Multivariate Visualization of Importance-Varying Data

NSF CAREER Grant:9996043



In many applications, information about the certainty of values is available or can be calculated from available information. In such situations, displaying the certainty information along with the variable values could greatly improve the usefulness of a visualization. Obviously, when visualizations are used to guide decision-making, values with lower certainty should not be given weight equal to more certain values. On the other hand, only displaying values with high certainty would be less informative, since uncertain values can provide valuable indications of broad patterns -- even when it would be unwise to draw conclusions about values at specific locations. In this way, uncertain values can provide the global context while highly certain values provide the fine detail. Multivariate visualization which incorporate certainty information can convey more available information, while giving each piece of information a visual weight corresponding to its certainty. The problem of incorporating certainty information into a visualization can be generalized to include other related data attributes. These include expected or calculated error, fuzzy classifications, high-dimensional fit, constraint satisfaction, sampling or modeling resolution, agreement among multiple measures, and relevance or interest, resulting in the category of importance-varying data.

This project investigates techniques for the display of importance-varying data in both two- and three-dimensional data domains. The approach to this problem emphasizes consideration of the perceptual characteristics of human viewers, dymanically manipulate display environments to increase the number of variables that can be clearly displayed, and experimental validation of the techniques developed. Portions of the project are conducted in collaboration with researchers in application fields such as meteorological monitoring, medical imaging, and public health decision-making. Specific project objectives include construction of a survey of existing importance visualization techniques, development of new techniques to represent importance-varying data, comparison and validation of techniques identified or developed, constructing of a dynamic environment for the exploration of importance-varying data, and development and dissemination of perceptual principles for effective multivariate display.


Publications

Educational Materials

The education plan has four basic components: the development and teaching of courses to comprise a specialty in computer graphics and visualization, the development and disemination of course materials for teaching visualization, the development and teaching of core courses in the undergraduate curriculum, and outreach to attract underrepresented groups to math, science, and engineering. Benefits of this plan include incorporation of research activities into undergraduate courses, collaboration across disciplines and institutions, innovative evaluation methods for student projects, development of freely available visualization course materials, and outreach activities to invole rural middle school girls in science activities.

Available course materials for courses in computer graphics and visualization, including syllabi, lecture notes, reading lists, examples, and assignments.

Data Visualization
Introductory Computer Graphics
Advanced Image Synthesis

Team members

© 1999