(Mathematica graphic of Minard's depiction of the fate of Napoleon's army, from Shaw & Tigg, Applied Mathematica.)

Data Visualization Discussion Group

The Statistical Consulting Service has organizied a discussion group on Topics in Data Visualization this year. We will meet monthly and include a mix of roundtable discussions, videos from the ASA Statistical Graphics Video Library, and presentations by inside and outside speakers on topics related to data display including (but not limited to):

Meetings take place on Monday mornings, 10:30 - 12:00 in N627 Ross, except where otherwise noted. (If your WWW browser supports images and imagemaps, click here for a York Campus map).

Current Meeting

Speaker: Ian Spence, University of Totonto

Date: Monday, April 10, 10:30 - 12
Location: N627 Ross.
Title: Graphs and psychophysics.

Meeting Schedule

Oct 17
Topic: Cleveland, Visualizing Data, Ch1-3: Univariate/Bivariate data
Nov 14
Topic: Cleveland, Visualizing Data, Ch4-6: Trivariate/Hypervariate & Multiway data
Dec 12
Topic: Perceptual issues in use of colour in data graphics.

Readings: "Human factors for color display systenms: Concepts, Methods and Research" (L.D. Silverman) "Investigating the Effects of Color" (E. Hoadley) "Color sequences for univariate maps (C. Ware)

Jan 09
Speaker: Michael Friendly

Topic: Dynamic models for Graphs of Quantitative and Categorical Data

Abstract: Visual representation of data depends fundamentally on an appropriate visual scheme for mapping numbers into graphic patterns. We are used to using position along a scale for representing quantitative data, but a corresponding visual mapping for categorical data is not widely used.

I'll describe mechanical, dynamic models for both quantitative and frequency data which give rise to natural visual respresentations of data. For quantitative data, position along a scale can be related to mechanical models in which fitting data by least squares or least absolute deviations correspond directly to balancing forces or minimizing potential energy.

An analogous conceptual model for categorical data likens observations to gas molecules in a pressure chamber. In this physical model frequency corresponds to pressure and fitting a statistical model by maximum likelihood corresponds to minimizing energy or balancing of forces. The model provides neat explanations of many results for categorical data, extends readily to mway tables, and provides a rationale for the graphic representation tables, and provides a rationale for the graphic representation of of counts by area or visual density.

Feb 3 (Friday, 9-12, 1-4)
Speaker: Forrest Young

Topic:ViSta: The Visual Statistics System

Mar 17 (Friday, 1:30-3pm)
Speaker: Wayne Oldford, University of Waterloo

Location: Room 291 Behavioural Science Bldg.
Title: Object-oriented graphics for quantitative programming environments: Design and implementation in the Quail system.

Abstract: [This is joint work with Catherine Hurley of the George Washington University]

In this talk, I will give an overview and demonstration of the Views visual system of Quail. Besides being object-oriented, the design of Views differs from that of other systems in many ways.

A fundamental design principle is that any meaningful graphic element is really a visual representation of some underlying data structure. The data structure might be a dataset, a single case, a batch, a mathematical function, etc. As its visual representation, the view is a conduit to the data structure and interaction with the display should permit this.

A second principle is that a single view can be displayed in multiple locations. Wherever it appears, in however many windows, it looks the same; changing it in one place changes it everywhere else.

A third is that all but the simplest views are hierarchical in that they contain subviews. The principal responsibility of such views is the layout of their subviews. Drawing etc, is the responsibility of each subview; their location is the responsibility of the containing view. Such views layout their subviews to display relations between between them --- a scatterplot being a case in point. Subviews of views can in turn have subviews and so on to arbitrary depth.

Spatial layout is can also be used to demonstrate relations between two variables mediated by a third. General layouts are themselves views and so can be subviews of other views. In this way, arbitrary interfaces can be made.

Apr 10
Speaker: Ian Spence, University of Toronto

Title: Graphs and Psychophysics

Abstract: Virtually all graphs employ physical objects to represent numbers. For example, bars are used to represent the GDP of a country in successive years; pie slices indicate the market share held by various automobile manufacturers; and the height of a line above an axis shows the price of a stock, and how it is changing, minute-by-minute. If such representations are to be successful the physical entity used to represent a number and the psychological sensation of size elicited by that entity should be proportional.

This linear relationship does not always hold and it is interesting to examine the ways in which bias can be introduced. In some situations the pattern of errors can be quite complicated and it appears that observers establish category boundaries in an attempt to improve accuracy, but this increase in precision is purchased at the expense of bias.

Psychophysical experimentation with graphical elements can provide interesting information for the study of graphical perception and can also illuminate basic processes in perception and cognition.

Workshop: Data Visualization in

Forrest Young, from the University of North Carolina at Chapel Hill will present a one-day workshop on the ideas and implementation of ViSta, an extensible system for research and development of visual statistics and data visualization techniques.

The workshop will be held Friday, Feb 3, 1995 in Administrative Studies, Rm 031. For further information, see the Abstract .

Other Data Visualization Links

Up Go up to the SCS home page