Early detection of disease outbreaks has been the focus of recent national efforts. I am currently working on exploring and evaluating univariate and multivariate monitoring methods that can be used to track multiple traditional and non-traditional time series for detecting an outbreak.
Research Group My research group conducts applied research in designing and evaluating statistical and data mining methods for monitoring biosurveillance data. Our group includes students in statistics, applied math, and business at UMD. We collaborate with researchers at the Johns Hopkins Applied Physics Lab. See relevant publications. Recent research projects include:
- Characterization and preprocessing of biosurveillance data
- Simulation of authentic multivariate data (See www.projectmimic.com for information, data, and code for generating semi-authentic multivariate biosurveillance data)
- Forecasting and modeling background syndromic series
- Monitoring series via statistical control charts (Matlab code for generating various control charts including wavelet-based charts is available here)
- Developing and evaluating univariate and multivariate monitoring algorithms.
A major theme is tackling multiplicity (multiple algorithms, multiple data sources, multiple data streams, etc.).