Advances in Quantitative Benchmark Risk Assessment for Environmental Health and Educational Settings

Abstract

This dissertation is a unifying document of three related projects developed during my dissertation research. The projects involve comparison of existing methods for estimating the benchmark dose (BMD) and its lower confidence limit (BMDL), a Bayesian calculation for mixed-factor quantal data to estimate the BMD and BMDL, and an application of BMD methodology in educational setting. Briefly, the first project compares the performance of current methods for BMD and BMDL calculation. We use 379 data sets from a curated Quantal Risk Assessment Database (QRAD) to compare performance of Bayesian model averaging (BMA), frequentist model averaging (FMA) and nonparametric regression modeling methods for BMD and BMDL calculations. The second project applies Bayesian strategies to a mixed-factor setting with a secondary qualitative factor possessing two levels to derive two-factor Bayesian BMDs and BMDLs. We present reparametrized dose-response models that allow for explicit use of prior information on the target parameter of interest, the BMD. We also enhance our Bayesian estimation technique for BMD analysis by applying Bayesian model averaging to produce BMDs and BMDLs. Lastly, the third project applies risk benchmarking methodology traditionally used in toxicology and environmental settings to educational data. We apply BMD and BMDL estimation techniques to students’ placement exam scores in order to calculate the minimum placement exam score required to place in a class and successfully pass it. Taken together, the three projects presented in this dissertation illustrate advances in quantitative benchmark risk assessment in environmental health and educational settings. Statistical Advisor: Dr. Walter W. Piegorsch