Evaluating Overall Quality of Graph Visualizations Indirectly and Directly
Huang, W, Evaluating Overall Quality of Graph Visualizations Indirectly and Directly, Handbook of Human Centric Visualization, Springer, W Huang (ed), New York, USA, pp. 373-390. ISBN 978-1-4614-7484-5 (2014) [Research Book Chapter]
Visualization is one of the popular methods that are used to explore
and communicate complex non-visual data. However, representing non-visual data
in a visual form does not automatically make the process of exploration and
communication effective. The same data can be visualized in many different ways
and different visualizations affect the process differently. Therefore, it is important
to have the resultant visualizations evaluated so that their quality in conveying
the embedded information to the end users can be understood. In designing an
evaluation study, at least three issues need to be addressed: what kind of quality is
to be evaluated?What methods are to be used? And what measures are to be used?
A range of methods and measurements have been used to evaluate visualizations in
the literature. Overall quality is often considered as a multidimensional construct
and the elements of the construct have limitations in evaluating overall quality.
In this chapter, we introduce two one-dimensional measures. The first one is an
indirect measure called visualization efficiency that is based on task performance
and mental effort measures, while the second is a direct measure that is based on
aesthetic criteria. These new measures take into consideration the elements of its
corresponding multidimensional construct and combine them into a single value.
We review related work, explain how these measures work and discuss user studies
that were conducted to validate them.