Wisconsin at Age 50

Stephen M. Stigler, 3 June 2010

It is a great pleasure to join in this celebration of a birthday - the 50th anniversary of the birth of one of the nation's most distinguished departments of statistics. Like most of you, I have felt like I was returning to one of my most important roots today. I arrived here in the Fall of 1967, fresh from four years at Berkeley. I cannot say I knew what to expect - I had met a few people on my interview trip that February, and they seemed nice. The department was newly relocated in a splendid new building, and it seemed a little odd that they were speaking of their previous location over a drugstore as "the good old days." But there was a warm feeling of belonging and a genuine interest in the science of statistics. Within days I was teaching my first graduate course, a course that soon shrunk to a single student, Hermann Habermann. I cannot tell you I foresaw his future role as a mover and shaker in national and international statistics, much less can I claim that I helped shape it, but the average quality of student in that class was high, as it was in nearly all my classes.

I was also a student here. I sat in on George Box's courses, and I frequently attended the Monday night "Beer and Statistics" consulting seminar at his home. George was working on the Box and Jenkins book at the time, and, as I recall, every problem he treated, no matter what the subject, no matter what the form of the data, was treated as a time series problem. It was a valuable lesson: in the hands of a master, any tool will do! I remained with the Department for 12 years and still regard it with warmth - it was for me a postdoctoral education. Consider me as a member of the class of 1979!

I could go on with such reminiscences, and will later to a small degree, but I believe I was invited to speak in a historical vein, a vein I have increasingly cultivated since I got tenure. And this being a scientific conference, I feel compelled to say something new to an audience that has already today heard several talks on the history of the department. But there is one corner of Wisconsin's history that has gone untouched in those talks.

Most people and all Statistics Departments believe the world began at their birth. Of course that is not literally true, but it is all we remember, and so it may seem to tell the whole story of what we are and what we do. I wish to dispel that belief, and tell you about statistics at Wisconsin BB (Before Box). In the process I will underscore the great achievement George and those who followed have accomplished.

I do not know the true beginning of statistics at Wisconsin. Perhaps it was the establishment of a meteorological station in the 1850s in a building that was soon nicknamed "Old Probabilities". Perhaps it was in the Religion Department in the 1870s when they presumably had a course in Cross-validation. I do know that statistics in the Wisconsin curriculum predated 1890, when George Comstock, the Director of Washburn Observatory published a textbook based upon his course on the Method of Least Squares, a course targeted to "students of physics, astronomy, and engineering." The first page of his book shows a density function as estimated by data, I suppose using some penalized likelihood algorithm. The second chapter was called "The Distribution of Residuals," and the book reads like an early draft of Draper and Smith. George Comstock has not been forgotten; one of the moon's craters is named "Comstock" in his honor. But it is on the far side of the moon, I am afraid.

The next few decades were relatively quiet. Probability was taught in the mathematics department, including a course offered by Warren Weaver, who was on the faculty from 1920 to 1932. Henry Scheffé took Weaver's course and achieved later eminence at Berkeley with his classic book The Analysis of Variance - Scheffé received both BA and PhD degrees from Wisconsin and taught here from 1935 to 1937. But he only taught pure mathematics.

Where was Wisconsin during the years of exciting development in England by Fisher and Neyman and Pearson and so many others? It is true that there was some activity here in the agricultural area, particularly in Agronomy, and Churchill Eisenhart taught here from 1937 for a while. But if you really wish to understand why it was only in 1960 that Wisconsin fully opened its arms to modern statistics, you must look back to the curious events of 1941, when there was a bold attempt to create a statistics program, and it crashed and burned.

