The Sensitivity of Concentration-Response Functions to the Explicit Modeling of Space-Time Dependence

University Of Texas At Dallas, Richardson TX

Investigators

Abstract

This research will assess the gains from explicitly incorporating space-time dependence in traditional air pollution concentration-response functions. It has a methodological and an application component. The methodological component involves evaluating the benefits and costs of various models that are based on sometimes conflicting statistical paradigms; e.g., Bayesian vs. frequentist time series extensions, geostatistical vs. lattice spatial models, and analytical vs. computational models. The application concerns the measurement of the effects of air pollution on human health, as measured by hospital admissions, in California's South Coast Air Basin. Previous applications have ignored the spatial variation in the data, losing potentially valuable information. The research requires (1) space-time interpolation to obtain reliable surfaces of pollutant concentrations; (2) exploratory spatial data analysis (ESDA) and visualization of model fit; and (3) estimation of space-time linear statistical models. The research agenda is designed to uncover and highlight the most promising areas where space-time analyses can contribute to policy analysis. An important aspect of research underlying U.S. environmental policy is concerned with the precise measurement of the relation between the ambient concentrations of criteria pollutants and socio-economic and health outcomes. The anticipated increase in precision and accuracy from space-time analysis facilitates an assessment of several key policy questions, including: what are the best empirical measures for air pollution concentrations; what is the role of various pollutants in explaining responses; are there no observable effects thresholds in concentration-response functions; and what is the range of uncertainty in the predictions from the various models. To the extent that, as a result of this methodological and empirical investigation, society gains insight into the accuracy and uncertainty from estimated models, this research will contribute to the refinement of the analysis of the costs and benefits of US environmental policy.