Spatial Statistics Applied to Environmental Health Problems

• Activity Number: 116
• Type: Topic Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): Annibale Biggeri* and Michela Baccini and Dolores Catelan and Corrado Lagazio
• Companies: University of Florence and University of Florence and University of Florence and University of Genoa
• Keywords: spatial misalignement ; informative sampling ; bayesian geostatistics ; tri-level bayesian hierarchical models ; impact assessment ; uncertainty

Abstract

Long-term effects of air pollution were estimated by cohort studies in several countries. Yet, appropriate statistical framework and uncertainty propagation is object of current research. Spatial misalignments (Gelfand et al. 2012), preferential sampling of monitoring networks (Diggle et al. 2010) and integration of deterministic model outputs (Berrocal et al. 2010) are recognized problem in this field. Health impact estimates are produced for policy planning and integrates several source of information. Uncertainty propagation is addressed by MonteCarlo techniques using probability distributions for each key model parameters (example in short-term impact estimates is in Baccini et al. 2015). Bayesian modeling of effect estimates for impact assessment is usually done by two-level Bayesian hierarchical meta-analysis. We show how to develop an integrated impact assessment of long-term effect of air pollution in a Bayesian framework. We integrate uncertainty of spatial predictions from deterministic model outputs and observed data derived from informative sampling with marginal posterior estimates of effects measures from tri-level Bayesian hierarchical models (Catelan et al. 2010).