Different Approaches to Model Averaging in Environmental Risk Assessment

• Activity Number: 78
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Section on Risk Analysis
• Author(s): Esben Budtz-Jørgensen*+
• Companies: University of Copenhagen
• Keywords: benchmark dose ; Bayesian statistics ; dose-response model ; model uncertainty

Abstract

The Benchmark approach is the most popular method for calculation of safe exposure standards. An important advantage of this method is that statistical uncertainty is taken into account. Thus the main result, the BMDL, is a lower confidence limit of the dose which leads to a specific increase in the risk. Still, standard applications of this method ignores model uncertainty and may therefore provide results that are biased toward higher a less protective dose levels. Model averaging has been proposed to solve this problem. This paper compares different approaches to model averaging. It is illustrated that the BMDL results may depend heavily on the choice of method as different philosophies lead to different assessments of the uncertainty in the model averaged estimates. Differences are illustrated theoretically and in epidemiological data on effects of exposure to mercury and arsenic.