Desktop Survival Guide
by Graham Williams
GGobi is a powerful, free and open source tool for visualising data, supporting two of the most useful interactive visualisation concepts, known as brushing and tours. GGobi is not R software as such, but is integrated with R through the rggobi package and the ggobi function. Key uses in a data mining context include the exploration of the distribution of observations for multiple variables, visualisations of missing values, exploration for the development of classification models, and cluster analysis ().
To use GGobi from the Interactive option of the Explore tab the GGobi application will need to be installed. GGobi runs under GNU/Linux, OS/X, and MS/Windows and is available for download from http://www.ggobi.org/.
GGobi is very powerful indeed, and here we only cover some basic functionality. With GGobi we are able to explore high-dimensional data through highly dynamic and interactive graphics that include tours, scatter plots, bar plots and parallel coordinates plots. The plots are interactive and linked with brushing and identification. Panning and zooming is supported. Data can be rotated in 3D, and we can tour high dimensional data through 1D, 2D, and 2x1D projections, with manual and automatic control of projection pursuits.
We are also able to interact with GGobi by issuing commands through the R Console, and thus we can script some standard visualisations from R using GGobi. For example, patterns found in data using R or Rattle can be automatically passed to GGobi for interactive exploration. Whilst interacting with GGobi plots we can also highlight points and have them communicated back to R for further analysis.