Abstract
A procedure of redundancy analysis for mixed measurement data, which is called QRDA, was suggested with a demonstration of its validity on artifical data(Saito & Nakajima, 1996). But so far it has not been applied for analysis of real data. It is characterized by incorporating the dependence relationship of one set of variables on the other into the model, and each set can include variables measured on continuous and/or nominal scale levels. In view of the characteristics, we apply QRDA to analyze sensory inspection data of offensive odors, showing clearer findings than those obtained by customary approaches. We conclude that QRDA serves to detect the relationship of dependence of a set of psychological variables on the other of physical variables, both measured on mixed scale levels, and to determine the number of latent variates to account for the relationship to a good extent.
