Session Topics

General Software Design

Simulation Environments for Economic Modeling
Friedrich Leisch, Andreas Weingessel

We are part of the center of excellence (short in Austrian: SFB) ``Adaptive Modelling in Economics and Management Science''. One of the long-term goals of the SFB is to create a model for an artificial economy, where adaptive models can be tested. Our group is currently working on a marketing simulation (using R), another group has a virtual stock market (using matlab) etc. The current plan is that some of these components will be connected in the future, e.g., to connect a marketing simulation with a production simulation. For this artificial economy to work we will need a lot of distributed computing, because the involved groups use rather heterogenous platforms/software. Another advantage of distributed sytems is that they are much closer to reality than a big homogenous simulation.
Download postscript file of slides.

Redesign of XLispStat
Luke Tierney
An update on the redesign of xlispstat, which is based on designing a core that can be implemented on top of C or Java and can be used as the base for xlispstat or R or Omega.
Download postscript file of slides.

Design of functions and objects
Jim Lindsey
Here are some ideas about topics I could discuss:
1. Developing functions for nonlinear, non-normal, multivariate models.
2. Designing objects for handling unbalanced repeated measurements allowing for various types of response variables (continuous, categorical, event history).
Download postscript files of slides: part 1, part 2,

Spreadsheet interfaces for economic applications
Erich Neuwirth
The only ``statistical'' packages economic students or many people from indusrty know are spreadsheets, hence having good interactions between spreadsheets and ``real'' statistics software would be beneficient both in teaching statistics as in many consulting jobs.

Distributed Computing / Omega

Connected Teaching of Statistics
Wolfgang Härdle, Sigbert Klinke
Statistics is considered to be a difficult science since it requires a variety of skills including handling of quantitative data, graphical insights as well as mathematical ability. Yet ever increasing special knowledge of statistics is demanded since data of increasing complexity and size need to be understood and analyzed. Although this changing demand on educated statisticians is visible, our methods of teaching statistics follow essentially the ideas developed by our grandfathers in the fifties. An attractive and powerful new way of incorporating today's and future demands is via tools based on an intra- or the internet. In this article we suggest a set of criteria for effective web based teaching and propose the first net based approach to meet these criteria.
Download powerpoint file of slides.

Future directions for statistical modeling software
Doug Bates, John Chambers, Duncan Temple Lang
Participants in the Omega project and others would like to encourage discussion about the next generation of software that deals with statistical models: What new capabilities are important? How can current languages and tools (e.g., Java) help? In this session, participants are invited to present ideas and suggestions. There is also a mailing list,, intended to be a forum for similar discussions.
View HTML version of slides.

(Possibly graphical) user interfaces
Anthony Rossini
I'm interested in talking/discussing a bit about ORCA (a "universal" graphical front-end for stat packages), and other (possibly graphical) user interface issues...

A Graphics library for Omega
Balasubramanian Narasimhan
Architecture, design and implementation of a Graphics library for Omega. This talk will rely heavily on the new release of Java 1.2.

Interfaces between existing statistical systems and Java/Omega
John Chambers
There are many ways to design interfaces between statistical systems. Java provides several attractive possibilities. We have been investigating interfaces built on the tools in the preliminary Omega language and on the Java Native Interface. We will discuss the progress to date and possible further directions.
Download postscript file of slides.

R / S

Things I think we did right, things I think we did wrong, and things I'm not sure about.
Ross Ihaka

Current quirks, an R (and S) wishlist
Peter Dalgaard
Download postscript file of slides.

R 2do list
Kurt Hornik

Numerical Accuracy in a Statistics Package: What Precision is Needed When ?
Martin Mächler
Download postscript file of slides.

Building GUIs for S-PLUS for Windows vertical applications
Brian Ripley
S-PLUS for Windows already allows a number of GUI connections, within a Windows-only environment. It has a programmable GUI interface of its own, and can interact with other Windows applications (such as spreadsheets) through `DDE' and `Automation'. We are currently building such interfaces to a cancer prognosis system for use with a laptop in an Oxford hospital. Hence, the experience gained in the simulation should be useful for devloping new concepts of information flow (data exchange) and data analysis in real life.
Download slides: postcript or PDF

Towards an R manual
Friedrich Leisch

Future of R
What are our goals for the future, how does this relate to projects like Omega etc.

Public Talks ("Adaptive Monday")

Directions for Statistical Systems
John Chambers
Download postscript file of slides.

Visualization and Pattern Recognition in Complex Datasets
Brian Ripley

Distributed Computing in the SFB
Friedrich Leisch, Andreas Weingessel
Download PDF file of slides.

Wrap Up Session

Lexical Scoping with R
Robert Gentleman

Random Number Generation for Parallel and Threaded Programs
Gregory Warnes
Download slides: postcript or PDF
Download C code.

Interactive Modeling
Stephan Lauer
Download postscript file of slides.

A Local Estimatio Approach in Multi-Categorical Varying-Coefficient Models
Silke Edlich
Download postscript file of slides.