Model Uncertainty in Health Effect
Studies
Merlise Clyde
There are many aspects of model choice
that are involved in health effect studies of particulate matter and other
pollutants. Some of these choices concern which pollutants and confounding
variables should
be included in the model, what type
of lag structure for the covariates should be used, and which interactions
need to be considered, and how to model nonlinear trends. Because
of the large number of potential variables, model selection is often used
to find a parsimonious model. Different model selection strategies
have led in some cases to very different models and conclusions for the
same set of data. As variable selection may involve numerous tests
of hypotheses, the resulting significance levels may be called into question,
and there is the concern that the positive associations are a result of
multiple testing. I will describe methods that can be used
to describe the level of uncertainty due to model selection, and can be
used to combine inferences by averaging over a wider class of models using
readily available summary statistics from standard model fitting programs.
MERLISE CLYDE
Box 90251
ISDS, Duke University
210A Old Chemistry
Durham, NC 27708-0251, USA
clyde@isds.duke.edu
Assessing Sources of Variability
in Measurements of Ambient Particulate Matter
Mark S. Kaiser and Michael J.
Daniels
Spatial and spatio-temporal models of ambient particulate matter (PM) have the potential to provide accurate and precise predictions of exposure to outdoor PM for regulatory purposes and to assess health effects. Formulation of such models requires an understanding of the types and levels of variability that are likely to be exhibited in data records of ambient particulate matter. We utilize a data set consisting of daily observations of PM_10 from air quality monitoring stations in Ohio, West Virginia, and Pennsylvania in the United States. A hierarchical model is used to investigate variability between two major groups of sites lying in separate drainage basins, temporal and spatial variability among individual sites, variability that can be modeled as functions of meteorological covariates, and micro-scale variability among observations obtained on the same day from several measurement apparatuses attached to the same monitoring towers. This information will prove valuable in subsequent efforts to develop spatial and spatio-temporal models for ambient particulate matter concentration and to connect predictions of such concentrations over space and time to models of human health outcomes.
MARK KAISER
Department of Statistics
Iowa State University
Ames, IA 50011, USA
mskaiser@iastate.edu
Modeling Short-Term Air Pollution
Health Effects Using
Surrogate Exposure Measurements
from Ambient Monitors
Lianne Sheppard
In this talk I will discuss modeling of short-term health effects from air pollutants. I will outline a plausible disease model conditional on true but unknown individual pollutant exposures. Since only surrogate exposure data are available from ambient monitors, I will link these to the disease model by proposing exposure and measurement models. I will discuss aggregation of the disease model given the typically available health outcome data from administrative records in a community. The aggregation will impact the surrogate exposure measurements as well as the weights needed in estimation. I will compare this approach with that usually taken in air pollution epidemiology studies and assess its impact in a data set of asthma hospital admissions in Seattle.
LIANNE SHEPPARD
Department of Biostatistics
University of Washington
Box 357232
Seattle, WA 98195-7232, USA
sheppard@biostat.washington.edu