All active development and support have shifted to ggobi. In contrast to xgobi, ggobi has richer color support, multiple plotting windows, better portability to Microsoft Windows and MacOSX, support for xml and csv file formats. ggobi has a multidimensional scaling plugin called ggvis, which supersedes xgvis, described below.


XGobi
A SYSTEM FOR MULTIVARIATE DATA VISUALIZATION

  • XGobi is a data visualization system for viewing high-dimensional data.
    The authors are Copyright of most code belongs to Telcordia (formerly Bellcore), where much of the work was done.

  • XGobi is FREELY AVAILABLE from this page, see below.

  • XGobi's primary views are SCATTERPLOTS and LINE DRAWINGS whose points and lines can be brushed and identified across LINKED VIEWS.

  • XGobi's auxiliary views include a PARALLEL COORDINATE window and a case list window with text labels, both linked to the primary scatterplot window.

  • XGobi has LINKED SCATTERPLOT MATRICES. See Di Cook's web page on this topic.

  • XGobi features INTERACTIVE DYNAMIC GRAPHICS: realtime zoom and pan, 3-D data rotations, grand tours, correlation tours, projection pursuit, and more.

  • XGobi can be used as a simple HIGH-DIMENSIONAL DRAWING PROGRAM: Data points in p-dimensional space can be moved around manually, and lines can be drawn to connect points.

  • XGobi can handle MISSING VALUES !

  • XGobi can be run on a PC running Windows -- with a little effort.

  • XGobi supports REMOTE PROCEDURE CALLS (RPCs) as a means for interprocess communication (IPC).

For more details, see the man page and this paper.



XGvis
A SYSTEM FOR MULTIDIMENSIONAL SCALING AND GRAPH LAYOUT IN ANY DIMENSION

  • XGvis is an interactive visualization system for proximity data as well as for graphs and networks.
    The authors are: Copyright of much of the code belongs to Telcordia (formerly Bellcore), where the work was begun.

  • XGvis is FREELY AVAILABLE.

  • XGvis uses XGobi as its VISUALIZATION ENGINE.

  • XGvis implements MULTIDIMENSIONAL SCALING (MDS) to analyze
    • dissimilarity data (to produce spatial configurations/maps of objects),
    • multivariate data (for dimension reduction), and
    • discrete graphs (for graph layout).

  • Some XGvis FEATURES:
    • classical inner-product scaling (Torgerson-Gower),
    • distance scaling (Kruskal-Shepard),
    • nonmetric MDS with mixing of isotonic and identity transforms,
    • metric MDS with power transformations,
    • animation of MDS optimization,
    • restarts from random configurations and random perturbations,
    • configurations in any dimension,
    • weights as a power function of the dissimilarities,
    • differential weights between/within groups of objects,
    • within/between/anchored MDS with regard to groups defined by colors and glyphs; special cases are multidimensional unfolding and external unfolding,
    • lower and upper trimming of dissimilarities,
    • random removal of dissimilarities for stability checks,
    • missing dissimilarity handling,
    • moving of configuration points with mouse dragging,
    • viewing of configurations with 3D rotations and grand tours,
    • linked views of covariates,
    • saving and printing of configurations,
    • XGobi window for Shepard diagram of fit ...
For more details, see the man page and this paper.



Downloading and Installing XGobi and XGvis:

[To the members of the DIMACS MDS working group: This is the newest version of XGobi/XGvis shown at the meeting. Use the extensive help pages of XGvis to get started.]

XGobi and XGvis (April 2002 version) are bundled in a single shar or tar file:

To install xgobi and xgvis, do the following: unpack `xgobi.sh' or 'xgobi.tar', creating a subdirectory `xgobi'. Go to `xgobi/src', create `Makefile' automatically from `Imakefile' by executing `xmkmf', compile by executing `make', put the executable in your path, and finally run by typing `xgobi filename' or `xgvis filename'. For example:

     tar zxf xgobi.tar.gz   # if you downloaded the gzipped tar file
  or
     gunzip xgobi.sh.gz     # if you downloaded the gzipped shar file
     /bin/sh xgobi.sh       # unpack the shar file
  then
     cd xgobi/src           # the source directory 
     xmkmf                  # create `Makefile' from `Imakefile'
     make                   # compile xgobi and xgvis

     # make the xgobi and xgvis executables available, 
     # eg, by adding this directory to your path:
     set path = ($path `pwd`)      # csh, tcsh, ... users
     export PATH=$PATH:`pwd`       # ksh, bash, ... users
     # This will work for the current session; for later sessions,
     # add the appropriate line to your startup file, replacing
     # "`pwd`" with the full pathname of the `src' directory.

