The density of points within multidimensional clusters can impact the effective representation of distances and groups when projecting data from higher dimensions onto a lower dimensional space. This paper examines the use of motion to retain an accurate representation of the point density of clusters that might otherwise be lost when a multidimensional dataset is projected into a 2D space. We investigate how users interpret motion in 2D scatterplots and whether or not they are able to effectively interpret the point density of the clusters through motion. Specifically, we consider different types of density-based motion, where the magnitude of the motion is directly related to the density of the clusters. We conducted a series of user studies with synthetic datasets to explore how motion can help users in various multidimensional data analyses, including cluster identification, similarity seeking, and cluster ranking tasks. In a first user study, we evaluated the motions in terms of task success, task completion times, and subject confidence. Our findings indicate that, for some tasks, motion outperforms the static scatterplots; circular path motions in particularly give significantly better results compared to the other motions. In a second user study, we found that users were easily able to distinguish clusters with different densities as long the magnitudes of motion were above a particular threshold. Our results indicate that it may be effective to incorporate motion into visualization systems that enable the exploration and analysis of multidimensional data.

Etemadpour, R., Forbes, A. G., Density-based motion for multidimensional data projection, Proceedings of SPIE-IS&T Electronic Imaging, Visualization and Data Analysis 2015, vol 9397, San Francisco, CA, February 11th, 2015.