Analysis of ranked preference data using latent class models

Consumer preference analysis using rating experiments can alternatively be performed by means of ranking experiments. Ranking procedures seem particularly easy to perform by a consumer, as are paired comparisons experiments. Among a variety of probability models on rankings, the Bradley¯Terry¯Mallows model is considered. Segmentation panel approach, as well as fitting procedure in the presence of covariates, are presented. The extension of the use of the BTM model in the case of partial rankings, obtained from a balanced incomplete block design, is also proposed. The preference rankings analysis including segmentation approach, covariates incorporation and/or incomplete block design is illustrated in terms of an example involving the ranking of seven extruded snacks by a panel of 56 consumers.