Organizer:

Antony Unwin, unwin@uni-augsburg.de

Speakers

10:30 a.m.
Graphical Interfaces for Statistical Software: Benefits, Drawbacks, and Future Directions
David Lane, Rice University

The overwhelming majority of modern spreadsheet, word processor, database, and other "desktop" software employ graphical user interfaces. The popularity of these interfaces has been justified by research showing that users are more productive using graphical than using command-based interfaces. Graphical interfaces are not as dominant in statistical software as they are in these other software categories. This is due, in part, to the difficulty of designing a good graphical interface for some of the tasks done with statistical software. Nonetheless, there are several excellent statistical analysis packages with graphical interfaces. This paper considers which tasks and user populations are best served by graphical interfaces and which are best served by command-based interfaces. Methods by which future graphical interfaces might accomplish tasks now best-done with command-based interfaces are discussed.

11:00 a.m.
GUI and Command-line - Conflict or Synergy?
Antony Unwin and Heike Hofmann, University of Augsburg, Germany

Interfaces rely on mutual understanding. Graphical interfaces tend to aid understanding and be easier to pick up again after a break. They are good for direct interaction without exact specification. Command-line interfaces are for specialists with deep understanding and offer powerful, flexible structures. They make it easy to store and repeat precise sequences. Ideally, we would like to have both forms of interface. To make progress we will have to understand better what statisticians and data analysts need and what the strengths (and weaknesses) of the two conflicting approaches are.

11:30 a.m.
Mental Models and Interactive Statistics
Wayne Oldford, University of Waterloo
Keywords: direct manipulation interfaces, analysis environments, Quail

The debate about the appropriate computer human interface is an old one and not likely to end anytime soon. Two things seem certain. First that the user needs to be presented with clear and consistent mental models for interacting with the machine. Second that all computer use is actually computer programming and so all but the most elementary interactions quickly demand all the usual features of a programming language.

Drawing on examples from Quail, this talk explores some of the issues involved in direct manipulation interfaces. Some principles and pitfalls are identified.