Triangle tests. Number of discriminators estimated by Bayes’ rule

A method based on Bayes’ rule was used to estimate the number of discriminators in triangle tests performed by consumers. It is assumed that the total population of consumers is constituted by two groups of individuals, fully discriminators and non-discriminators, and that random numbers of persons from each group form the test panel. Individuals are exchangeable within groups but not between groups and only the responses of non-discriminators follow a binomial distribution with p = 1/3. Given an observed result from a triangle test, it is used to estimate the most likely number of discriminators and non-discriminators and their confidence limits.