Title: Canonical covariance analysis for three-mode three-way data by using connector matrix
Authors: Jun Tsuchida - Doshisha University (Japan) [presenting]
Hiroshi Yadohisa - Doshisha University (Japan)
Abstract: Given two three-mode three-way data sets, such as panel data sets, two types of factors are generally investigated: common factors, which show relationships between the two data sets, and unique factors, which represent the uniqueness of each data set. We propose a method for investigating the common and unique factors simultaneously. Canonical covariance analysis is one such method; however, this method has been proposed for two-mode two-way data and regards the same variable under different conditions as being two different variables. Moreover, this method makes it difficult to distinguish between the two types of factors because parameters of common factors are not separated from parameters of unique factors. To address these problems, we introduce a connector matrix and extend the canonical covariance analysis from two-mode two-way to three-mode three-way data sets. Using a connector matrix makes it easy to distinguish between the two types of factors. Furthermore, we can choose different numbers of dimensions for the canonical variable between two data sets.