SPES Sessions at the Joint Statistical Meetings
Toronto Aug 8-12, 2004

Stan Young, NISS
2004 SPES Program Chair
Angela Dean, Ohio State University
2004 SPES Program Chair-Elect

Roundtable Topics
Short-course: Mixture Experiments
Awards for Outstanding Presentations
Champion SPES/Q&P Mixer!
Young Investigator Award Recipients

Roundtable Topics

Using Data from Customers to Improve Engineering Decisions
Jeff Robinson, General Motors R&D Center

Companies often collect a lot of data about their prospective customers and product feedback from their current customers. But using the data effectively is a challenge. This roundtable discussion will summarize some of the issues related to using customer data for effective engineering decision making in the auto industry. It is intended for those wanting to learn a little more as well as for experts from other industries willing to share their experience. We are likely to raise more questions than answers, but the goal is to encourage statisticians to think more seriously about ways to help engineers make better decisions. Some of the data sources could include warranty data, customer satisfaction surveys, and product preference data from customers. Some of the decisions engineers have to make are how to run the plant, how to allocate resources to solve product problems, and how to incorporate customer feedback into future designs.

What's happening in computer experiments?
Bill Notz, The Ohio State University

This roundtable will provide an overview of computer experiments followed by a discussion of some current issues in the area. We will begin by explaining what a computer experiment is, give some examples of computer experiments, describe some of the statistical issues that one faces, and conclude with a discussion of topics that are of current interest. This roundtable is intended both for people who are interested in the area but know little about it as well as researchers who would like to share some ideas about topics that they believe are of current interest. The goal is to encourage new people to enter this area of research as well to allow active researchers to share some ideas.

Issues in design and analysis of discrete choice experiments
Stephen Jones, The Boeing Company

This roundtable will give an overview of the popular market research tool of discrete choice experiments, illustrating the methods with some examples from product development at the Boeing Company. We will then discuss some of the statistical issues that distinguish discrete choice experiments from the typical application of Design of Experiments methodology. This roundtable is intended both for practitioners and researchers. The objective is to encourage new people to enter this area of research as well to allow practitioners and researchers to share some insights about issues that they believe are of current interest from both a practical and a research perspective.

Web-based catalogue of designs
John Morgan, Virginia Tech

Where do you go to find a good design for a comparative experiment? Textbooks contain limited tables of standard designs, most commonly balanced incomplete block designs and less frequently lists of lattices, partially balanced incomplete block designs and members of generalized cyclic families. While the last thirty years have seen tremendous progress in finding optimal designs, corresponding tables, when they exist, are scattered throughout the literature. Design search software is likewise limited, with inability to allow a multi-criteria approach to design selection and no guarantee that an optimal design will be produced. In 2001 the United Kingdom Engineering and Physical Sciences Research Council funded a project entitled "A Web-based Resource for Design Theory." From this award is growing the Design Theory Resource Server (DTRS) for combinatorial and statistical designs at www.designtheory.org. DTRS is developing

1. a database of designs, including serving as a repository of designs as they continue to be discovered,
2. software for constructing and manipulating designs, and investigating their statistical and combinatorial properties,
3. documentation which will explain the types of design, their uses, and their representations in the database.

DTRS is being built to serve all parties interested in combinatorial design, from statistical practitioners to mathematical researchers. Of special interest to statisticians are evaluations of designs from many optimality, robustness, and efficiency perspectives. Thus DTRS will offer statisticians an extensive and ever-growing catalog of designs that can be sorted and compared from a variety of perspectives, bringing practical meaning for real experiments to the term "optimal design." This roundtable will be a forum for discussion of the DTRS project: its goals, its implementations, its modes of communication and how they may best serve the statistical community. We will also discuss advances in computational tools for design construction and enumeration, including the recently released GAP Design package.

Top

Short Course: "Methods for Designing & Analyzing Mixture Experiments"

Sunday, August 8, 2004, 8:15 AM - 4:15 PM

At the 2004 JSM, SPES is co-sponsoring with ASA a 1-day (Sunday, August 8, 2004) short course Methods for Designing & Analyzing Mixture Experiments that will be taught by the two leading experts in the area, John Cornell and Greg Piepel. Mixture experiments involve changing the proportions of product components and observing the changes in the product's characteristics. Mixture component proportions cannot be varied independently (as in factorial experiments) because they must sum to 1.0 for each run in the experiment. Mixture experiments are very useful in many product development areas, including foods, materials, fertilizers, textile fibers, drugs, and many others.

