The Effects Of Representation And Juxtaposition On Graphical Perception Of Matrix Visualization

Analyzing multiple networks at once is a common yet difficult task in many domains. Using adjacency matrices for this purpose, however, can be effective because of its superior ability to accommodate dense networks in a small area. We evaluate various representations and juxtaposition designs for visualizing adjacency matrices through a series of controlled experiments. We investigate the effects of using square matrices and triangular matrices on the speed and accuracy of performing graphical-perception tasks. Based on human symmetric perception, we propose two alternative juxtaposition designs to the conventional side-by-side juxtaposition, and study how users perform visual search and comparison tasks regarding different juxtaposition types. Our results show that the matrix representations have similar performance, and the matrix juxtaposition types perform differently. With the design guidelines derived from our studies, we present a compact visualization termed TileMatrix for juxtaposing a large number of matrices, and demonstrate its effectiveness in analyzing multi-faceted, time-varying networks using real-world data.