The difficulty in taking the wider robustification approach is that we cannot expect to get good results unless we are really prepared to engage in the hazardous undertaking of finding out what the world is really like. It requires us to have some knowledge of reality.
I believe we do have members, we may have ASA fellows, possible we even had ASA presidents who really do not care what the world is really like. Some years ago, a friend of mine told me about his daughter who was then at oxford University. She was a very bright girl, but she got interested in politics (it was in the 1960s); she got behind in her studies, and the time of graduation was approaching. You may know that in the English system, there are many different grades of bachelor's degrees. The young lady started to worry: Was she going to get a “pass” degree (which is almost like the University spitting at you), or was it to be a third class, lower second class, upper second class, or a first class honours degree? She decided to ask her tutor about it. Finding him buried in the dust of one of the Oxford colleges, she eventually got around to asking him the delicate question, ‘Would it matter in the outside world if I didn't get a very good degree?’ He looked very startled and said, “Outside world? What do I know about the outside world?”
It wasn't long after the new department was started that I became afraid that the teaching of the students might be overly theoretical. This prompted me to begin the “Monday Night Beer Session,” which met every Monday evening in the basement of my house (Figure 7.3). It was not a formal university course; you got no grades or credits, and you just came when you felt like it. It could be attended by students and faculty from any department.
Figure 7.3 Announcement for Monday Night Beer Session.
People came, piled their coats on the ping pong table or on Harry's train set, and we talked. We sat on an odd array of chairs and on an old sofa that had seen better days. There was a cupboard door that we painted black, and it became a substitute blackboard. Svante Wold arrived one evening announcing that he had a surprise for me: It was a chalkboard eraser with a hole drilled in it and a good length of string attached. Not a lot of beer was consumed, but it was always available. Brian Joiner reminds me that I sometimes forgot all about the beer until the last minute, when I would call him and ask him to pick some up on the way over.
Some of the “regulars” at our weekly sessions acted as “talent scouts,” looking for people from fields such as engineering, medicine, and business who had problems they wanted to discuss. Usually a graduate student took on this responsibility, and they would visit the engineering school and the College of Agricultural Sciences, as well as other departments, to invite professors to come and speak, or to recommend speakers. When Kevin Little arranged for our speakers, for example, he had Ken Potter from civil engineering speak on time series applications to hydrology, and Warren Porter came from zoology to talk about modeling problems in animal physiology.4
I tried to simulate the experience I had gained in industry by letting students experience the catalysis to discovery that occurred from discussions using statistics. The meetings were a great success. Often people who had brought a problem and received some help would let us know in subsequent sessions how their project was getting on, and a number of discoveries and more than one publication with multiple authors came about as a result. These meetings went on pretty much until my retirement, and former students and other alumni have often told me that what they found most helpful at Madison was the “Monday Night Beer Session.”
Much later, my daughter Helen, who gives creative gifts, presented me with a beer-making kit. When we tried it out, things went well until it was time to decant the beer into bottles. Somehow there was an unexpected explosion that left us both covered in beer and reduced to helpless giggles. Later, after I had mastered the bottling stage, I served the homemade brew at the Monday Night Sessions. Unfortunately my students didn't like my homemade brew and I decided that beer making was not my best talent.
In 1966, the General Electric Company offered me $20,000 a year for three years to spend as I saw fit. I used the money to finance the University of Wisconsin Statistical Consulting Laboratory and along with it, the “Statistician in Residence” program. We brought in experienced statisticians whose only duty was to work half time in the lab, and to be available to help anyone at the University with statistical problems. This allowed the students, who spent a semester in the lab as a course requirement, to experience “real-world” statistical problems and to learn from experts to be consultants themselves. From 1967 through 1974, we had six statisticians in residence who came from academia and from industry. These individuals were as follows:
1967–68: J. Stuart Hunter from Princeton University
1968–69: Graham Wilkinson from the Commonwealth Scientific Industrial Research Organization in Adelaide, Australia
1969–70: Donald W. Behnken from American Cyanamid Company
1970–71: G. Morris Southward from the Department of Experimental Statistics at New Mexico State University
1972–73: Harvey Arnold from Oakland University
1973–74: Svante Wold from Umeå University, Sweden
When funding from General Electric ended, the position was supported by the Wisconsin Alumni Research Foundation, which paid half of the resident statistician's salary. Beginning in 1974, Brian Joiner became the statistician in residence, and kept the job until 1983. Brian had spent eight years as a consulting statistician for the National Bureau of Standards and four years as a professor of statistics and director of statistical consulting at Pennsylvania State University. He was one of the three developers of the statistics package Minitab, which he introduced to the University when he came. Brian had a background in quality-type statistics, and had first met Dr. W. Edwards Deming in 1963.
