Unfolding item response model using best-worst scaling

Abstract

In attitude measurement and sensory tests, the unfolding model is typically used. In this model, response probability is formulated by the distance between the person and the stimulus. In this study, we proposed an unfolding item response model using best-worst scaling (BWU model), in which a person chooses the best and worst stimulus among repeatedly presented subsets of stimuli. We also formulated an unfolding model using best scaling (BU model), and compared the accuracy of estimates between the BU and BWU models. A simulation experiment showed that the BWU model performed much better than the BU model in terms of bias and root mean square errors of estimates. With reference to Usami (2011), the proposed models were applied to actual data to measure attitudes toward tardiness. Results indicated high similarity between stimuli estimates generated with the proposed models and those of Usami (2011).