George E. P. Box Remembered

On March 28, 2013, the world lost a person whom many consider to be a major contributor to the world of industrial statistics: George E. P. Box. Relatively unknown outside the world of statistics, Box was certainly very well known by those who have studied or practiced industrial statistics.

His death at the age of 88 followed only nine months later by the death of another notable in statistics, Genichi Taguchi (see the article, “The Legacies of Genichi Taguchi”). As will be seen below, Box and Taguchi were also statistical sparring partners.

Although a brief overview of Box’s life and accomplishment is covered in the recent article, “Statistics for the Rest of Us” by Quality Digest contributor, Matthew Barsalou, we thought it would be interesting and informative to hear from the many statisticians and practitioners who have worked with Box, heard him speak, or have benefitted from his contributions.

What follows are some abridged comments that Quality Digest received from a variety of our readers, contributors, and associates.

Statistics for experimenters

George Box was a regular participant in the annual Gordon Conference in Statistics in Chemistry and Chemical Engineering. It’s amazing how many ideas developed by Box and his colleagues during these conferences eventually made it into industrial practice, especially the chemical industry, in the 1950s through the 1980s.

Statistics for Experimenters (Wiley-Interscience 2nd ed., 2005) is considered the classic book on statistical design of experiments. Box co-authored it with two other venerable gents and regular Gordon Conference participants, William C. Hunter and Stu Hunter. This quote appears before the title page:

“When the Lord created the world and people to live in it—an enterprise which, according to modern science, took a very long time—I could well imagine that He reasoned with Himself as follows: “If I make everything predictable, these human beings, whom I have endowed with pretty good brains, will undoubtedly learn to predict everything, and they will thereupon have no motive to do anything at all, because they will recognize that the future is totally determined and cannot be influenced by any human action. On the other hand, if I make everything unpredictable, they will gradually discover that there is no rational basis for any decision whatsoever and, as in the first case, they will thereupon have no motive to do anything at all. Neither scheme would make sense. I must therefore create a mixture of the two. Let some things be predictable and let others be unpredictable.  They will then, amongst many other things, have the very important tasks of finding out which is which.”
—From E. F. Schumacher’s Small is Beautiful

In chapter 1, one is immediately introduced to one of Box’s favorite diagrams:

Simply:  Moving from “hypothesis” towards “data” is always labeled “deduction.” The other arrow begins at the tail of the previous, moving downward to the right from “data” to “hypothesis” and is always labeled “induction.”

Box loved this diagram. And used it extensively during the discussion phases of the Gordon Conference talks. He’d walk up to the blackboard, give that wry smile of his, telegraphing, “Wait for it....” And then proceed to draw the diagram, usually to the laughter of everyone in the room. This one particular time, he prefaced this with one of his favorite expressions, “All models are wrong, but some are useful,” drew the diagram, put all of this in context of the talk, and said that every time one did a “deduction” to “induction” cycle it was “Nearer My God to Thee.” The room roared.

He was a brilliant theoretician—a true one-in-a-generation (or maybe two or three). A great teacher, witty, and a very approachable, humble human being, he was the ultimate statistician’s statistician—and every bit the gentleman. The public at large wouldn’t know him, but statisticians of my generation certainly did and universally respected him.
—Davis Balestracci
Author of Data Sanity (Medical Group Management Association, 2009)
MS in statistics

Statistics: Science or mathematics?

George Box was certainly one of the most original thinkers in statistics. One of his more revolutionary observations was that, “the domination of statistics by mathematics rather than by science has greatly reduced the value and the status of the subject.” This quote, from the abstract of “Scientific Statistics, Teaching, Learning and the Computer” (CQPI Report, June 1996), discussed the value of statistics in an iterative learning paradigm, very different from the theorem-proof paradigm of mathematics.

He noted: “An important issue of the 1930s was whether statistics was to be treated as a branch of science or of mathematics. To my mind unfortunately, the latter view has been adopted in the United States and in many other countries. Statistics has for some time been categorized as one of the mathematical sciences, and this view has dominated university teaching, promotion, tenure of faculty, the distribution of grants by funding agencies, and the characteristics of statistical journals. All this has, I believe, greatly limited the value and distorted the development of our subject.”

There are too few influential statisticians who think that way. The loss of one of the most influential thought leaders in our subject will be keenly felt. This paper, and the first chapter of Statistics for Experimenters, should be required reading at the beginning of every statistics course.
—Rip Stauffer
Manager, Quantitative Analysis and Improvement, WBB Inc.

Statistic’s renaissance man

George Box was surely statistics’ Renaissance man. I was first exposed to him as the young author of a chapter on response surface plans in an early book (edited by Owen Davies) on experimental design.

Box-Cox transformations, evolutionary operation (EVOP), and Box-Jenkins time series models are only a few of his many notable contributions. He was a prime force in broadening the use of statistics in industry. His sayings and philosophy (e.g., “All models are wrong, but some are useful”) are for the ages.

