Fixed width confidence interval estimation for general continuous responses under a response adaptive design

Abstract

. A fixed width confidence interval for the treatment difference is obtained under heteroscedastic normal responses following a response adaptive allocation targeting Neyman optimum allocation. The developed procedure is further extended for general location-scale model based responses under a compound response adaptive allocation. The developed procedures are studied theoretically and are supported empirically assuming various response models. A real data study is also provided following the proposed procedure for the purpose of illustration.