Evaluating rates of true and false positives in Bayesian disease mapping
Publisher: Universidad de Murcia. Departamento de Estadística e Investigación Operativa
ISBN: 978-84-691-8159-1
Year of publication: 2009
Congress: Congreso Nacional de Estadística e Investigación Operativa (31. 2009. Murcia)
Type: Conference paper
Sustainable development goals
Abstract
Empirical Bayes (EB) and Fully Bayes (FB) approaches have been used for smoothing rates in disease mapping. However, these techniques are not free from inconveniences as an excess of smoothing might hinder the detection of true high-risk areas. Identifying regions with extreme risks minimizing the misclassi cation of normal areas is a primary goal in epidemiology. The FB approach exploits the posterior distribution of the relative risks de ning Bayesian decision rules to detect raised-risk areas. These rules can not be applied under the EB approach because only point estimates are available. Then, second order correct estimators of the mean squared error (MSE) of the log-relative risk predictor can be used to derive con dence intervals for the relative risks. The aim of this work is to compare both procedures in terms of sensitivity (true positives) and speci city (1-false positives).