The history of GGobi

Out of sight, out of mind

Where did the GGobi name came from?

The name arose with an earlier version of the software, XGobi. It was developed under the fostering of Bellcore. Bellcore was the the research lab for all the local telecommunications companies in the late 80s and early 90s. To be able to release the software, it had to be approved by the lawyers, who also insisted on a name.

Originally XGobi was called Xdataviewer, as it had evolved from an older software called Dataviewer, written in Lisp by Andreas Buja, Catherine Hurley, John McDonald and Werner Stuetzle, while at the University of Washington, Seattle. The lawyers objected to this name (Dataviewer was too generic and used by too many other people), and insisted we chose a new one. All the sensible names were rejected by the lawyers so Debby and Andreas started throwing out nonsense names, names of deserts because multi-colored rotating scatterplots, the primary purpose of the software,q resemble the swirling sands of a desert. Gobi was the desert name that we all preferred.

Why was GGobi created?

It was initially developed to look at data matrices. We commonly have data that takes a matrix form, the rows are the samples/cases/subjects, and the columns are the values/variables that we have measured on them. The dimension of a data set is the number of variables, and high dimensional has more than two variables. Each variable is mapped to an axis in a Euclidean coordinate system, and the number of axes is the dimension.

It initially arose from our own interests in finding ways to draw pictures of high-dimensional data. It has evolved to handle other types of data, including missing values and network/graph data. There were users at Bellcore, and then at AT&T interested in telecommunications networks which drove the network methods. XGobi included maps by linking to other software (developed as part of a research grant in spatial analysis), and this work is being picked up again with GGobi. We are also now looking at methods for related tables, catering to longitudinal measurements. This is driven by general interests, consulting work and teaching.

When did the project start?

Dataviewer was started in the mid-80s, XGobi started in 1988/9, and GGobi started in about 1999. Each change in the software was driven by technology changes. In the late 80s X Windows was the dominant windowing system on unix workstations. That didn't eventually port so well to Microsoft Windows, so in the late 90s Debby began looking for other graphics libraries to build on that were portable across platforms and more modern looking. We chose gtk, which gives the first "G" in GGobi.