A cluster approach to analyze preference data: Choice of the number of clusters

We consider the clustering of a panel of consumers according to their scores of liking. The procedure is based on a cluster of variables approach proposed by Vigneau et al. [Vigneau, E., Qannari, E. M., Punter, P. H., & Knoops, S. (2001). Segmentation of a panel of consumers using clustering of variables around latent directions of preference. Food Quality and Preference, 12, 259–363]. We aim at setting up a hypothesis-testing framework in order to determine the appropriate number of clusters. The procedure consists of two steps. Firstly, a cluster tendency test determines if there is more than one cluster. Secondly, a hierarchical algorithm is performed and cluster validity tests at the different levels of the hierarchy indicate the appropriate number of clusters. Once the number of clusters is determined, a partitioning algorithm is implemented by considering as a starting point the partition obtained from the hierarchical algorithm. We illustrate the method on preference data from a European sensory and consumer study on coffee [ESN (1996). A European sensory and consumer study: A case study on coffee. European Sensory Network] and we undergo a simulation study in order to assess the efficiency of the procedure.