Other photos Duncan Temple Lang, Associate Professor (Temple Lang is the surname) Address: Department of Statistics 4210 Mathematical Sciences Building One Shield Avenue Davis, CA 95616 USA e-mail: duncan at wald dot ucdavis dot edu Phone: (530) 752 4782 Digital signature: GPG Primary Projects The S Language. Omegahat GGobi StatDocs S packages Current Activities

Research

The focus of my research is innovations in information technology and integrating computer science research concepts with the process of scientific and statistical research. The goal is to think about where statistical computing should be in the future, e.g. 5 years from now, and to determine how to get practitioners there. While we develop software for the present practitioners, the primary motivation is enabling things in the future and not merely reimplementation of existing concepts. The hope is to enable statisticians to do things that they currently cannot attempt, to export statistical methodology to non-statisticians and to facilitate interchange and direct collaboration between disciplines.

An important aspect of my work is to facilitate the integration of software from different communities. It is important that statisticians be able to easily access both software and data from other disciplines. Similarly, it is vital that people in other fields using their familiar tools be able to use state-of-the-art statistical methodology developed by statisticians in their favorite environment, and be able to integrate them into their work process. I have developed a general and flexible model for integrating programming languages and systems to enable this interchange. We call this the inter-system interface model. This involves embedding systems within others, and also enabling the possiblity of programming using multiple languages.

I returned to academia from industrial research with the purpose of introducing modern statistical computing to the statistics curriculum. Rather than focussing almost exclusively on mathematical statistics, it is becoming increasingly important that students and researchers be able to work with data both in new formats and also in the increasing size we are continually experiencing. Data technologies, software development and basic computing proficiency are as important as statistical methodology in the context of the scientific process.

Projects

The S Language.

I am one of the developers of R, and the S4 engine, originally developed at Bell Labs, that underlies S-Plus.

Omegahat

The Omegahat project is both

• an experimental, next generation language and interpreter for statistical computing that is
  • designed for both interactive and programming/development use, by providing a language that is a hybrid of both S and Java, using optional type-checking, dynamically compiled extensions to allow integration across the interpreted and base languages (Omegahat and Java);
  • a component-based infrastructure that permits one to create modified and new systems from the fundamental structures,
  • a multiple-interpreter, multi-threading, distributed computing environment that can support numerous derived "languages",
  • extensible at both the user and system level.
• a collection of packages that extend the S language and environment for use in very new ways such as embedding these within other systems and vice-versa, using meta-computing for generating code, graphical interface technology, data exchange, inter-system interfaces.

  There are a large number of packages available from Omegahat.

GGobi

GGobi is a dynamic, interactive visualization system that provides linked plots for exploratory data analysis. Descending from XGobi, it is significantly more flexible, allowing it to be embedded in other applications and to be extended using high-level functions in R, Perl or Python. A plugin mechanism allows it to be directly extended in the core system for different goals.

StatDocs

StatDocs is a project to develop both a technology infrastructure and content for authoring and rendering both dynamic and interactive. These documents are dynamic in the sense that when processed, the contents are generated by evaluating the code that one would run manually. This leads to reproducible results. They are interactive in the sense that the contents can contain interactive GUI components.