Conference Overview

Hotel & Travel

Call for Posters

Deadlines

Interest List

Tell a friend
Enter a friend's
e-mail address:

Presentation Abstracts

Keynotes

Data Mining Technology and Its Support of Six Sigma/Lean Initiatives

Michèle Boulanger, JISC Consulting

Data mining is attractive for its ability to retrieve valuable information embedded in large, complex, disparate sources of data. Yet, basic statistical paradigms have precluded its ready acceptance by quality practitioners. With the increased generation rate of very large data sets housed on multiple data bases across the IT network - the types involving customer transactions or product measurements - there is a wealth of knowledge hidden in data warehouses that often pertains directly to the tasks at hand in Black Belt projects. Extracting that knowledge involves specially designed tools and techniques that constitute the essence of data mining. We will review data mining opportunities for Six Sigma/Lean practitioners and infer that, by linking the Six Sigma/Lean community to the Information Technology area of a company, both parties can reap the benefits.

Genomic Insights into Chronic Pain

Susan Dorsey, University of Maryland Baltimore

Although we are midway through the Congressionally-declared Decade for Pain Control and Research, chronic pain continues to be a leading public health epidemic, affecting more than 50 million Americans and costing more than $165 billion per year in treatment expenses and lost work productivity. Moreover, despite recent advances in treatment, most patients do not obtain adequate pain relief. An important focus for drug development has been to identify new molecules that participate in the development and persistence of chronic pain. Rather than taking a candidate approach to the identification of new targets, we chose to exploit the power of genomics to examine gene expression changes in an animal model of antiretroviral therapy-induced chronic pain. Using Affymetrix arrays, we were able to explore changes in gene expression and alternative splicing in drug vs. vehicle-treated animals. We have further verified differentially regulated genes in drug treated animals via Quantitative PCR methods. Verified candidate genes will be examined in more detail both in vitro and in vivo for potential therapeutic relevance to treat chronic pain.

Overcoming Fear of Confounding in Screening Design

Bradley Jones, JMP

Two-level fractional factorial designs are the standard tool for screening many potentially active factors to find the vital few factors that have the greatest effect on the response. In a first course on DOE students learn to create these designs by starting with a full factorial design and creating new columns by multiplying two or more columns together. This procedure necessarily confounds the main effects of these designs with higher order interactions. Resolution III designs confound main effects with two-way interactions and Resolution IV designs confound two-way interactions with other two-way interactions and main effects with three-way interactions. This confounding often makes the results of such experiments ambiguous. Instructors point this out and instill fear of confounding in the students. The fear of ambiguous results can discourage a novice from even trying screening experiments.

In the last decade a number of new orthogonal designs have been discovered that do not create complete confounding of main effects with interactions. These new designs are called irregular fractional factorial designs. Using these designs it is possible to avoid the ambiguity of pure confounding that comes from using the standard methods. Thus, these designs are more model robust than the traditional (regular) fractional factorial designs, which will be demonstrated using a case study.

Futures Trading Strategies via Robust Product Design and Predictive Modeling

Christopher J. Nachtsheim, University of Minnesota

In this presentation, we summarize research on the identification of robust futures trading strategies based on designed computer experiments, robust product design techniques, and data mining methodologies using JMP. We seek trading strategies that are robust to random changes in market conditions and employ Taguchi-inspired robust product design experiments to generate candidate strategies. Neural networks and regression trees are used to analyze the experiments, to identify promising trading strategies and to validate results.

Extreme Six Sigma using JMP

Damon Stoddard, Microsoft

The classical approach to developing Six Sigma analytical capability in an organization starts by training key individuals over a 4 month period of time with application of the methods to key business projects. While extremely effective, there are times when 4 months is too long.

Mr. Stoddard has pioneered an approach at Microsoft in building analytical capability and driving rapid results. This approach, dubbed "Extreme Six Sigma" builds on principles developed in Extreme Programming-pairing up individuals for instant feedback on the quality of their work while accelerating the learning curve on JMP's analytical capabilities. JMP enables Extreme Six Sigma through its seamless integration of databases and powerful data visualization techniques.

