Last edited on 20.Jan.97 by gs
Our apologies for the slightly disorganised nature of these notes.
There is a lot of information here, and links to other material, but it
is not well organised. That is one of the things that might happen when
one tries to gather material and publish it in real time
Design and Implementation of
Data Analysis Systems
Materials Available, and Summary. European AM, Saturday 25 March.
The opening session on Thursday evening
was a presentation by Piet Groeneboom (TU Delft)
(P.Groeneboom@twi.tudelft.nl) and his
colleague Dimitri Tischenko
We have the materials for Alan Wilk's presentation on
The future of S
Robert Gentleman and Ross Ihaka gave a presentation about R.
R (an alternative to S) written by Ross Ihaka and
Robert Gentleman of the University of Aukland. R is implemented
via an underlying Scheme interpreter that is itself written in C.
One of the reasons to implement R is to have an S-like environment
that can work on small Macintoshes. The current version is used for
teaching introductory courses on Macs with less than 4mb of memory.
Ihaka and Gentleman's R software is available at the University of
Aukland via FTP
Interesting discussion about the implementation, and particular the
modes of evaluation (lazy vs. eager) of S and R, with particular
comments by Luke Tierney.
M Theus of Augsburg started with a brief basic introduction to DataDesk,
showing graphics capabilities and drag and drop, including things like
dragging one plot into another to ensure that the two plots are the
same size (and the same scale).
Luke Tierney gave a presentation about XLispStat.
Luke started with a simple explanation of (X)LispStat,
including some of the modelling tools that are available.
Graphics is the strength of the system, particularly dynamic graphics
(like spinning point clouds) and linking of graphs for data
exploration. He showed a kernel density estimate with a scroll bar to
modify the smoothing parameter. An interesting example was a kernel
density estimate that was done by continually updated by bootstrapping
from the data and redrawing the display on the fly. Many of the
"simple" extensions that Luke described were extremely powerful. A
theme of the presentation was that the system provides tools that
allow anyone to extend the graphics and particular commands.
There was some discussion about Luke's Grand Tour implementations,
which just reinforced his point that the open nature of the system
allowed users to try different versions of many plots, including the
Where is the package going.
The software is available from the
StatLib server and from Luke's FTP
- The underlying Lisp. Has received considerable work and is
"basically done", including garbage collection and compilation (via a
- Graphics. Improvement of use of the native GUI, portability,
widget, merging of dialogs and plots, heirarchical plots.
- Objects. extensions to multiple dispatching, meta-object protocols,
tail recursion, ditributed objects (like OLE and CORBA), and
- Visual Interface. Graphical support for visual syntax.
Frank Weitek gave a talk about a graphic editor for epidemiological models
Bill Cleveland and Rick Becker gave a tag-team presentation
on Trellis Displays.
Allan Wilks gave an interesting presentation on Pictor, the next
generation of graphics for S. Simple notes on this talk are
MANET talk and information about
Radical Effective Graphical
Analysis of Regional Data are available
D. Keim (München) will talk on
data mining (Note: big files, w. colour.)
Rick Becker and Allan Wilks presented some material on maps (in S).
Technical Report is available on netlib. It is a compressed postscript
Our faithful conference organiser, Günther Sawitzki (StatLab Heidelberg)
will speak on On a Voyage to Oberon, Part II