As I said earlier, the first course I had to teach at Madison was the “Advanced Theory of Statistics,” later known as course number 709. Afterward, in the hands of the mathematicians who took it over, this became a “killer” course, as did its “sequel,” course number 710. Both courses were required of all graduate students in the department, and both were so difficult that they deterred good students from coming to Madison. The mathematicians, of course, regarded what I had taught as a lightweight version of the real thing, when in fact what I was teaching was a subject that was altogether different from the one they taught. Brian tried to introduce a new curriculum more in line with the original intent of the early days of the program, but by then the theoretical statisticians were in the majority and voted against his proposal. Brian resigned and formed a successful consulting company that I will speak about shortly.
Despite divisions in the Statistics Department, we came together for parties and celebrations, the most notable being the festivities that surrounded the fiftieth anniversary of the department in 2010. Also memorable were the annual Christmas parties that took place at my house (Figure 7.3). It became a tradition that the students presented a skit and the faculty presented a skit. These were high-class performances on which a good deal of effort was expended. The competition was intense, and often the students were ahead. This pleased me because I believed that if you could write a first-class skit, you could also write a first-class thesis. Originality and wit are very close.
An especially good student skit in the late 1970s was based on the first Star Wars movie. On another occasion, the faculty skit took place in the years before there were personal computers. The department had a large, stand-alone computer, and a particular student and his professor were known to have a very complicated and special program involving calculations in multidimensional space. It was impressed on all of us that we must on no account do anything that might affect this program.
In the skit, the first scene showed Brian Joiner at the computer at night. He could not resist fiddling with the machine, and suddenly there was a huge flash and all the lights went out. When they were restored, Brian had completely disappeared and it was concluded that he must have gotten
lost in n-dimensional space. So in the next scene there was much technical discussion on how to retrieve him. Finally a program was written that would bring him back—one dimension at a time. A very large cardboard box was brought in and connected up. At that time Brian had a beautiful mustache, so after some intensive computing, slowly, from one corner of the box, rose an automobile antenna on the end of which was a lifelike imitation of Brian's mustache. This was the first dimension. After a good deal more discussion and much computing, a sheet of cardboard emerged from one side of the box with a very large photograph of Brian Joiner (Figure 7.4).This was his two-dimensional representation. Finally, after much business and further struggle, there was a flash and the box flew open and Brian jumped out—in all three dimensions!
Figure 7.4 Brian Joiner retrieved from outer space.
For Christmas parties, some of us sometimes wrote and performed songs. Norman Draper wrote some very good songs, a memorable one being “The Chairman's Lot is Not a Happy One” based on “A Policeman's Lot is Not a Happy One” from Pirates of Penzance. At one party, three former chairmen sang it together. Other songs reflected topical statistical interests. The song “There's No Theorem Like Bayes' Theorem” was written at a time when there was still a great deal of argument about the validity of a particular means of using data to make inferences, sometimes called “inverse probability.” It was propounded by a clergyman, the Reverend Thomas Bayes, who lived from 1701 to 1761 and ministered to his flock at Tunbridge Wells, Kent, about 25 miles from where I was born. He did not publish his ideas, but his friend, Richard Price, communicated them to the Royal Society after Bayes' death.
Since its inception, Bayes's theorem has been looked on with varying degrees of enthusiasm, and other theories—the Neyman-Pearson theory and Fisher's Fiducial Inference—were proposed to avoid the inverse argument. However, a number of statisticians, including myself, came to believe that Bayes had been right to begin with, hence the following, which is sung to the tune of Irving Berlin's “There's No Business Like Show Business”:
VERSE (1)
The model, the data you can't wait to see
The theta, beta, sigma, and the rho
The Normal, the Poisson, the Cauchy, the t
The need to specify what you don't know
The likelihood for data you acquire
The perspicacious choosing of the prior
REFRAIN
There's no theorem like Bayes' theorem
Like no theorem we know
Everything about it is appealing
Everything about it is a wow
Let out all that a priori feeling
You've been concealing right up to now
There's no people like Bayes' people
All odd balls from the urn
The other day you thought that you had got it straight
Take my advice and don't celebrate
A paradox by Lyndley could arrive quite late
Another Stone to unturn5
REFRAIN
There's no theorem like Bayes' theorem
Like no theorem we know
You can lose forever that perplexed look
If you start to study it right now
Even more enthralling than a sex book
You'll find that textbook
By Box and Tiao
There's no dogma like Bayes' dogma
It's great knowing you're right
We know of a fiducialist who knew the lot
We thought at first he had hit the spot
But after three more seminars we lost the plot
We just could not see the light
REFRAIN
There's no theorem like Bayes' theorem
Like no theorem we know
Fisher felt its use was quite restricted
Except in making family plans for mice
But there, he said, for pinning down a zygote
I'd give it my vote
And not think twice.