The story begins in June 1940, when the University of Wisconsin invited Milton Friedman to visit the Economics Department with the specific charge of strengthening their statistical offerings. Friedman is of course much better known now as an economist. He won the Nobel Prize in 1976 for his work on monetary theory and the consumption function, and in addition to his scientific work he influenced and still influences policy as much as any economist in the 20th century. But in 1940 Milton was more of a mathematical statistician than economist. He had published his rank test for two-way ANOVA in 1937. It was possibly at the suggestion of his thesis advisor Harold Hotelling that Friedman came to Wisconsin in Fall 1940 as a visiting Lecturer. Harold Groves of the Economics Department played a key role in getting the Dean and President on board. While here, Milton produced a detailed report on the dismal situation faced by any Wisconsin student interested in studying statistics, and he included as one suggestion the creation of a new Department of Statistics. I quote the final sentence:

"A student cannot secure training at the University of Wisconsin sufficient to qualify him to teach advanced statistics or to do independent work in the field of statistical methods. Even if he takes all the work offered he will be but indifferently qualified to do research involving the application of modern statistics."

The administration, particularly Dean Sellery, was greatly impressed, and in the spring they offered him a position as Associate Professor with tenure at the princely salary of $3500 a year. He was charged to come and implement the program he had outlined. He indicated he would accept, but before the appointment could be approved by the Regents, a dispute broke out that reached the public press. A group of senior professors in the Economics led by Edwin Witte demanded that the Regents cancel the offer, claiming that as able as Friedman might be, he was too young for such an appointment. Witte was a labor economist and a statistician of the old school; one wag quipped that his name was half-accurate. In a switch from the stereotype, the old guard took to the streets to protest the young interloper. On the other hand, the students in the economics department backed Friedman, and the students picketing included such future leading liberal figures as Walter Heller (later Chairman of the Council of Economic Advisors under Presidents Kennedy and Johnson) and Joseph Pechman (later at Brookings).

The headlines in the Daily Cardinal tell the story: On May 15, "Econ Profs Oppose Promoting Lecturer; Youth, Inexperience are Cited in Protest." On May 16: "Milton Friedman Becomes Object of Controversy." May 27: "Faculty Probe Asked by [Governor] Heil." The headlines in The State Journal and The Cap Times were not much different. The dispute simmered over a couple of weeks, and on June 2, 1941, Friedman decided he did not want to come where he was not wanted, and he withdrew. The Daily Cardinal headline on June 5 read, "Friedman Drops Job Application." Friedman returned to a position at the National Bureau of Economic Research in New York at a salary of $5000. Within the next few years Friedman would play a pivotal role in a number of statistical developments, including the introduction of sequential analysis: He invented the first sequential procedure for testing binomial data, a test he informally christened the Super Colossal Test because it outperformed the Uniformly Most Powerful Test. And he gave Wald the briefing that produced Wald's major theoretical contribution to this subject.

Friedman emerged intact from this debacle, but Wisconsin did not. Once burned, the Wisconsin administration retreated, and it would be nearly 20 years before they dared a second attempt. I mention all this to highlight the difficulties that George Box faced and overcame in founding the Department. Where Friedman may have been too direct and brash, George was skillful in negotiating with all manner of potential partners, and in demonstrating the great value statistics could bring, without exciting the narrow-minded defensive reaction the senior economists showed in 1941.

We celebrate that founding today, and we celebrate all the people who have played vital roles in this exciting period. Not all are with us today - Bill Hunter, Jerry Klotz, John Gurland, Jim Hickman, John Van Ryzin, Jerry Senturia, Greg Reinsel, come quickly to mind. Quite a few of my colleagues from that first decade are thankfully here today - George Tiao, Norman Draper, Bernie Harris, Grace Wahba, Rich Johnson, George Roussas, Bob Wardrop. Others have escaped to the four corners of the world - Irwin Guttman, Gouri Bhattacharyya, Don Watts, Jim Bondar, John Crowley, Bob Miller, Brian Joiner. There were important visitors like Stu Hunter, Svante Wold, Graham Wilkinson, and Don Behnken.

We will fondly remember all these colleagues and all our stellar students whose success we can be justly proud of even if we cannot claim credit for all their great achievements. I, for example, am immensely proud to have been associated with Alan Agresti, LJ Wei, and Hermann Habermann, among many others. We remember the student-faculty Christmas skits. We remember the office staff - Mary Esser, Mary Ann, Wanda, Candy, and all their successors.

I propose a toast to the entire Wisconsin Statistics Department family!