Statistical Education Project

Summary

Since the inception of the Statistical Engineering Division by Churchill Eisenhart in 1947, education and training of the NIST staff has been a core component to the division's service, consulting, communication, and research mission. The benefits of such education are 5-fold:

1. Understanding of relevant probabilistic, data analytic, and experiment design methodology by the NIST technical staff, with subsequent improved (= greater efficiency, cost-saving, and insight) SRM certification, calibration, interlab studies, engineering experiments, and scientific investigations.
2. An appreciation and understanding by SED (Statistical Engineering Division) staff of the types of scientific/engineering problems being dealt with by NIST staff. This keeps SED staff "grounded" and avoids laboring on elegant solutions to problems of no relevance to the NIST scientists/engineers. Jack Youden was a superb example of such a "solving real problems" focus.
3. A synergistic augmentation of SED's consulting function: many staff students subsequently become SED consulting clients, bringing both short-term and long-term projects to the focus of SED staff.
4. A further synergy occurs when SED teaching staff and NIST staff students subsequently develop a collaborative relationship to work on and publish (statistical and engineering) methodological improvements which occur as a natural culmination of the 3-node teaching-consulting-collaboration cycle.
5. SED teachers frequently go on to publish their statistical methodologies in papers, monographs, and books. Youden's Monograph Series for the American Chemical Association, Natrella's Experimental Statistics book, and the web-based NIST/SEMATECH Engineering Statistics Handbook are examples of products which have resulted from the teaching experiences of the SED authors.

The tradition of excellence in the SED education program started in the 1950's with the superb teaching talents of the division's Jack Youden and Mary Natrella, both of whom are honored annually via awards and/or sessions at the national statistical professional society meetings. The epitome of NIST-relevant teaching was captured by SED's Mary Natrella's subsequent book "Experimental Statistics" (1978) which is #2 on the all-time NIST best-selling books (400,000+) (after AMS55: Handbook of Mathematical Functions").

In the 1970's and 1980's, training remained an important function for SED-NIST communication, with an aggressive and ongoing series of talks and specialized courses.

In the 1990's, SED teaching became more diversified, via several different formats:

1. Short courses (on-site and off-site) on specific topics (e.g., experiment design, uncertainty analysis).
2. Series of short courses on a variety of topics (e.g., the Statistics for Scientists series).
3. Web-based training via e-books (e.g., The NIST/SEMATECH Engineering Statistics Handbook).

We highlight statistical education with 4 examples:

1. NIST/SEMATECH Engineering Statistics Handbook (e-book).
2. Experiment Design for Scientists and Engineers Short Course.
3. Statistics for Scientists Short Course Series.
4. Professional Society Short Courses.

Example 1: NIST/SEMATECH Engineering Statistics Handbook (e-book)

Inspired by the style and content of Mary Natrella's "Experiment Statistics" book (1978), SEMATECH approached NIST in 1995 regarding the possibility of writing a modernized update of Mary's book (in both content and form--e-book). A CRADA was subsequently formed involving a working team of 8 staff members (4 from SEMATECH/AMD and 4 from NIST). Work on this extensive (2000 web-page) e-book is near completion. It is anticipated that it will serve as an invaluable resource for industry and university training alike. It comes with a "course builder" feature that automates the topic selection process for easy tailoring to a classroom environment.

Example 2: Experiment Design for Scientists and Engineers Short Course.

With the advent of improved data analysis techniques, the major impediment in scientific/engineering investigations shifted from the ability to analyze data to the ability to generate meaningful data. This week-long course on experiment design--techniques and tools to construct meaningful and bias-free experimental plans--was initially presented in 1990. It has subsequently been presented on an annual basis to NIST staff and on a less frequent basis to audiences from industry/government/academia.

Example 3: Statistics for Scientists Short Course Series.

The variety of problems encountered by the NIST scientist/engineer is broad and diverse--as is the variety of statistical solutions needed to solve these problems. In this context, a unified effort was initiated in 1995 to present an integrated, comprehensive and ongoing series of short courses on topics of specific import and relevance to NIST. Topics include exploratory data analysis, intervals and uncertainty analysis, regression, ANOVA, time series analysis, and more.

