ANGULAR TREEMAPS - Information Visualisation: Techniques, Usability and Evaluation - page 104

During the usability study, subjects were given static images of both
methods. Images were sized at approximately 1000×800 pixels. Images
were presented on a 24-inch wide monitor with 1920×1200 pixel
resolution. Subjects were seated approximately 80 cm from the monitor.
With the angular treemap method, the target applied angular rotation; in
the other method, the vertical and horizontal treemap technique was
applied. The order of methods varied randomly from subject to subject, to
prevent order bias. In each trial, subjects loaded a new image by pressing a
key and were asked to write down answers as soon as they were found.
Between each trial a blank background was displayed. Then, the user
reported where the targets were located to the experimenter, who noted it.
Subjects were asked to complete seven questions for each task, in
sequence of increasing difficulty. Task completion time was measured.
The reading time and writing time was not counted in completion time.
The average completion time for each task was measured. At the end,
subjects were interviewed as to preference and were also encouraged to
give feedback about further improvements.
Overall, subjects' performances were consistent. In a single case
subject behaviour deviated significantly from the average. This case has
also been included in the result analysis. Consequently, in all three tasks
subject performances for the angular method were better than for
traditional treemaps. Fig. 5.16 shows average completion time of the three
tasks for both methods. Compared with traditional treemaps, the time
spent using angular treemaps to locate one node on level 1 was 17% less
than using traditional treemaps; users efficiency in using the Angular
Treemap application for locating a particular node on level 2 increased by
19%.
The advantage of Angular Treemaps is greatly reflected in task 3.
When users need to identify two different focuses on different levels of the
hierarchy, the performance using Angular Treemaps improved by 29%.
Last but not least, all users selected Angular Treemaps as their preference
in the interview session.
Conclusion
Current treemap techniques provide the capability to visualize large
attributed hierarchical data inside a limited display space. However, most
algorithms have a common limitation: they can only partition a rectangle
into sub-rectangles in vertical-and-horizontal directions. This restricts not
only human perceptual skills for recognizing visual objects but also the
adaptability to apply treemaps to various styles and display areas, such as