|
Activity Number:
|
280
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
WNAR
|
| Abstract - #310161 |
|
Title:
|
Optimal Designs for Dose-Finding Experiments in Toxicity Studies
|
|
Author(s):
|
Weng-Kee Wong*+ and Holger Dette and Andrey Preplyshev
|
|
Companies:
|
University of California, Los Angeles and Ruhr-Universitaet Bochum and St. Petersburg State University
|
|
Address:
|
Department of Biostatistics UCLA, Los Angeles, CA, 90095,
|
|
Keywords:
|
continuous design ; design efficiency ; Elfving theorem ; local c-optimal design ; robust optimal design
|
|
Abstract:
|
We use optimal approximate design theory to construct experimental designs for toxicology studies on a prototype dose interval. The specific goals are to estimate the effective dose of fetal malformation rate, prenatal death rate and the overall toxicity using information on the number of implants, number of prenatal deaths and number of live fetuses. We assume the intra-litter correlation coefficient to depend on the dose and use Weibull models to represent probabilities of malfunction and prenatal death. Locally optimal designs are found, along with analytical conditions on the number of dose levels required for the study. We also compare our designs with popular equidistant designs and investigate sensitivities of our optimal designs to misspecification in the nominal values.
|