In this talk Mr. Stoddard will share principles of Extreme Six Sigma with case studies while demonstrating the power of integrating JMP with large scale databases (using SQL) for rapid discoveries in data.

Session Speakers

Sine-Cosine modeling of hourly electric system load with analyses of temperature, time and customer-class contribution to system load: a graphical demonstration using JMP

Larry Arvidson, Central Hudson Gas & Electric Corp

Part 1 illustrates the use of JMP spectral analysis to identify the cyclical components of hourly electric system load (1960-2008); then, the use of JMP to build time series models containing sine and cosine functions based on the identified cycles. Visual evaluation of the model (via Overlay Plots, Scatterplot Matrix & Matched Pairs) supports fine-tuning the model's R-square by adding cross product & polynomial terms for hourly temperature. To forecast future hourly loads, the model utilizes forecasts of hourly temperature; then, comparison of predicted hourly load (using actual hourly temperatures for March through May 08) to realized hourly load, yields a final evaluation of the Sine-Cosine model.

Part 2 dissects the customer class contribution to hourly electric system load over three contiguous years, Nov 04 through Oct 07, from load research data. Overlay Plot, Contour Plot and Surface Plot are used to clearly demonstrate the relative importance of temperature & class contribution to hourly system demand. The relationship of class average hourly demand to time of day, day of week, and temperature are further explored visually using many of JMP's analytical & graphical features: Spline fit of Y by X; Surface Plots of load vs. time by class; dynamic prediction profiles vs. time & temperature variables; and, Sine-Cosine modeling as applied to selected customer-classes.

Data Mining with JMP

Tom Bohannon, Analytical Business Solutions

This presentation with demonstrate how JMP can be used for data mining applications using a well known dataset. The talk will illustrate the use of JMP in the application of SEMMA to a data mining project. The modeling tools applied in this application are: decision trees, regression, and neural networks. Statistics for comparing the various models will be discussed and used to select a model. The chosen model will be used to score a dataset.

Classroom Experiences Using JMP Software for Data Mining

Michael D. Conerly, University of Alabama

I have used JMP software to perform data mining in Masters level courses at the University of Alabama. This presentation will discuss some of the modeling techniques used as well as some lessons learned in creating predictive models. Methods such as cluster analysis, discriminant analysis, logistic regression, decision trees and neural networks can be easily done with JMP. Creating hold out data for testing purposes and constructing tables to compare competing models will be discussed.

JMP Through the Clients' Eye

Karen Copeland, Boulder Statistics, LLC

I am a statistical consultant who is a generalist - that is, I don't focus on one industry or one area of statistics. What I do focus on is helping clients understand their data, be it historical data, or data from a planned study or survey. Graphics, especially JMP graphics, provide the bridge between the statistics and the meaning in my client's data. This talk will be a "sampler" of my favorite JMP platforms for displaying data. Examples taken from my consulting will range from mundane to those with great "ah-ha!" moments, all in an attempt to show the audience some ways to present data that they may not have thought of before. I'll also show some examples that are so basic and "obvious" that we experts forget about them, however, they're sometimes exactly what we need to remember to show our non-expert clients.

A New Classification of Variables in Design of Experiments

John J. Flaig, HP Emeritus, Fellow ASQ, Applied Technology

The standard system factor classification of process input variables as controllable or uncontrollable does not reflect the observed structure of some processes. Specifically, some factors that are classified as controllable are actually only semi-controllable. In this paper the classic variable structure is extended to controllable, semi-controllable, and uncontrollable. This extension is proposed in an effort to deal more accurately with real world problems.

Presentation Quality Graphics in JMP

Xan Gregg, JMP

What makes visualizations "presentation quality" and how do we create them in JMP? The data visualization world recognizes a dichotomy consisting of interactive displays for exploration versus polished graphs for presentation. JMP generally focuses on the former category. We'll discuss the hurdles users are facing and share tips for making JMP graphics ready for publication.

