International Workshop on Robust Statistics and R (07w5064)

Arriving in Banff, Alberta Sunday, October 28 and departing Friday November 2, 2007

Organizers

(Universita' Ca' Foscari, Venezia, Italy)

Peter Filzmoser (Vienna University of Technology)

Matias Salibian-Barrera (The University of British Columbia)

Arnold Stromberg (Department of Statistics, University of Kentucky)

Objectives

The main objectives of the proposed workshop are to facilitate and coordinate the development of Robust Statistics tools in R and to promote the interaction and collaboration between researchers interested in the computational aspects of Robust Statistics. The long term goal is to lay the foundations for the development of a solid Robust Statistics library in R.

In the last few years the R-project has established itself as a widely available, powerful and versatile computer program for statistical analysis. In particular, R is distributed under the GNU public license, it is open-source, and has been already widely accepted and adopted by a broad community of students, practitioners and researchers in many disciplines (not only Statistics).

A library with all relevant developments of robust statistical methods in an attractive computing environment like R would be important for many people world-wide: for experts in the field since they could incorporate their own developments into this library and thus make new methods easily available to a wide audience; for professors, instructors and students since it would be much easier to teach/learn robust statistics using a powerful and uniformly designed computing tool; and for practitioners in general since they would have a quick overview of the best robust methods available for different models, and could easily apply these methods to their data and problems.

This workshop will also strengthen collaboration and communication among different research groups helping in the creation of new research networks and consolidate (mostly informal) existing ones. The proposed workshop participants include leading researchers from the most important groups worldwide working on robust statistics. They also include young researchers interested in both robust statistics and the computational challenges that remain to be addressed in large-scale applications.

The proposed workshop will continue the work started in the first International Workshop on Robust Statistics and R (Treviso, Italy, 2005). The work we initiated in 2005 will need to be revisited after each research group had the opportunity to work on the discussions that took place in Treviso. After 18 months, a preliminary version of the library is expected to exist and a large collection of programs to have been incorporated. There should already be clear guidelines to unify the different implementations (classes, common methods, input / output style, common graphical displays, help pages, etc.)

The proposed BIRS workshop will be necessary to make critical decisions about the general structure of the programs developed so far (accesibility for non-experts, integration with other packages already existing in R, degree of unity and internal coherence, missing methods / techniques that need to be incorporated, etc.) We would also use this opportunity to revise our development road map and discuss new targets, for example: updates to algorithms and estimation / inference methods (new developments in the area that need to be incorporated) and the ability of the current version of the library to manage large scale applications (scalability). This workshop will also provide a good opportunity to discuss and compare new proposals in the area that may have emerged between 2005 and 2007.

We believe that this workshop will be a unique and very important opportunity for leading and young researchers in the field of robust statistics to work together towards the computer implementation of the best robust methods in an open-source, widely-used and high-quality software package. The BIRS workshop structure is ideal for this purpose and we believe that we could not achieve the same results, with comparable efficiency, in a "classical" scientific meeting. We also anticipate that the practical and theoretical challenges identified by the development of this library will become a driving force for new research and scientific activity in the field.