|
Activity Number:
|
442
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #302413 |
|
Title:
|
Informative Priors for Computing the UCL of a Set of Asbestos Measurements with Low Total Counts
|
|
Author(s):
|
Bradley C. Venner*+ and Bill Brattin and Terry Schulz and Timothy Barry
|
|
Companies:
|
U.S. Environmental Protection Agency and Syracuse Research Corporation and Independent Contractor and U.S. Environmental Protection Agency
|
|
Address:
|
PO Box 25227, Denver, CO, 80232,
|
|
Keywords:
|
asbestos ; Poisson-lognormal ; exposure point concentration ; Superfund
|
|
Abstract:
|
USEPA guidance on risk assessment at Superfund sites recommends calculation of an exposure point concentration as an upper confidence limit on the mean of the samples. This approach is difficult when measuring asbestos. These measurements often occur with few if any observed fibers due to the cost of manually counting asbestos structures under high magnification. Another complication is that there are two major sources of variability---in the observed number of structures due to counting errors and the air concentration of interest. Although a compound Poisson-lognormal model may be used to describe both sources of variability, this model suffers from a lack of identifiability at low number of counts. We propose the use of informative priors for the log-normal parameters, present the results of simulation studies on the sensitivity of this prior, and discuss experimental design.
|
