Remote sensing data, chemical transport
simulations, and land use terms have complementary advantages
in predicting air pollution , and these may interact with
meteorology. Ensembles of machine learners are good at
combining such predictors including nonlinearities and product
terms. I will illustrate this with models predicting PM2.5,
NO2, and O3, and applications to studies of air pollution and
mortality, and fine scale risk assessments within cities.
About the Speaker:
Joel Schwartz is a Professor in the departments of Environmental
Health and Epidemiology at the Harvard School of Public Health,
on the steering committee of the Harvard University Center for
the Environment, and Director of the Harvard Center for Risk
Analysis. His major research interests include health effects of
air pollution, of heavy metals, climate change, and drinking
water, epidemiological methods, risk assessment and cost benefit
analyses. He has examined these questions using a variety of
methods including time series, case-crossover, and case-only
analyses of administrative data, survival and repeated measures
analyses of cohorts, repeated measures analyses of panel
studies, etc. He is particularly interested in
quasi-experimental designs and other causal models. These have
included a range of outcomes including cognitive function, lung
function, asthma, heart attacks, strokes, deaths, blood
pressure, lipid levels, biomarkers of inflammation and oxidative
stress, markers of biological aging, and epigenetic changes. He
is particularly interested in social and other factors conveying
increased susceptibility. In addition, he has been involved in
exposure modeling, using machine learning to combine land use
data, remote sensing data and chemical transport models. He is
also interested in methodological issues, including
dose-response modeling, causal modeling, and data fusion. Dr.
Schwartz' benefit-cost analysis on lead in gasoline was
responsible for its elimination in the United States, and his
methodology for valuing the benefits of reducing toxins that
have cognitive effects is widely used. He is the recipient of a
John D. and Catherine T. MacArthur Fellowship, and the John
Goldsmith Award from the International Society for Environmental
Epidemiology.