A B ayesian Approach to Analyzing Replicated Preference Tests

Replicated preference tests have become important tools for assessing consistency in consumer preferences in repeated tests as well as overall consumer preference. Replicated preference tests can also provide the means for separating discriminators from nondiscriminators. Despite the increasing popularity of replicated preference tests, there are relatively few statistical tools for their analysis, especially when one is interested in examining the consistency of consumer preferences across tests. This paper presents flexible B ayesian methods for examining overall consumer preference and consistency of consumer preference in replicated preference tests. In particular, this paper presents B ayesian methods for forced‐choice preference testing with two tests and then extends this methodology to include forced‐choice preference testing with more than two tests and replicated preference testing with a no‐preference or no‐choice option. The methods produce intuitive and easily interpreted probabilities. These methods are applied to various replicated preferences test data from the literature. Practical Applications The B ayesian methods presented in this paper will help sensory scientists and statisticians working with sensory data to explore data from replicated preference tests more thoroughly. Much of the currently used methodology only allows scientists to assess the overall preference for a particular product, although a measure of overdispersion is sometimes incorporated. The methodology in this paper allows for a wider array of questions to be answered. In particular, this paper focuses on consumers' ability to consistently choose the same product. Initiallly, the methods apply to forced‐choice tests, but they are later extended to include a no‐preference option. Allowing for a no‐preference options is another important contribution of this work.