Interpretation, validation and segmentation of preference mapping models
Peer reviewed, Journal article
MetadataShow full item record
Original versionFood Quality and Preference. 2014, 32 198-209. 10.1016/j.foodqual.2013.10.002
In this paper we discuss an extension to preference mapping of the method proposed in [Endrizzi, I., Menichelli, E., Johansen, S. B., Olsen, N. V., & Næs, T. (2011). Handling of individual differences in rating- based conjoint analysis. Food Quality and Preference, 22, 241–254 ] for accommodating both population averages and individual differences in the same model. The method, based on average estimates and residuals, is a combination of ANOVA, PCA and PLS-DA, which are well-known techniques that can be run in almost all statistical software packages. Main attention is given to the relation between the double- centred residual matrix which highlights differences between consumers in their relative position as compared to the average consumer values and the standard centring in preference mapping. This approach has been found particularly useful for highlighting differences in preference pattern among the consumers. Furthermore, the interpretation and the segmentation, that is here taking place based on differences in acceptance pattern, are graphically oriented. In addition, some possible alternatives to the generally used validation method in PCA are suggested. The approach is then illustrated using two data-sets from consumer studies of apple and raspberry juice, showing that when individual differences are analysed by the present method, interesting results regarding individual differences in response pattern are detected.