Abstract

Binary responses are routinely observed in practice whether it is medicine, geology, defense or day to day life situations. Logistic regression methods can be used to capture the binary responses. Modeling becomes critical when there is sensitivity analysis involved, and the selection of the settings of variables depends on sequential design methodology. A total number of experimental runs is also an important factor since cost is directly related to it. In this research different experimental approaches for logistic regression modeling are investigated to improve the estimation of median quantile, to reduce the number of experimental runs as well as to improve overall modeling quality. We present the Break Separation Method which guarantees an overlap in the data such that the Maximum Likelihood Estimation may be used to estimate the model parameters. We also investigate and discuss the augmentation after the BSM.

Publication Date

5-17-2017

Document Type

Thesis

Student Type

Graduate

Degree Name

Industrial and Systems Engineering (MS)

Department, Program, or Center

Industrial and Systems Engineering (KGCOE)

Advisor

Rachel Silvestrini

Advisor/Committee Member

Brian Thorn

Campus

RIT – Main Campus

Share

COinS