Quantitative Data Visualization in Compartmented Force-Directed Graphs using Calibrated Columns

Abstract

Networks are commonly used to represent data and relationships. However, when mapping nodes to quantitative data, it is often difficult to accurately identify both the values of individual nodes and the overall relationships among nodes. Simultaneous group detection and precise quantitative data reading are necessary for scientists interpreting critical data. Here, we hypothesize that the calibrated columns method for encoding large-range quantitative values will provide a more accurate reading of data values than the common approach of variable-area circles. We also hypothesize that the addition of subtle halos around nodes will support accurate grouping of spatially distributed nodes in a network. We have conducted a pilot study with seven critical tasks in order to understand quantitative data reading and group inferencing in networks having up to three-levels of complexity. Our results show that (1) network size has a significant effect on confidence levels in grouping tasks; (2) the grouping encodings do not have a significant effect on confidence levels, but do significantly affect accuracy in the overall task; (3) the quantitative data encodings do not have a significant effect on confidence levels, but do significantly affect accuracy when determining exact node values; and (4) the colored halos and calibrated columns encodings are particularly useful in tasks involving both precise and global perception of the network. This work has contributed to understanding effective construction of quantitative networks and their groupings and our results suggest design guidelines broadly applicable to inform visualization design in domains such as biology and the social sciences.