The short course will provide an overview of methods used in designing mixture experiments and analyzing the resulting data. Topics to be covered include: (1) designs for simplex-shaped and irregular-shaped regions (the latter resulting from additional constraints on the component proportions), (2) various types of mixture models for fitting mixture data, (3) graphical techniques for interpreting component effects, (4) including process variables and/or a total amount variable in mixture experiments, and (5) graphical and analytic methods for developing mixtures with optimum properties. Numerous examples will be used to illustrate the topics discussed.

The course is designed for anyone (statistician or non-statistician) wanting to know about statistical methods for designing mixture experiments and analyzing the resulting data. Prerequisites are an understanding of elementary statistics concepts and some previous exposure to experimental design and least squares regression.

For more information about the course, see Greg Piepel's page about it on the MIXSOFT web site (http://members.aol.com/mixsoft/mixsc2.htm), or email either of the two presenters directly.

Top

SPES Awards for Outstanding Presentations

Ana Ivelisse Avilés, Awards Chair, National Institute of Standards and Technology

The Section on Physical and Engineering Sciences is pleased to announce the results of its annual Outstanding Presentation competition for papers given at the 2003 JSM in San Francisco. (Results for papers given at the 2004 JSM in Toronto, Canada, will be announced in Spring 2005.) These awards are based on audience evaluation of papers contributed to SPES-sponsored sessions at the Joint Statistical Meetings. The purpose of the awards is to encourage continuous improvement in the presentation of statistical information by recognizing at least some of the truly excellent talks given each year in SPES sessions. The winners received cash awards as part of their recognition: $100 for best presentation, $50 for runner-up, and $25 for honorable mentions. Plaques were also awarded. The awards for papers given in 2003, which were presented at the SPES/Q&P mixer during the meetings in Toronto, Canada, are:

Outstanding Presentation Award Dr. Thierry Duchesne, of the Universite Laval, for his SPES Contributed Paper presentation, “Dropping the Masks: Simple Inference Procedures Based on the EM Algorithm for the Competing Risk Model with Masked Causes of Failure”.	No photo
Runner-Up Outstanding Presentation Award Dr. Christopher H. Tong, of the Merck & Co., Inc, for his SPES Contributed Paper presentation, “Quantitative Structure-Activity-Relationship Modeling Using Leo Breiman's Random Forest”.	No photo
Honorable Mention Outstanding Presentation Award William F. Guthrie, of the National Institute of Standards and Technology (NIST), for his SPES Contributed Paper presentation, “A Bayesian Analysis of a Chain of Calibrations Used in the Preparation of Electrolytic Conductivity Standards”.
Honorable Mention Outstanding Presentation Award Christian Ritter, of the Catholic University of Louvain, for his SPES Contributed Paper presentation, “Dual Response Displays: The Incredible Flying Carpets”.	No photo
Honorable Mention Outstanding Presentation Award Chih-Ming Wang, of the NIST Boulder Laboratories, for his SPES Contributed Paper presentation, “Fiducial Approach for Propagation of Uncertainties in Measurement”.
Honorable Mention Outstanding Presentation Award Steven E. Rigdon, of Southern Illinois University, for his SPES Contributed Paper presentation, “The Duane Plot for Repairable Systems”.
Honorable Mention Outstanding Presentation Award Dr. Christopher J. Malone, of Kansas State University, for his SPES Contributed Paper presentation, “Separating Location Effects from Dispersion Effects in an Unreplicated Factorial Design”.

I would also like to thank all the volunteers for organizing and submitting the evaluations. The SPES Awards Committee Members are Cheryl Dingus and Tena Katsaounis from The Ohio State University, Will Guthrie, Adriana Horníková, and Sarah Streett from NIST, and Jazmin Rae Fulkersin from Cresset Christian Academy in Durham NC.

Top

Champion SPES/Q&P Mixer!

Top