Identifying random-scoring respondents in sensory research using finite mixture regression models

Abstract This paper illustrates the application of finite mixture regression modelling to consumer sensory studies, exploring the segmentation of consumers using latent class methodology as an alternative to conventional techniques. An extended model is proposed that enables the effects of ‘random scoring’ respondents to be filtered out, with the aim of providing a clearer identification of the effects of sensory drivers on liking and interpretation of the results. The extended model is used to compare the results of two alternative preference elicitation techniques, suggesting that much of the observed scoring variation is due to the elicitation task itself rather than inability to discriminate between the products.