Multiplicative mixed models for the analysis of sensory evaluation data

Multiplicative mixed models have recently been proposed for the analysis of data from series of plant variety trials. Such approaches may be useful for other applications, including the analysis of data from sensory evaluation experiments. Implicit in these data are two-way tables of product by assessor effects. Several authors have suggested the need to accommodate heterogeneity of product and error variance between assessors. This provides information on the discriminating ability and reliability of individual assessors. Additionally it is of interest to examine the level of agreement between assessors. These issues can be explored with the use of a mixed model that includes a multiplicative model for product by assessor effects. This allows examination of product by assessor effects whilst accommodating important covariate information, for example, associated with the experimental design. In this paper we use the approach to analyse data from two sensory evaluation experiments.