Measuring the Influence of Alternative Treemap Visualizations on Human Judgment Tasks

Visualizations of large data sets provide insight to features of the data, improve the accuracy of mental models of the information, and locate data regions of particular interest. A visualization technique that has gained popularity is the treemap. The treemap uses a recursive algorithm to display a hierarchical data set in the form of nested rectangles of varying size, orientation, aspect ratio, relative placement, and color to represent selected aspects of the data. This dissertation extends previous work on human perception and rectangular relative area judgments to determine how these display features affect judgments that are supported by treemap visualizations. The effort includes a novel treemap generation algorithm, which utilizes lexicographical order theory to generate treemaps that group like data elements together, to support human gestalt perception capabilities and the results of an experiment comparing this alternative approach to the current, squarified algorithm that is a current practice. A final experiment explores the use of treemaps for supporting decision making in a specific application area -- review and interpretation of surgical quality data -- and further characterizes the performance of decision makers to correctly interpret data displayed in alternative treemap formats. The combined effort thus improves on current knowledge of the capacities of individuals to make relative and proportional judgments in hierarchical visualizations, with an innovative method for visualizing hierarchically structured information in treemaps.