Constructing a confidence region for the optimum values of factors in multiple response problems

Abstract

This thesis focuses on the construction of a confidence region for localization of the set of optimal values of factors in multiple responses problems, using as objective the loss function proposed by Young-Hyun, Kwang-Jae, & Chi-Hyuck (2005). Data were generated using the bootstrapping techniques consisting of resampling with replacement errors to generate new observations. This resampling enabled the creation of new optimal solutions. The correlation between responses was introduced in examples to simulate real life situations. This method of constructing the confidence regions was compared with the method proposed by Meza (1998). To be able to compare between the efficacity of these two methods, the surface area of the confidence region was used as comparison criteria.