|
Abstract:
|
Bayesian analysis is applicable when the number of events under consideration is small. If the event is rare, we can think of the event to follow a Poisson distribution. We examined child blood lead surveillance data in small geographic areas and assumed the Poisson parameter i.e. the prior follows a gamma distribution. Since for children < 72 months old, BLL 5-9 µg/dL is considered unsafe,investigating children with BLL 5-9 µg/dL is important. We used CDC child blood lead surveillance data (public domain) for children aged < 72 months who were tested for lead at least once. Based on 2015 BLL data for children with 5-9 µg/dL in 8 counties in Georgia, we estimated the predictive density of children with BLLs 5-9 µg/dL in X County, Georgia in 2015 with similar level of BLL among children. These estimated predictive probabilities for 0-10 children were 0.169, 0.275, 0.247, 0.161, 0.085, 0.039, 0.016, 0.006, 0.002, 0.001, 0. The results implied that the county was expected to have between 0-3 children with BLLs 5-9 µg/dL . We will compare prior with posterior and also apply the Bayesian model to other counties in Georgia.
|
