STATISTICS COLLOQUIUM
Gavino Puggioni
Assistant Professor
Department of Computer Science & Statistics
The University of Rhode Island
Bayesian Spatio-Temporal CAR Models for Tropical
Diseases Outbreak Prediction
ABSTRACT
This projects proposes the first comprehensive spatio-temporal model that links reported cases of Dengue in Puerto Rico (monthly records from 1990 to 2015 in 76 municipalities) with climatic variables collected at 34 weather stations and land use satellite data. Dengue and Zika are mosquito-borne tropical diseases, reported with increasing rates in the last decade. Early warning systems help in predicting outbreaks and allow public health decision-makers to implement preventive measures. Several factors have been linked to the increase in reported cases: changes in temperature, precipitation, urbanization, and other spatial variables. Several space-time CAR specifications are implemented to assess the relative risk of these factors, as well as to set a predictive framework. The proposed models use a two stage approach to account for the different spatial support of predictors and response. Estimation is carried using MCMC algorithms and ideas for dynamic extensions are discussed.
DATE: Wednesday, April 20, 2016
TIME: 4:00 pm – 5:00 pm
PLACE: Philip E. Austin Bldg., Rm. 105
Coffee will be served at 3:30 in the Noether Lounge (AUST 326)
For more information, contact: Tracy Burke at tracy.burke@uconn.edu