|
Abstract:
|
Air pollution scientists and policymakers rely heavily on time series study estimates of air pollution health effects. As a design time series studies are a paradox. By comparing day-to-day changes in air pollution concentrations with health event counts, their ecological structure suggests loss of information and bias in the parameter estimates. However, because they rely on huge populations from entire metropolitan areas, such studies are able to detect very small effects. In this talk I discuss the utility and interpretation of time series studies. I frame the discussion in terms of an individual-level disease model to describe the association between personal air pollution exposure and health. I discuss the impact of the ecological design, the specific risk and exposure models, and the need to substitute ambient concentration for personal exposure in the time series analysis.
|
