A Hierarchical Bayes Approach to Estimation and Prediction for Time Series of Counts

, , and

BibTeX reference

In this paper, we are concerned with the statistical methodology of epidemiological surveillance; that is, the ongoing procedure of analyzing and interpreting public health data of infectious disease incidence. In particular, we propose a hierarchical Bayes approach for the estimation of generalized linear mixed models for time series count data, and their use in the prediction of counts for future time periods. The estima- tors are obtained by Gibbs sampling and their performance is compared to those of other methods on the polio data originally analysed by Zeger (1988), which consist of the monthly number of U.S. polio cases between 1970 and 1983. Their properties are also investigated via simulation. Our aim is to illustrate how easily the hierarchical Bayes methodology lends itself to model checking and model comparisons. The pro- posed methodology, in particular, hierarchical Bayes prediction, is applied to a series of Campylobacter infection cases in the Montreal-Center region.

, 20 pages

This cahier was revised in February 2007


G-2003-25.pdf (100 KB)