Example 4: Professional Society Short Courses.

The NIST scientists/engineers which SED staff consult with belong to a variety of discipline-specific professional societies. To assist in the transfer of methodology to external professionals who rely on NIST services and expertise, SED staff and NIST clients have collaborated in presenting a variety of discipline-specific short courses at (or in conjunction with) various professional societies. These short-courses (e.g., Uncertainty Analysis Techniques, Data Mining, etc.) have been particularly well received and have served to reach out to the broader scientific community.

Description

The NIST laboratory has over 1300 scientists and engineers formulating investigations, planning experiments, collecting data, analyzing data, and deriving scientific/ engineering conclusions. All of these scientific-method components may benefit from both the general and NIST-specific statistical principles and techniques developed over the years in conjunction with SED staff expertise/experience. The ultimate objective in statistical training is to assure valid, supportable, repeatable scientific/engineering conclusions while maximizing insight and minimizing experimentalist time/cost. It is the aim of SED training to maximize the leverage of the limited-size SED by putting the best statistical tools in the hands of the NIST scientist/engineer.

Example 1: NIST/SEMATECH Engineering Statistics Handbook (e-book)

• Audience : Semiconductor industry engineers, NIST scientists/engineers, general scientific/engineering/ industrial community.
• Objective: To provide a comprehensive and complete Internet-accessible on-line handbook of statistical methods (including experiment design) for (nominally) the semiconductor engineer but (in fact) for the broader science/engineering community.

Example 2: Experiment Design for Scientists and Engineers Short Course.

• Audience: NIST/industrial scientists and engineers.
• Objective: To provide state-of-the-art techniques for the construction and analysis of designed experiments in scientific/engineering applications.

Example 3: Statistics for Scientists Short Course Series.

• Audience: NIST scientists/engineers.
• Objective: To provide in-depth information on statistical methodology appropriate for NIST programs in SRM, calibration, science, and engineering.

Example 4: Professional Society Short Courses.

• Audience: Industrial and academic scientists and engineers.
• Objective: To provide detailed instruction on the latest statistical strategies/methods relevant to a specific scientific/engineering area.

Major Accomplishments

Example 1: NIST/SEMATECH Engineering Statistics Handbook (e-book)

• Carroll Croarkin: wrote Measurement Process Characterization chapter, edited Exploratory Data Analysis and Product Comparisons chapter, project management, web design.
• Jim Filliben: co-wrote Exploratory Data Analysis chapter, edited Production Process Characterization and Process Improvement chapters, software development, web design.
• Will Guthrie: wrote Process Modeling chapter, web design.
• Alan Heckert: co-wrote Exploratory Data Analysis chapter, added case studies in several other chapters, software development, web design.
• Nien-fan Zhang: edited Process Monitoring and Control chapter.

Example 2: Experiment Design for Scientists and Engineers Short Course.

• Jim Filliben: present part of 5-day course.
• Eric Lagergren: present part of 5-day course.

Example 3: Statistics for Scientists Short Course Series.

• Jim Filliben: present Exploratory Data Analysis.
• Will Guthrie: present Regression Models.
• Stefan Leigh: present ANOVA.
• Mark Vangel: present Statistical Concepts.
• Nien-fan Zhang: present Time Series Analysis.

Example 4: Professional Society Short Courses.

• Carroll Croarkin: present mass measurement course.

Additional Technical Details

Example 1: NIST/SEMATECH Engineering Statistics Handbook (e-book)

The SEMATECH semiconductor consortium companies have access to the Handbook from the SEMATECH web page (and will be able to mirror the SEMATECH site to their own local web pages).

The Handbook is available to NIST scientists and engineers.  

The Handbook is available to any industry/company with web-access.

Example 2: Experiment Design for Scientists and Engineers Short Course.

This course has been taught to many NIST scientists and engineers. In addition, numerous manufacturing companies have sent their engineers and scientists to this course.

Example 3: Statistics for Scientists Short Course Series.

These courses are provided to NIST scientists and engineers.

Example 4: Professional Society Short Courses.

to be provided