Linear model analysis for ordinal response in a mixture experiment

Abstract

Mixing of ingredients is commonly applied in food and beverage industry activities such as making cookies. An applicable method to get an optimal formula for mixing ingredients is a mixture experiment. The mixture experiment is an experiment in which levels are the proportions of the factors which are ingredients. The standard mixture design is developed based on linear models. However, the response for organoleptic test in food industry is ordinal scale. The goal of the research was to evaluate some mixture designs of linear model for ordinal logistic model which is more appropriate model for ordinal scale. The mixture designs were the D- optimal and I-optimal design. Based on the D-optimality criterion of ordinal model and D- efficiencies of the D-optimal design compared to other designs, the D-optimal design was the best design.