P. K. Sen and C. R. Rao, eds., Handbook of Statistics, Vol. 18
© 2000 Elsevier Science B.V. All rights reserved.
k.y
Hierarchical and Empirical Bayes Methods
for Environmental Risk Assessment
Gauri Datta, Malay Ghosh and Lance A. Waller
The study of geographical variability in mortality rates for diseases such as cancer
has received a great deal of attention from statisticians in recent years. A con-
siderable amount of effort has been devoted to the production of disease atlases,
i.e., collections of maps in which the geographical distribution of incidence of a
specific disease is presented across a country or region. Typically, the study area is
partitioned into smaller spatial districts, and the disease map displays local rates
(incidence or mortality) associated with each district. The main problem faced in
production of these atlases is the choice of a suitable measure of local mortality
(or incidence of disease) to be mapped.
This paper reviews hierarchical and empirical Bayes methods for the estima-
tion of mortality rates. Unlike standardized mortality ratios, these methods are
particularly well-suited to account for spatially structured heterogeneity in disease
risks. The inference is based on certain spatial generalized linear models. Spatio-
temporal models are also discussed. The papers concludes with a data example.
1. Introduction
The study of geographical variability in mortality rates for diseases such as cancer
has received a great deal of attention from statisticians in recent years. A con-
siderable amount of effort has been devoted to the production of disease atlases,
i.e., collections of maps in which the geographical distribution of incidence of a
specific disease is presented across a country or region. Typically, the study area is
partitioned into smaller spatial districts, and the disease map displays local rates
(incidence or mortality) associated with each district. The main problem faced in
production of these atlases is the choice of a suitable measure of local mortality
(or incidence of disease) to be mapped. Until recently, the two widely used
measures were ratios of standardized rates (such as standardized mortality ratios)
and P-values of statistical significance of local deviations of risk from the overall
risk averaged across the entire study region.
223