Info

Notes: Experiments are described in the text. In each column, the winning method is highlighted.

Notes: Experiments are described in the text. In each column, the winning method is highlighted.

many interesting statistical details. Here, we provide pointers to other popular or promising approaches.

• Wavelets are another popular approach to smoothing time series data, and wavelets of varying resolutions can be used to model data properties at both broad and fine time resolutions. Examples used in biosurveillance include Zhang et al. (2003) and Goldenberg et al. (2002).

• Changepoint statistics (Carlstein, 1988, Buckeridge et al., 2005, Baron, 2002) are increasingly popular in the statistics literature for noticing when an underlying process changes.

• Kalman filters (Hamilton, 1994) or hidden Markov models (Rabiner, 1989) would be appropriate for biosurveillance time-series modeling. Madigan (2005) contains an overview of the use of hidden Markov models in surveillance.

• Much of this chapter used a Gaussian model for deriving confidence intervals and thus alarm levels. It is common to use different distributions such as a Poisson model.

Was this article helpful?

0 0

Post a comment