Putting Stats into Perspective: A New Way to Think About Stats for Industry

John Haury, www.applied-stats.com Consultant

Integrating the major tools of industrial-strength statistics to help prioritize efforts and to understand why one set of tools is used in preference to others: It depends on the goals and the available data! This talk is a step towards a unified theory of applied stats. It will be useful to and understandable by the non-statistician. Too often the path taken to find an answer is chosen arbitrarily. This talk will take the arbitrariness out of statistics. The strong interdependency of Measurement Quality, Process Predictability and Experimentation is not easily recognized by clients. Common to all three is the need to collect samples. My presentation gives a new and simplified statistics paradigm for non-statisticians. Outline of my presentation:

The necessity and the challenges of sampling
Using samples to understand and improve measurements
Using measured results to understand processes
Using experimentation on a stable process to improve it

Using JMP to Make Multiplex less Complex

Aviva Jacobs, Osmetech Molecular Diagnotics

The development, manufacture, and validation of an FDA-regulated multiplex DNA diagnostic product require rapid and rigorous analysis and interpretation of large amounts of data. The large number of variable factors that contribute to product performance have required the use of the powerful analysis and graphical representation functions of JMP. For several years, we have used JMP as a tool to design, analyze and interpret complex experiments, while benefiting from JMP's scripting capabilities and from collaborations with SAS statistical consultants. In this presentation, we will describe some of the challenges we have faced with our data analysis and experimental designs, and how JMP has contributed to our success as a cutting-edge diagnostic company.

Data Preparation for Forensic DNA Typing

Mark Johnson & Jack Ballantyne, University of Central Florida

JMP 7 and other user friendly packages with data mining capabilities allow the investigator to plunge rapidly into predictive analytics. However, for the primary data set considered here (as well as most other sets), data preparation is critical. Data on Y chromosome short tandem repeats for about 14,000 males are investigated for ethnicity prediction. Although the data set as provided was relatively "clean" there remained considerable work to put the information in a form suitable for JMP or SAS EM analyses. Functionality in JMP is ideal for transforming nominal values into multiple numerical variables having predictive power. This data set was used for the final project in a data preparation class taught in the recent spring semester at the University of Central Florida. The results of this competition will be reviewed. Additionally, a somewhat simpler case for data preparation will be made in the context of a Titanic Disaster data analysis used effectively in a professional MBA course this past spring.

Managing Multiple Sources of Data in the Food Service Industry Using JMP

Joni L. Keith, Independent Consultant

I am not a JMP expert. In fact, I have been a JMP user for less than a year. However, in just a few short months, I have been able to utilize many of JMP's platforms and modules to simplify my life as a statistical consultant, and I have also been able to provide my client a more comprehensive evaluation of their research data. In this session, I will describe how JMP is being used at Bush Brothers and Company (a family-owned corporation based in Knoxville, Tennessee, best known for its baked beans) to organize and analyze analytical, descriptive, and consumer data. You will also see how the JMP Profiler allows the researchers to interact with their data so they can make the best decisions for their research, development, innovation and business needs.

Engineering Analytics with JMP 7

Jose Ramirez, W.L. Gore and Associates

Nonlinear models are common in engineering and science. This is usually the case when modeling degradation phenomena like changes in the strength of material over time and varying temperatures. The Formula Editor and the Nonlinear model platform make it easy for us to specify and fit such models, while new tools like the motion-enabled Bubble Plot and the 3-D scatter plot help us visualize the results in a dynamic way.

Products that are robust to environmental conditions lead to better performance and satisfied customers. A cake mix example, in which the goal is to find the amounts of flour, shortening and egg powder (control factors) that will make the taste of the cake insensitive (robust) to variations in oven temperature and baking time (noise factors), will be used to show how the analysis of robust designed experiments can be done using the Fit Model platform. We will also show how the support for noise factors in the Profiler makes the task of finding robust settings of control factors a breeze.