     # finally run first examples: 
     xgobi ../data_xgobi/places      # the places rated data
     xgvis ../data_xgvis/morsecodes  # the morsecode data
If something doesn't work, read `xgobi/Readme.install'.



Link between XGobi and ArcView

The link connects the geographic information system ArcView with XGobi to facilitate exploratory analysis of spatially-referenced data. The link provides connections between the geography and:

  • multiple attribute plots
  • variogram clouds
  • spatial CDFs
  • multivariate variogram clouds
  • lagged scatterplots
The additional files and some references can be downloaded from this page.



References:
  • ``XGvis: Interactive Data Visualization with Multidimensional Scaling''
    [gzipped PostScript, 254 Kb] [PDF, 813 Kb]
    A. Buja, D. F. Swayne, M. Littman, N. Dean, H. Hofmann (2001).
    This is the most up to date work on XGvis and covers its 1998 and 2001 redesigns and extensions.
    [Tentatively accepted for publication in the Journal of Computational and Graphical Statistics.]

  • ``Visualization Methodology for Multidimensional Scaling''
    [gzipped PostScript, 197 Kb] [PDF, 592 Kb]
    A. Buja, D. F. Swayne (2001).
    A companion paper to the above. It shows what how the XGvis functionality can be used for specific MDS problems.

  • ``XGobi: Interactive Dynamic Data Visualization in the X Window System,''
    [gzipped PostScript, 95 Kb] [PDF, 253 Kb]
    D. F. Swayne, D. Cook, A. Buja (1998).
    Journal of Computational and Graphical Statistics 7 (1) (1998).
    This is the most up to date overview of XGobi and is the closest thing to a manual.
    It includes the 1996-98 redesign of the human interface, and new features such as parallel coordinate displays, jitter plots for categorical variables, missing data support, dragging points, and more.

  • ``Missing Data in Interactive High-Dimensional Data Visualization,''
    D. F. Swayne, A. Buja (1998).
    Computational Statistics 13 (1) (1998).
    A similar paper with a different data example will appear in the Proceedings of the 1997 American Statistical Association Meetings (1998).
    This is about recently implemented methods in XGobi for the exploration of data with missing values.

  • ``Manual Controls For High-Dimensional Data Projections,''
    D. Cook, A. Buja,
    Journal of Computational and Graphical Statistics 6 (4) (1997).
    A paper on methods for manual control of 2-D projections from high-dimensional data spaces.

  • ``Interactive High-Dimensional Data Visualization,''
    [gzipped PostScript, 60 Kb] [PDF, 206 Kb]
    A. Buja, D. Cook, D. F. Swayne,
    Journal of Computational and Graphical Statistics 5(1) 78-99 (1996).
    [The above files have no figures; click here to see the figure gallery.]
    This paper contains a taxonomy of interactive data visualization based on the notions of focusing, linking, and arranging views of data. A few case studies of XGobi applications illustrate the taxonomy.

  • ``Interactive Graphical Methods in the Analysis of Customer Panel Data,'' with discussion,
    M. Koschat, D. F. Swayne,
    Journal of Business and Economics Statistics 14(1) 113-132 (1996).

  • ``Dynamic graphics in a GIS: a link between ARC/INFO and XGobi,''
    J. Symanzik, J. Majure, D. Cook, and N. Cressie,
    Computing Science and Statistics: Proc. of the 26th Symp. on the Interface, 431-535 (1994).

  • ``Higher Dimensional Representations of Graphs,''
    A. Buja, N. Dean, M. Littman and D. Swayne,
    Technical Report 95-47, DIMACS, Piscataway, NJ (1995).
    A paper on the Netpad system for graph manipulation, and XGvis/XGobi for higher dimensional graph layout.

