A0461
Title: Optimal designs with multiple correlated responses for experiments with mixtures
Authors: Hsiang-Ling Hsu - National University of Kaohsiung (Taiwan) [presenting]
Abstract: A mixture experiment within the (q-1)-dimensional probability simplex is a specific experimental setup in which the q factors are non-negative and adhere to the sum of all factors equals one. The issue of the optimal approximate designs is investigated with the k-correlated response mixture experimental models. In the multiple correlated response mixture models, the improvement design class is explored, known as the complete class, in relation to the Kiefer ordering for a given design. Based on the complete class results, the properties of optimal designs are delved into for multiresponse models using the well-established equivalence theorem. For specific multiresponse model settings under the D- or A-optimal design criteria, the optimal results can reduce multiresponse experimental design problems to single-response experimental design problems. An illustrative example showcasing optimal designs for correlated response mixture experimental models is presented.
