QPRC 2016

Exact, D-optimal Mixture Designs for Ordered Categorical Data


Mickey V. Mancenido, Ph.D.

Arizona State University 


Mixture or formulation experiments are prevalent in industries where several mixture ingredients or components are treated as the experimental factors. In many cases, formulation experiments in industrial settings yield responses that are measured in an ordinal scale. This situation often arises when chemical formulations are evaluated by a panel of experts or randomly selected participants for sensory attributes, such as taste, feel, or clarity. While research on designing experiments for continuous responses in formulation experiments is abundant in experimental design literature, this is not so the case with ordinal data. In this talk, we propose an adapted coordinate-exchange algorithm for the construction of an optimal mixture design for an ordinal response.