Convex Hull Brushing in Scatter Plots - Multi-dimensional Correlation Analysis

Abstract

Interactive Visual Analysis has been widely used for the reason that it allows users to investigate highly complex data in coordinated multiple views, showing different perspectives over data. In order to relate data, multiple techniques of brushing have been introduced. This work extends the state of the art by introducing the Convex Hull (CH) Brush, which is a new way of selecting and interpreting high dimensional data in scatter plot (SP) views. By using a combination of brushes through linked views, the CH-Brush allows the selection and clustering of values that are not typically defined by SP ranges, in spite of sharing similarities. In CHBrushing is also able to visually report the existence of correlation between variables. Furthermore, we discuss CH-Brushing sensitivity and the application of smoothness. We use synthetic data to support our rationale and clarify the intrinsic meanings of CH-Brushing in scatter plots. We also report on the first experience on using the CH-Brush in a real-world medical case.