EcoSta 2019: Start Registration
View Submission - EcoSta2019
Title: Visualization of heterogeneous class specific tendencies in categorical data Authors:  Mariko Takagishi - Osaka university (Japan) [presenting]
Michel van de Velden - Erasmus University Rotterdam (Netherlands)
Abstract: In multiple correspondence analysis (MCA), both individuals and categories can be represented in a biplot that jointly depicts the relationships across categories or individuals. To enhance the interpretation of such a biplot, adding class information of individuals (e.g., gender, nationality) can be helpful. We consider interpreting class-specific tendencies in a biplot. By obtaining average points for each class, we can depict class-specific tendencies in the biplot. However, this approach only reveals tendencies of many individuals within a class. When a relatively small group in a class has a strong tendency towards a particular category not selected by the majority group in that class, this tendency would not be visible in the biplot. Such minority tendencies could still be interesting to consider, especially in order to characterize tendencies within classes. Therefore, we propose a new approach to find class-specific clusters, and depict them together with the category points. The resulting visualization allows us to identify different heterogeneous tendencies within classes in a single biplot, as well as the perceived relationships among classes.