical engineering (Bakshi 1998; Aradhye 2003). They have proven useful for monitoring non-stationary processes with unknown anomalous structure. The univariate method is Multiscale Statistical Process Control (MSSPC), and its multivariate extension is Multiscale Principal Components Analysis (MSPCA). We show that these methods are useful for syndromic data, which usually cannot be assumed to be normal, time-independent, or even stationary, and where the manifestation of the outbreak in the data is unknown. Using real data, we show that these wavelet-based techniques outperform the EWMA-based monitoring method, as each vies to detect out-of-control processes early.
Discrete Wavelet Transform, Statistical Process Control, Anthrax, Principal Components Analysis, False Discovery Rate, Multivariate Monitoring
Bernard L. Dillard, firstname.lastname@example.org
Galit Shmueli, email@example.com
READING THE ARTICLE: You can read the article in
portable document (.pdf) format (526681 bytes.)
NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.
This page has been accessed 3079 times since MARCH 14, 2010.
Return to the Home Page.