Choice of Microbial Dose-Response Models Under Consideration of Attendant Sampling Error. A. S. Vicari, R. A. Morales, L. A. Jaykus, and P. Cowen, North Carolina State University-Raleigh, Research Triangle Institute
Microbial risk assessors have often gone to extreme lengths in justifying the choice of a particular dose-response model, and eventually in stating the reliability of their model’s prediction. In doing so, consideration of the uncertainty potentially inherent in a specific model choice has not been the object of analogous scrutiny. In this study, the bootstrap method is applied to characterize the sampling error involved in fitting the beta-Poisson model and the log-logistic model to two different data sets from volunteer feeding studies (Campylobacter jejuni and Shigella dysenteriae). Bootstrapping is carried out through three different resampling schemes - of which two are first proposed here - that represent three distinct aspects of sampling error. The results show that, while the uncertainty due to model choice dominates at doses smaller than 10 bacterial cells, the resampling scheme becomes somewhat more critical at the highest doses (>1E+6). However, within the range of the experimental data, the results are virtually equivalent regardless of the dose-response model and/or the resampling scheme. This study demonstrates that, when uncertainty of microbial dose-response models is considered, the relevance of the risk assessors’ choices depends on the dose range under consideration.
Go to . . .
2002 SRA Annual Meeting Table of Contents
2002 SRA Annual Meeting Author Index
Main Abstracts Menu Page
RiskWorld Home Page