A0188
Title: Ensemble Kalman filtering for Bayesian disease map surveillance
Authors: Andrew Lawson - Medical University of South Carolina (United States) [presenting]
Abstract: Kalman filtering has been commonly used for time series modeling and the prediction of time-based systems. In the exploration of application to disease mapping, the conversion of count data is considered from small areas to a Gaussian-like variable: Log(SMR). Because the log(SMR) behaves closely like a Gaussian, the use of KF methods is explored for space-time data. However, for fitting purposes, especially for online surveillance, the need to invert large matrices is unattractive. Instead, the use of ensemble KF is explored, which bypasses the inversion by using simulation-based updating, closely related to particle filtering. It has been found that very fast updating of system and measurement models can be made without (much!) recourse to MCMC, and so it would appear to provide a useful tool in applications where fast updating is required, such as during large infectious disease outbreaks.
