Date of Award

12-1986

Degree Name

Doctor of Philosophy

Department

Mathematics

Abstract

Past literature concerning drug combination studies is reviewed. This literature has both biological and statistical origins. The justification of the logit transformation for dichotomous responses in a linear models framework is made by observing its similarities to models proposed by others. Using the logistic linear model, optimal experimental designs are derived. Methods for obtaining parameter estimates are discussed as are diagnostic techniques for logistic regression. Two new robust techniques for logistic regression are proposed and compared to a robust technique suggested in the literature via Monte Carlo simulation. Finally, estimation of the median effective dose (ED50) and the resulting confidence intervals are presented. Two new techniques for ED50 estimation that are resistant to outlying observations are presented.

Access Setting

Dissertation-Open Access

Share

COinS