Analyzing paired-comparison data in the situation where judgment is affected by multiple factors

Abstract

Paired-comparison is a well-known and effective method for scaling stimuli. In situations where judgment is affected by multiple factors, it is necessary to take these multiple effects into consideration. However, there is no very effective method to deal with this problem, especially when the data is binary. This study is to propose a new method that expands the Bradley-Terry model to overcome this problem. The proposed method allows estimation of both the scaled value of each stimuli and the influence of each factor. This article includes an analysis of the real data of a sensory test to verify the efficacy of the proposed method. Lastly, a possible expansion of the BradleyTerry model for more general situations is discussed.