  • ``Visualizing the Embedding of Objects in Euclidean Space,''
    M. L. Littman, D. F. Swayne, N. Dean, A. Buja,
    Computing Science and Statistics: Proc. of the 24th Symposium on the Interface, 208-217 (1992).
    A paper about the XGvis system for interactive Multidimensional Scaling (MDS). XGvis uses XGobi to view embedded configurations. Kruskal's optimization algorithm can be viewed in progress. Debby Swayne recently revived this program; it has nothing lost of its usefulness. We hope to be able to offer it to the public sometime soon.

  • ``XGobi Meets S: Integrating Software for Data Analysis,''
    D. F. Swayne, N. Hubbell, A. Buja,
    Computing Science and Statistics: Proc. of the 23rd Symposium on the Interface, 430-434 (1991).

  • ``XGobi: Interactive Dynamic Graphics in the X Window System with a Link to S,''
    D. F. Swayne, D. Cook, A. Buja,
    Proceedings of the 1991 American Statistical Association Meetings (1992).
    This is the first published document about XGobi.

  • ``Theory and Computational Methods for Dynamic Projections in High-Dimensional Data Visualization,''
    A. Buja, D. Cook, D. Asimov, C. B. Hurley.
    A monograph on rendering techniques, mathematics, and algorithms for tours and related methods.

  • ``Prosection Views: Dimensional Inference through Sections and Projections,'' with discussion, (paper, figures, rejoinder)
    G. W. Furnas, A. Buja,
    Journal of Computational and Graphical Statistics 3, 323-385 (1994).
    Explores the question of what can be inferred about high-dimensional objects when looking at their projections and sections. An interesting duality between projections and sections is described.

  • ``The Grand Tour via Geodesic Interpolation of 2-Frames,''
    D. Asimov, A. Buja,
    Visual Data Exploration and Analysis, Symposium on Electronic Imaging Science and Technology (IS&T/SPIE) (1994).
    An early technical paper on grand tour construction. Superseded by the above monograph on "Theory and Computational Methods ...".

  • ``Grand Tour and Projection Pursuit,''
    D. Cook, A. Buja, J. Cabrera,
    Journal of Computational and Graphical Statistics 4(3) 155-172 (1995).
    One of two papers from Di's thesis.

  • ``Projection Pursuit Indices based on Orthogonal Function Expansions,''
    D. Cook, A. Buja, J. Cabrera,
    Journal of Computational and Graphical Statistics 2 3 225-250 (1993).
    Di's other thesis paper. Expands on work by J.H. Friedman and P. Hall on exploratory projection pursuit based on expansions.

  • ``Analyzing High-Dimensional Data with Motion Graphics,''
    C. B. Hurley, A. Buja,
    SIAM Journal on Scientific and Statistical Computing 11, 1193-1211, (1990).
    Describes a marriage of mulitvariate analysis with motion graphics. Unfortunately the functionality shown here is not currently available in any software system. It would be desirable to have it in xgobi or now ggobi.

  • ``Elements of a Viewing Pipeline for a Data Analysis,''
    A. Buja, D. Asimov, C. B. Hurley, J. A. McDonald,
    in: Dynamic Graphics for Statistics, eds.: W. S. Cleveland, M. E. McGill; Wadsworth Statistics/Probability Series (1988).
    Introduces important pipeline concepts for the construction of complex data visualization systems. Based on a Lisp-based predecessor system of XGobi.

  • ``Grand Tour Methods: An Outline,''
    A. Buja, D. Asimov,
    17th Symposium on the Interface of Computer Science and Statistics 63-67 (1986)
    The earliest paper that describes grand tours as currently implemented in XGobi, and now in GGobi.

  • ``Interactive Data Visualization using Focusing and Linking,''
    A. Buja, J. A. McDonald, J. Michalak, W. Stuetzle,
    Visualization '91, 156-163 (1991).
    Still of interest: an early paper on two important data visualization concepts.

  • ``Painting multiple views of complex objects,''
    J. A. McDonald, W. Stuetzle, A. Buja,
    OOPSLA/ECOOP '90 Proceedings, 245-257 (1990).
    A paper related to the one above. More oriented towards systems aspects.