Multivariate Difference Testing with Multiple Samples
Rousseau, B. and Ennis, D. M. (2002). IFPress, 5(2) 2-3.
Variation in the manufacture of consumer products presents a challenge when investigating the suitability of process and formulation changes to a particular product. This issue is particularly relevant when investigating small sensory differences, like those involved in discrimination testing. In some cases, a product is best represented by several variants. The traditional discrimination protocols such as the triangular, duo-trio and m-alternative forced choice methods do not allow the consideration of manufacturing variation, unless multiple paired tests are performed. This approach can be costly and time-consuming. In an earlier newsletter, we introduced Torgerson’s method of triads. Using an appropriate model, this method allows for the simultaneous estimation of multiple product differences. In this report, we discuss a multivariate model for Torgerson’s method.
Figure 3. Strawberry flavored yogurts: Representation of product similarities using the multidimensional Thurstonian model.