Data Visualization Workshop
Professor Forrest Young, Psychometric Laboratory at the University
of North Carolina, Chapel Hill, will present a one-day workshop on Friday, February 3, 1995 in the Administrative Studies Building, Room 031 at York University, Toronto, ONT, describing...
The workshop will consist of two sessions: 9am-12 noon and 1pm-4pm. (If your WWW browser supports images and imagemaps, click here for a York Campus map).
9am-12: Visual Methods for Statistical Data Analysis
ViSta, a new statistical data visualization system, will be discussed
and demonstrated in a non-technical, intuitive and entirely visual
presentation. The goal of the system is to provide the statistical
data analyst with visual methods that improve the quality, accuracy and
satisfaction of the data analysis process. The system works under
Microsoft Windows, on Macintosh computers, and under XWindows.
The
system includes visualization methods which fall into five major areas:
- Visual Guidance (GuideMaps):
To guide novice data analysts, the
steps which expert data analysts think should be taken to complete a
particular data analysis are presented as a map. The map is dynamic,
and through computer/analyst interactions, guides the analyst through
the analysis.
- Session Visualization (WorkMaps):
To aid analysts at all levels of
expertise, a map is constructed of the steps taken during the data
analysis session. This map acts as a memory aid, and can be used to
review earlier steps or commence new analyses.
- Structure Visualization (SpreadPlots):
To aid hypothesis formation
and confirmation, pictures are presented showing the structure of data
and the structure and fit of data analysis results.
- Model Re-Vision (Interactive Graphical Modeling):
To further aid
hypothesis formation and confirmation, model SpreadPlots include
graphical tools for revising parameter estimates. The consequences, in
terms of the model's structure and fit, are immediately shown to the
analyst, who can then make further re-visions if desired.
- Visual Guidance Tools (GuideTools): To guide novice data analysts,
the expert data analyst has visual tools for creating guidemaps. To
create the maps, the expert uses the system to perform prototypical
data analyses, indicating the decision steps as well as the analysis
steps.
1pm-4pm: The Implementation of ViSta in LispStat
ViSta, discussed in the morning session, is implemented using
the LispStat object-oriented environment for statistical computing and
dynamic graphics. A brief introduction to LispStat is presented, with
emphasis on its object-oriented and dynamic graphics features. Then,
ViSta's object-oriented system architecture is presented, covering its
data, model, map, icon, menu and tool objects and the underlying
ViSta/Lisp-Stat data analysis language.
This presentation culminates
in showing how a programmer develops new model objects within ViSta's
open programming environment. The relation of ViSta's architecture to
hypertext, hypercode, dataflow languages, visual programming and
program visualization will also be discussed. Finally, a cognitive
theory about data analysis will be briefly mentioned.
About Forrest Young
Forrest W. Young, Professor of Psychometrics and Biostatistics at the
University of North Carolina at Chapel Hill, received his PhD in
Psychometrics from the University of Southern California in 1967. His early
research interests focused on Multidimensional and Optimal Scaling (for
which he was elected the President of the Psychometric Society and received
the American Market Research Association's O'Dell award, both in 1981). His
research interests then shifted to Statistical Visualization, including the
development of Guided Tours and SpreadPlots.
In recent years Prof. Young has worked on visualization techniques for
guiding novice data analysts through complex data analyses. This work is
based on cognitive-science and artificial-intelligence notions concerning
the structuring of the environment in which data analysis occurs. Prof.
Young's long-range goal is to develop a public-domain,
academically-oriented, extensible visual statistics package that will be
useful for teaching introductory statistics, for data analysis by
statistically naive researchers, and for advanced research and development
in visual statistics.
Obtaining Vista
ViSta is available from Forrest Young's WWW site, at
http://forrest.psych.unc.edu/research/ViSta.html
- PostScript documentation
- Macintosh: Download the documentation as a compressed
self-extracting-archive (it's 4 Meg).
Or, you can ftp vistadoc.sea.hqx from www.psych.unc.edu/pub/forrest/vista/mac
- Windows/Unix:Download the documentation as a
zipped file (it's 3 Meg).
Or, you can ftp vistadoc.zip from www.psych.unc.edu/pub/forrest/vista/mswindows
- Documentation on Xlisp-Stat.
Go up to the SCS home page