Dissertation/Thesis Abstract

Model uncertainty and model averaging in the estimation of benchmark dose
by Kim, Steven B., M.S., California State University, Long Beach, 2010, 93; 1486380
Abstract (Summary)

Food-borne infection is caused by intake of contaminated foods or beverages. Since there is no "default model" for microbial risk assessment, model averaging in the estimation of benchmark dose has been studied to analyze microbial dose-response experiments. In this research, several dose-response models including four two-parameter models and four three-parameter models are used. Parameters of the statistical models are estimated by maximum likelihood method. The benchmark dose is estimated by a weighted average of effective dose estimates from the eight models, and the weights are determined by Kullback information criterion to account for model uncertainty. Both model uncertainty and data uncertainty are addressed to compute the variance of the benchmark dose estimate, and a bootstrap-based 95% confidence interval of benchmark dose is constructed. To evaluate the coverage probabilities of the confidence limits, a Monte Carlo simulation study is conducted in various conditions based on a real data set in human volunteers.

Indexing (document details)
Advisor: Moon, Hojin
School: California State University, Long Beach
School Location: United States -- California
Source: MAI 49/01M, Masters Abstracts International
Subjects: Applied Mathematics, Mathematics, Statistics
Publication Number: 1486380
ISBN: 9781124247700
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy