|
Activity Number:
|
85
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #300572 |
|
Title:
|
Threshold Effects in Air Pollution Mortality Associations
|
|
Author(s):
|
Lise A. Tole*+ and Gary M. Koop
|
|
Companies:
|
Leicester University and Leicester University
|
|
Address:
|
Dept. of Economics, Leicester, LE7 1RH, United Kingdom
|
|
Keywords:
|
nonlinear time series ; Markov chain monte carlo ; dose-response relationship ; Bayesian model averaging
|
|
Abstract:
|
Important policy decisions hinge on whether air pollution has significant effects on human health. A topic of particular interest is whether there are thresholds below which pollutants are harmless. Numerous studies have used daily time series data on a pollutant (often particulate matter) and a health outcome (often mortality) to investigate possible thresholds in the dose-response relationship. The statistical methods used in previous studies typically use splines or non- or semi-parametric methods such as the general additive model. We argue that such approaches may not be suitable when a myriad of possible explanatory variables (and lags) could trigger threshold effects. We introduce a flexible parametric model that allows for the possibility that threshold effects could occur in one (or more) of many possible ways. Multiple potential thresholds are defined according to levels and growth rates of a pollutant (or a weather variable or an interaction between a weather variable and a pollutant) on the current day or the recent past (or averaged over the recent past). We develop Bayesian methods for statistical inference and apply our methods to an extensive dataset.
|
