Implications of Using Ambient PM Concentration as a Surrogate for PM Exposure in Statistical and Epidemiologic Analyses of Time Series Data
The question What information may be obtained from a time-series regression of health effects on ambient PM concentration?, is of interest to exposure analysts and epidemiologists, especially since exposure analysts find low and frequently non-significant correlations between daily ambient PM concentrations and daily personal exposures to total PM. Zeger et al. (200) provide a framework for the analysis of the effect of PM exposure error on the results of epidemiologic analyses using ambient PM concentrations. In their analysis, they made three important assumptions. However, an alternate set of assumptions are possible. This paper will examine the experimental evidence for each set of assumptions, conclusions resulting from each set of assumptions, and key questions posed for exposure analysts and epidemiologists. (Both Zeger and this paper assume that the relationship between exposure and health effect is linear.)
Wilson Jr., W E. Implications of Using Ambient PM Concentration as a Surrogate for PM Exposure in Statistical and Epidemiologic Analyses of Time Series Data. Presented at Joint International Conference of the ISEE/ISEA, Vancouver, BC, Canada, August 11-15, 2002.