There's no answers like Bayes' answers
Transparent, clear and precise
Stein's conundrums you can solve without a blink6
Best estimators in half a wink
You can even understand what makes ‘em shrink
Their properties are so nice
VERSE (2)
There's Raiffa and Schlaifer, Mosteller and Pratt
There's Geisser, Zellner, Novick, Hill, and Tiao
And these are all people who know what they're at
They represent Statistics' finest flower
And tho' on nothing else could they agree
With us they'd join and sing in harmony.
REFRAIN
There's no theorem like Bayes' theorem
Like no theorem we know
Just recall what Pearson said to Neyman
Emerging from a region of type B
“It's difficult explaining to the Lehmann7
I fear it lacks Bayes' simplicity.”
There's no haters like Bayes' haters
They spit when they see a prior
Be careful when you offer your posterior
They'll try to kick it right through the door
But turn the other cheek if it is not too sore
Of error they may yet tire
REFRAIN
There's no theorem like Bayes' theorem
Like no theorem we know
Critics carp at Bayes' hesitation
Claiming that his doubts on what he'd done
Led to late posthumous publication
We will explain this to everyone
When Bayes got up to Heaven
He asked for an interview
Jehovah quickly told him he had got it right
Bayes popped down earthwards at dead of night
His spectre ceded Richard Price the copyright
It's very strange but it's true!
As I have said, as result of the differences in the department, Brian left the University in 1983 and with his wife, Laurie, started a quality improvement consulting firm called Joiner Associates. The company was tremendously successful, and soon had a world-wide reputation. But by the time the firm was a decade old, Brian became concerned that none of the businesses that he advised had any interest in the pressing environmental issues that were then so much in evidence. In 1996, after attending the World Quality Congress in Japan, he visited China, which he realized was pursuing a resource-devouring industrial strategy that mirrored the one in the United States. To clear his head, he and his twin sons David and Kevin went on a 23-day trek through Nepal, where the people seemed content without “stuff.” When he returned to the United States., Kevin Little, a statistics department Ph.D. who worked at Joiner Associates, gave him Bill McKibben's book, Hope, Human and Wild: True Stories of Living Lightly on the Earth. The book, which describes three communities around the world that addressed environmental problems and required no affluence to develop caring communities, changed his life. Brian and Laurie sold Joiner Associates to the first bidder, and Brian has been a full-time community activist since.8
My present feeling is that statistics should be split into two disciplines. One would be theoretical statistics, and the other might perhaps be called “Technometrics.”9 Courses in technometrics would be required not only for chemists, engineers, and so on, but also for mathematical statisticians. Such courses would concentrate on problem solving using the statistical design and analysis of investigations and avoid specious assumptions such as the independence of sequential data.
Some reservations about the manner in which statistics is currently taught are expressed in the following song I composed for a Christmas party. It was based on Gilbert and Sullivan's I Am the Very Model of a Modern Major General from the Pirates of Penzance:
I am the very model of a professor statistical
I understand the theory both exotical and mystical
The logic of my argument it is that matters most to me
My chance of making erro
rs is exactly what it's s' posed to be
I relentlessly uncover any aberrant contingency
I strangle it with rigor and I stifle it with stringency
I understand the different symbols be they Roman, Greek, or Cuneiform
And every distribution from the Cauchy to the uniform
Chorus
And every distribution from the Cauchy to the uniform
And every distribution from the Cauchy to the uniform
And every distribution from the Cauchy to the uni-uni-form
With derivations rigorous each lemma I can justify
My every estimator I am careful to robustify
In short in matters logical, mathematic, idealistical
I am the very model of a professor statistical
Chorus
In short in matters logical, mathematic, idealistical
He is the very model of a professor statistical
I am the very model of a professor statistical
My art it is immaculate and thoroughly puristical
Judge me by my inner soul and not by my external face
Understand my mind's away in reproducing Hilbert space
With repetitious pleasantries my students all must learn to live
For my wit's one hypothesis to which there's no alternative
I never stoop to folly nor to action reprehensible
I always state assumptions whether ludicrous or sensible
Chorus
He always states assumptions whether ludicrous or sensible
He always states assumptions whether ludicrous or sensible
He always states assumptions whether ludicrous or sensi-sensi-ble
My manner it is modest and not the least hysterical
My errors they are normal both elliptical and spherical
In short in every aspect whether specious or realistical
I am the very model of a professor statistical
An Accidental Statistician Page 13