Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Identification of clusters and underlying latent components in sensory analysis.
Evelyne Vigneau

Last modified: 2013-06-14

Abstract


The Clustering of Variables around Latent Variables (CLV) approach aims to identify groups of features in a data set, and, at the same time, the prototype, or the latent variable, of each group. The procedure makes it possible to search for local groups or directional groups. Moreover, constraints may be added on the latent variables in order to introduce, if available, additional information on the observations or/and the variables. This approach is illustrated in two different contexts encountered in sensory analysis: (i) The clustering of sensory descriptors by taking into account their redundancy but also the type of the sensory perception investigated (texture, flavour,…); (ii) the segmentation of a panel of consumers according to their liking, by taking account of external information on the products and on the consumers.

Full Text: PDF