On Detection of the Unique Dimensions of Asymmetry in Proximity Data

As methods of analyzing the relationship of n objects of proximity matrix, multidimensional scaling (MDS) has been developed and applied to many data sets. However, we take care to apply them to an asymmetric proximity matrix. Okada and Imaizumi have been proposed so-called “circle (or radius–distance) model” by introducing the radius of each object to extract asymmetric parts in data. In this paper, we overview these models for a one-way two-mode asymmetric proximity matrix and propose a method for the detection of asymmetric dimensions by losing the positiveness of asymmetric dimension weights.