Reliability of the sensory analysis data of a panel of tasters

Abstract

A method, the quantum measurement technique, is described which, by means of the conceptualisation of the information underlying a set of data which are considered as manifestations of a latent variable or theoretical construct, allows one to detect which are the data in sensorial analysis that, after obtaining a measure, do not fit the formulated conception. The theoretical foundation, based on Rasch probability and item response theory (IRT), detects and quantifies by means of the misfits the data that respond to unexpected scores. A detailed analysis of their residuals aids in finding the causes of these misfits. The technique is applied to the data of a tasting panel that form part of the sensory evaluation of virgin olive oil. © 2000 Society of Chemical Industry