34 – Advances in Analysis of Categorical Data
Con�dence Intervals for the Difference Between Two Proportions for Correlated Binary Response
Krishna Saha
Central Connecticut State University
Interval estimation for the risk difference in the correlated binary data is often an important problem in many biomedical applications. For instance, in the study of cancer and leukemia group B randomzed trials (Cooper et al., 1993), it is often of interest to compare two chemotherapy treatments with respect to success rates of the patients with multiple myeloma who survived at the end of this study. In this project, the interval procedures are developed for estimating the difference between success rates of the two chemotherapy treatments in this study. An extensive simulation study i conducted for the purposes of evaluating and comparing the performance of the proposed intervals, in terms of coverage and expected lengths. An application to biomedical data is used to illustrate the proposed methods.