RAND Statistics Seminar Series

Visual Exploration of Graph Data in GGobi

Presented by Deborah Swayne - AT&T Labs - Research
with Duncan Temple Lang - Lucent Bell Laboratories
Di Cook - Iowa State University
Andreas Buja - The Wharton School, University of Pennsylvania
Thursday, March 18, 2004, 4:00 pm
Main Conference Room


Graphs have long been of interest in telecommunications and social network analysis, and they are now receiving increasing attention from statisticians working in other areas, particularly biostatistics.

Most of the visualization software available for working with graphs has come from outside statistics and has not included the kind of interaction that statisticians have come to expect. At the same time, most of the exploratory visualization software available to statisticians has made no provision for the special structure of graphs.

Graphics software for the exploratory visual analysis of graph data should include the following: graph layout methods; a variety of displays and methods for exploring variables on both nodes and edges, including methods that allow these covariate displays to be linked to the network view; methods for thinning a dense graph. In addition, the power of the visualization software is greater if it can be smoothly linked to an extensible and interactive statistics environment.

In this talk, I'll describe and demonstrate how these goals have been addressed in the GGobi data visualization software. GGobi is a descendant of XGobi, with many additions and improvements, including multiple plotting windows, flexible color management, and XML file handling.

GGobi has been designed so that it can be extended in two ways. First, it can be embedded in other software and controlled using an API (application programming interface); one of its embedding environments is R. Second, it can be extended using plugins. The plugins can be used to handle different data formats, and to add algorithms, graphical user interface controls, or plot types, and they can be written in C, Java, or the S language.