Preference data analysis using a paired comparison model

Ranking procedures are very suitable for measuring consumer preference. This paper deals with the segmentation of a consumer panel after a ranking trial. A latent class approach is used to estimate segments from the preference data. The statistical analysis is based on the Bradley-Terry-Luce model, classically used to interpret paired comparison data. This segmentation procedure provides configurations of the stimuli in the different classes and the weights of these segments. A likelihood ratio test can be performed in order to determine the adequate number of latent classes. An application of this segmentation model on preference data is presented. (C) 1997 Elsevier Science Ltd. All rights reserved.