This chapter begins with two simple illustrations of the power of combining multiple data sources for surveillance. We then survey four representative multivariate approaches. The first two (multiple regression and the Hotelling T-squared test) are conventional and time-honored statistical approaches. Next, we describe how a famous probabilistic approach called hidden Markov models can be used for outbreak diagnosis from many data streams. Finally, we discuss WSARE, a system that searches for anomalous subsets of multivariate records.
Was this article helpful?