Author ORCID Identifier
https://orcid.org/0009-0001-2657-610X
Date Available
5-3-2025
Year of Publication
2024
Document Type
Doctoral Dissertation
Degree Name
Doctor of Philosophy (PhD)
College
Arts and Sciences
Department/School/Program
Statistics
Advisor
Dr. Katherine L. Thompson
Co-Director of Graduate Studies
Dr. Arnold J. Stromberg
Abstract
Median effective dose (ED50), benchmark doses (BMDs), or median lethal dose (LD50) are a key concept in toxicology and medicine. Traditional dose-response analyses often neglect other relevant factors beyond dose considerations, inducing the need to explore the inclusion of additional explanatory variables.
This dissertation evaluates established computation methods, including the profile likelihood method, the Wald interval, and a bootstrap percentile method with and without additional explanatory variables through a comprehensive simulation study involving GLMs with different link functions. It also investigates the behavior of these estimators in small sample sizes.
Findings indicate that the profile likelihood method suffices for single-variable models. The bootstrap percentile intervals provide a better coverage rate of the true ED50 in scenarios involving additional explanatory variables or log-transformation of the dose variable. Moreover, the coverage rate of bootstrap intervals is influenced by group size, with larger groups yielding higher coverage rates while other factors, such as the quality of the data and the number of treatment levels (valid points) affect the width of bootstrap intervals.
Digital Object Identifier (DOI)
https://doi.org/10.13023/etd.2024.222
Recommended Citation
Park, Jiyeon, "Deriving Confidence Intervals for ED50/LD50 in Generalized Linear Models" (2024). Theses and Dissertations--Statistics. 76.
https://uknowledge.uky.edu/statistics_etds/76
