Details

Autor(en) / Beteiligte

• Gryparis, Alexandros

Titel

Bayesian semiparametric regression methods with applications to environmental epidemiology

Ort / Verlag

ProQuest Dissertations & Theses

Erscheinungsjahr

2006

Quelle

ProQuest Dissertations & Theses A&I

Beschreibungen/Notizen

• The potential effects of ambient air pollution on health are a major public health issue and have received much attention in recent years. The last decade more environmental epidemiologists have focused on spatial variability in air pollution and its effects on human health. This thesis focuses on developing methods for assessing the health effects of ambient air pollution on humans. In Chapter 1 we propose a semiparametric latent variable regression models for modeling multiple surrogates of a single pollution source. Our proposed models, which combine attractive features of geoadditive models for spatial data and latent variable models for multiple exposures, allow for both flexible, nonlinear effects of covariates and for unexplained spatial and temporal variability in exposure. We use a penalised spline formulation to specify temporal and spatial correlations on the latent pollution variable. Our penalised spline formulation of the model leads to a natural Bayesian Markov Chain Monte Carlo algorithm for model fitting. In Chapter 2 we provide a framework for spatial measurement error modeling, showing that smoothing induces a Berkson-type measurement error with nondiagonal error structure. From this viewpoint, we review several existing approaches to estimation: direct use of the spatial predictions, multiple imputation, and Bayesian approaches. We also propose two new approaches to estimation, one based on regression calibration and another based on iterative weighted least squares. We evaluate and compare all these methods both in principle and via simulations. In Chapter 3 we focus on measurement error in hierarchical models. A major concern in studies that address the health effects of air-pollution is whether an observed association between one pollutant and one outcome is due, all or in part, to the correlation between that exposure and either a second pollutant or a confounder. The addition of measurement error to such data, complicates matters further. In Chapter 3 we describe a fully Bayesian extension that extends the two-stage approach in Schwartz and Coull (2003), and compare its performance to the two-stage approach via simulations.