Peltier, C.3; Brockhoff, Per B.1; Visalli, M.3; Schlich, P.3
1 Department of Applied Mathematics and Computer Science, Technical University of Denmark2 Statistics and Data Analysis, Department of Applied Mathematics and Computer Science, Technical University of Denmark3 Centre des Sciences du Goût et de l’Alimentation
Assessor performances in sensory analysis are usually represented by three indicators: repeatability, discrimination and agreement. However, assessors can also differ on the range of their scores, the so-called “scaling effect”. Brockhoff, Schlich, and Skovgaard (2013) proposed the mixed assessor model (MAM) which, as the original assessor model (Brockhoff & Skovgaard, 1994), takes this effect into account, but also allows for the product effect to be tested against a new interaction free of the scaling effect.The present paper proposes a unified system for monitoring assessor and panel performances based on the MAM. In addition to the product effect (tested at panel and individual levels), scaling and repeatability panel heterogeneities are tested and, when significant, assessors who were responsible for this heterogeneity are identified. Further, the pure disagreement is decomposed over assessors allowing for the individuals significantly responsible for it to be highlighted. Those numerous statistics are summarized into a few diagnostics presented in a synthetic single table called the MAM CAP table.
Food Quality and Preference, 2014, Vol 32, Issue Part A, p. 24-27