Using Six Sigma to Trade Performance for Profit

Rob Reul, Isometrics Solutions

Securing sustainable growth points to the fundamental common denominator of all businesses - profit. Profit is defined here simply as revenue in excess of operating costs. Ensuring future profits successfully sustains business growth. The honed disciplines of Six Sigma have for decades contributed to profitable growth of businesses in every SIC code, domestically and internationally. Up until now, these significant contributions have been principally focused on reducing operating costs: eliminating inefficiency (Lean) and reducing variation (Six Sigma). This has been the predominate leverage point, and rightfully so, as project results have proven.

But now we can use the disciplines of Six Sigma to focus beyond operations to the revenue side of the profit equation with astonishing results. Six Sigma initiatives focusing on the Voice of the Customer "upstream" can greatly maximize revenue potential. This is accomplished by examining the relative importance of a product's (or service's) various features using a trade-off analysis that allows providers to better price and position their offerings for their marketplace.

The presentation will discuss this technique currently deployed by a Fortune 200 multi-national manufacturer whose ultimate aim is to maximize pricing with global sensitivity while providing customer segments with the exact product characteristics they most desire.

Key Learnings:

Participants will learn how to advance their voice of customer efforts to marketing choice-based experiments.
Participants will learn how to construct conjoint marketing experiments to be able to compare the relative worth of increasing performance of differing aspects of their businesses.
Participants will learn how the results of conjoint marketing experiments can be used to optimize market pricing to maximize profit while delivering on the required levels of performance.

Preview of JMP8

John Sall, JMP

A tour of some of the features we are working on for JMP8.

Using JMP for System of Systems Testing

Matthew Shike, Joint Navigation Warfare Center

The Joint Navigation Warfare Center (JNWC) is a US Strategic Command organization tasked with becoming the Global Navigation Satellite System (GNSS) knowledge center for the Department of Defense. GNSS includes the US Global Positioning System (GPS), the European Union's new Galileo system, and several others. In pursuit of this goal, the JNWC was tasked by the United States Navy to evaluate several integrated GPS/INS (Instrument Navigation System) navigation systems used on various aircraft. A test was conducted in order to evaluate the navigation systems, and JMP was the primary tool used to make sense of the resulting data. Data from these systems was semi-standardized, but there were some differences which complicated our analysis. Topics include using scripts to manipulate large data sets, using scripts to reduce analysis time, the importance of configuration management (even with a small team), using data filters to isolate certain conditions, and presenting data to the layman.

Using JMP Scripting to Make Time for Insightful Data Analysis

Lise Varner, Cambrios Technologies

Many industries and disciplines use statistical software such as JMP to organize, analyze and track large amounts of data. Without scripts, routine charting and analysis can consume so much time that in-depth analyses or special comparisons among related data sets get pushed to the back burner. This could have negative business impacts such as increased product development time due to delays in identifying key process parameters. Using an example of reliability data from a new product under development, this presentation will compare the chore of manual routine analysis to the fun of clicking "Run Script" and watching hours of work completed in seconds. A two-minute tutorial will show any JMP user how to get started with scripts. Then demonstrations of scripts with user input will illustrate different ways of looking at the same data, ways for revealing key variables and for presenting data clearly to both JMP and non-JMP audiences. How one engineer minimally proficient in JMP Scripting Language can make all of these analysis methods accessible to all JMP users in a company should be readily apparent by the end of this presentation.

JMP 8: Not Just a Pretty Face, New Statistical Abilities, and Lots of New Visualization Capabilities, but also the Most Complete Programming Environment for SAS

Jon Weisz & Jeff Perkinson, SAS

JMP is really three products in one: a point and click interactive statistical analysis for researchers, a graphical data visualization for business users and an application development/analysis tool for SAS coders and analysts. How do these three personalities live together? Come see how magic can happen when you mix these three core capabilities of JMP together. This talk will highlight new features of JMP 8 and case studies showing how JMP can act as a front end to SAS data management and analytic capabilities.