A New Visualization Method for Asymmetric Proximity Data

• Activity Number: 257
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Graphics
• Author(s): Anna Quach* and Adele Cutler
• Companies: and Utah State University
• Keywords: Multidimensional Scaling ; Asymmetric Proximities ; Visualizing Asymmetric Proximities

Abstract

Multidimensional Scaling (MDS) plots are widely used to visualize symmetric proximity data. However, not all proximity matrices are symmetric. The traditional approach for asymmetric proximities would be to symmetrize the proximities by averaging before applying MDS. While this would be appropriate if the proximities were slightly asymmetric, visualizing with large asymmetries in this way would be misleading and would lose information. We propose a new model to visualize asymmetric proximities and compare it with existing visualization methods for asymmetric proximity data.