Similarity testing in sensory and consumer research

Similarity testing is an important methodology in sensory and consumer research. It is of practical and theoretical significance. In many comparative experiments in sensory and consumer field, the objective is not to demonstrate difference but to demonstrate equivalence or similarity between treatments. It is widely acknowledged that the conventional hypothesis test with the null hypothesis of “no difference” between two treatments is appropriate for difference testing but inappropriate for similarity testing. Schuirmann [Biometrics 37 (1981) 617] introduced the use of interval hypotheses testing in the context of bioequivalence. For the interval hypotheses testing, the null hypothesis is that the difference between two treatments is larger than a specified non-zero value defining equivalence or similarity and the alternative hypothesis is that the difference is smaller than the specified value. If the null hypothesis is rejected, equivalence or similarity is then concluded. Based on this concept and approach, the present paper discusses some statistical procedures of similarity testing used for different situations in sensory and consumer research. The procedures include similarity testing for hedonic or intensity rating means, similarity testing using conventional discrimination methods (forced choice methods and the methods with response bias) and similarity testing for preference proportions.