University of Michigan
154 Hurley Hall
Addressing The Change Of Support Problem (COSP) For Environmental Exposure Assessment
The Change Of Support Problem (COSP) is a classical problem in spatial statistics, very pervasive in applications involving spatial and spatio-temporal datasets. In this talk, we will present two approaches to address the COSP with the common goal of generating environmental exposures to be used in social and environmental epidemiology. In the first talk, we will focus on air pollution and we will present a Bayesian hierarchical spatio-temporal model that fuses two data sources with different spatial resolution to yield improved estimates of daily air pollution at point level. In the second talk, we will work with estimates of neighborhood socio-economic indicators developed from the American Community Survey, a multi-year nationwide survey conducted by the U.S. Census Bureau, and we will present a Bayesian hierarchical model to disaggregate these estimates spatially and temporally while accounting for the survey design effect.