I had the great privilege of learning from him on many occasions and have benefited immeasurably from his wisdom. On a personal level, I was most interested in exchanging World War II experiences in England. I saw him last at the launching a few years ago of the second edition of his classic book (with Bill and Stu Hunter), Statistics for Experimenters. He will surely be missed by all of us.
—Gerry Hahn
Manager, Applied Statistics, GE Global Research, retired

Box vs. Taguchi

It is often overlooked that George was a practicing statistician as well as world-class researcher. I had the privilege of being a co-editor of his collected works (The Collected Works of George E. P. Box (Chapman and Hall, 1984). My role was to focus on his work on applications of statistics. It was an enlightening and stimulating experience. I saw George iterating between practice and theory as he solved problems. I marveled at his clear and interesting writing style. He made extensive use of graphics and schematics of all types. I saw George, the talented scientist as well as George, the statistician, at work. In one article he commented that “statisticians must learn to be good scientists, a talent which, I think, has to be learned by example.”

In the 1980s Taguchi methods became a popular alternative to the design of experiments approach being used in industry at the time. George was in the middle of the discussion. This marked George’s return to a focus on design of experiments research, which he had moved away from to focus on other research endeavors. Two major meetings on the issues that I attended in the 1980s were the Mohonk Conference and an ASQ Annual meeting. Both George and Taguchi spoke and debated the issues.

I marveled at how George handled the charged situation. He was, as always, a highly polished professional, relying on sound scientific and statistical thinking and keeping emotion out of his arguments. In private conversation it was clear he had plenty of emotion around the issue. Taguchi, of course, made many important contributions, including getting the quality and statistical professionals to the focus on product and process robustness. Some feel another important Taguchi contribution was getting George Box to focus on design of experiments again.

Around the year 2000 I wrote to George requesting a meeting at a place in Madison, Wisconsin, with me and four or five of my colleagues to discuss Six Sigma. I got a phone call from George in response telling me that he could not participate in such a meeting as his health and doctor’s orders would not permit. He said that while he would greatly enjoy such a meeting, he would get too excited during such a discussion, which would not be good for his health.

He then went on to talk about Six Sigma, its value and promise. He was enthusiastic about, and supportive of, the approach to improvement. He noted the great positive impact it had already had on industry around the world and, in particular, noted that Six Sigma had gained the support of upper management as no other approach had been able to do. It was clear that he had thought deeply about the topic and its strengths and opportunities for statisticians, quality professionals, and society in general. I was overwhelmed by his comments. His depth of understanding and vision of the role and benefits of Six Sigma were clearly communicated in just a few moments.
—Ron Snee, Ph.D.
Assistant professor, Statistics Center, Rutgers University, from 1966–1968
Founder and president, Snee Associates, LLC

“It’s still wrong”

I met George Box for the first time at a Gordon Research Conference on statistics in chemistry and chemical engineering, at which he was somewhat of a regular. He was always passionate about whatever he was discussing, and in many cases, what he was debating. I remember one speaker gave a talk on time series, and George got very animated during the discussion period, challenging the speaker’s work (which was not at all unusual for the Gordon Conference).

George asked in a loud voice: “Where did you get that equation? It’s wrong!”

The speaker thought for a moment, and then replied: “Well, I actually got it out of your book, George,” meaning the Box and Jenkins time series book.

You could have heard a pin drop at this point; the attendees were pretty much all holding their breaths.

Not to be outdone, George replied without hesitation: “Well, it’s still wrong!” and sat down.
—Roger Hoerl
Professor of statistics at Union College, Schenectady, New York

Clarity of thought

I first became aware of Dr. Box when I was starting out in my career in the 1970s. His work on design of experiments and response surface analysis was already legendary. Unlike most other authors of technical topics in quality, Dr. Box's writing style was down-to-earth and unpretentious. He made the subject of design of experiments, which can be daunting, fun to learn. His other works were no less accessible.

Some years later I attended the great “Box vs. Taguchi” debate at the ASQC annual conference. It was there that I discovered his humor and wit. His arguments were, as usual, clear and easy to follow. By that time I’d already decided that Dr. Taguchi’s revolutionary accomplishments were conceptual rather than statistical, but I enjoyed the spirited verbal banter nonetheless.

Dr. Box continued to make contributions to the body of knowledge in quality engineering and statistics until his death. His writings appeared in Quality Engineering as recently as November, 2012.

The section headings in that article titled, “Innovation, Quality Engineering and Statistics” and co-authored with William H. Woodall, illustrate the clarity with which Dr. Box thought and wrote:

What is Innovation?
Innovations in Statistical Science
Inductive-Deductive Discovery
Use of Lateral Thinking
Use of Discussions and Group Interactions
The Use of Statistics and Experimentation
Leadership
General Advice to Individuals
Use of Analogy
Recommendations for the Quality Profession

All that in 10 concise pages. Amazing.

I—we—will miss this great man, a true pioneer and leader in quality.
—Tom Pyzdek
President, The Pyzdek Institute LLC

“On the shoulders of giants”

Recently I have been working with a client on designing a semiconductor fabrication experiment. And, as you know, one great contributor to DOE was George Box. George’s contributions to statistics and quality were large and fundamentally significant. His Statistics for Experimenters and Times Series Analysis (Wiley, 2008) are classics. His work at the University of Wisconsin in quality and productivity was of tremendous value to society. As Newton said, “If I have seen farther than others, it is because I stood on the shoulders of giants.” George Box was one of those giants who allowed us to stand on his shoulders, and for this we owe him our eternal gratitude.
—John J. Flaig, Ph.D.
Managing director, Applied Technology

About The Author