| Title: | Part 2. Development of Enhanced Statistical Methods for Assessing Health Effects Associated with an Unknown Number of Major Sources of Multiple Air Pollutants |
| Author(s): | Park ES; Symanski E; Han D; Spiegelman C; |
| ISSN/ISBN: | 1041-5505 (Print) 1041-5505 (Linking) |
| Abstract: | "A major difficulty with assessing source-specific health effects is that source-specific exposures cannot be measured directly; rather, they need to be estimated by a source-apportionment method such as multivariate receptor modeling. The uncertainty in source apportionment (uncertainty in source-specific exposure estimates and model uncertainty due to the unknown number of sources and identifiability conditions) has been largely ignored in previous studies. Also, spatial dependence of multipollutant data collected from multiple monitoring sites has not yet been incorporated into multivariate receptor modeling. The objectives of this project are (1) to develop a multipollutant approach that incorporates both sources of uncertainty in source-apportionment into the assessment of source-specific health effects and (2) to develop enhanced multivariate receptor models that can account for spatial correlations in the multipollutant data collected from multiple sites. We employed a Bayesian hierarchical modeling framework consisting of multivariate receptor models, health-effects models, and a hierarchical model on latent source contributions. For the health model, we focused on the time-series design in this project. Each combination of number of sources and identifiability conditions (additional constraints on model parameters) defines a different model. We built a set of plausible models with extensive exploratory data analyses and with information from previous studies, and then computed posterior model probability to estimate model uncertainty. Parameter estimation and model uncertainty estimation were implemented simultaneously by Markov chain Monte Carlo (MCMC*) methods. We validated the methods using simulated data. We illustrated the methods using PM2.5 (particulate matter |
| Keywords: | "Air Pollutants/*adverse effects/chemistry/pharmacology Air Pollution/*adverse effects/analysis Artificial Intelligence Bayes Theorem Computer Simulation Data Interpretation, Statistical Environmental Exposure/*adverse effects/analysis Environmental Monito;" |
| Notes: | "MedlinePark, Eun Sug Symanski, Elaine Han, Daikwon Spiegelman, Clifford eng Research Support, U.S. Gov't, Non-P.H.S. 2015/09/04 Res Rep Health Eff Inst. 2015 Jun; (183 Pt 1-2):51-113" |
