History of Statistics at UCLA

Statistics at UCLA — 1939 to 1998
Contributed by Don Ylvisaker, Professor Emeritus

Genesis
It could be said that Statistics arrived formally at UCLA when the Mathematics Department hired Paul Hoel in 1939. Hoel had received his Ph.D. from the University of Minnesota in 1933, writing a thesis on approximation theory under Dunham Jackson. He had taught at Rose Polytechnic Institute and Oregon State, then came to UCLA and stayed on until his retirement some 32 years later.

Exactly how and when Hoel came to Statistics is not clear, but his progress in this direction was no doubt aided by spending the 1936-37 year in Oslo under Ragnar Frisch, as a Fellow of the American-Scandinavian Foundation. Already by 1940, it was noted by his department chair that Hoel had been “especially clever in handling a number of delicate matters concerning the statistical work on this campus” and that it seemed he would be “able to coordinate the various courses in statistics in such a way that a logical program in this field” could be offered at UCLA.

In early 1943, Hoel’s interests were such that Sam Wilks asked him to join in a National Research Committee project at Princeton as a “mathematical statistician”. He did spend some time in Colorado Springs during the war years “digging up data and attempting to analyze it in simple terms”. His contributions were apparently notable, especially as to “increasing the accuracy of heavy and very heavy bombing”, but he missed his family and Los Angeles and soon returned to campus.

Years spent on various approaches to teaching statistical theory to students without advanced calculus led him to the writing of his classic book “Introduction to Mathematical Statistics” that came out in 1947. His idea was “to give the beginner a fairly broad introduction to both classical large-sample and modern small-sample methods” while presupposing only elementary calculus. The success of this vision can be seen in the large number of texts produced by others in following his lead, the translation of the book into several different languages, and in the fact that the sixth edition of the book remains in print in 2011, and in use.

A main research interest of Hoel’s followed neatly from his graduate training in approximation in his 1950’s focus on continuous design theory. He was a fine teacher at all levels, and his output of 12 Ph.D. students as of 1960 was then “by far the record of anyone else in the Mathematics Department”, this in “competition” with such prolific producers as Magnus Hestenes and Leo Sario. Of special note, Bob Jennrich obtained his Ph.D. under Hoel in 1960 and returned to the Statistics faculty at UCLA in 1962. Though formally retired from the university in 1971, Hoel maintained a presence around the department for an additional twenty years.

There were fresh developments in the 1950’s with other statisticians arriving on the local scene. Alexander Mood and George Brown had each gotten PhD’s at Princeton in 1940, under Wilks, had taught in the Statistics Department at Iowa State, and had subsequently come west to the Rand Corporation. Brown was known for his knowledge of computers and came over to UCLA as a professor of administration in 1952, especially in connection with an IBM project. He left in 1966 to head the Graduate School of Administration at UC Irvine. After five years at Rand, Mood formed a consulting firm, the General Analysis Corporation, later CEIR. He went to UC Irvine in 1967, following Brown. During his career, Mood achieved considerable distinction, serving as the President of the Institute of Mathematical Statistics (1957) and President of the Operations Research Society of America (1963); he was given the Wilks award by the American Statistical Association in 1979.

Mood wrote the long-running standard text “An Introduction to the Theory of Statistics” with Franklin Graybill (1963). Joint work of Mood and Brown included a median test, a nonparametric test that is now all but supplanted by the Mann-Whitney Wilcoxon test of the equality of two distributions. Another noteworthy accomplishment of the two of them was to lure Ray Jessen away from the Department of Statistics at Iowa State to join with them in their Los Angeles consulting firm, this in 1957. Jessen, a noted sampling expert, taught at UCLA beginning in 1958, and was a faculty member in Management from 1962 until his retirement. The book “Basic Statistics for Business and Economics”, jointly written with Paul Hoel, was published in 1971.

Another large statistical figure, Will Dixon, came to UCLA from the University of Oregon in 1955, with a joint appointment in the Department of Preventive Medicine in the School of Medicine and the Biostatistics Division of the School of Public Health (its first member). His 1951 book, “Introduction to Statistical Analysis”, written with Frank Massey who later joined Dixon at UCLA, had been well-received and widely used; in its 1984 fourth edition, the book is still standardly cited.

At UCLA, Dixon quickly grew the Biostatistics Division in Public Health and, in 1967, initiated the Biomathematics Department within the School of Medicine. In particular, he did not think the development of Statistics through the Mathematics Department was possible, or even healthy, and was generally dismissive of mathematical statisticians as divorced from applications and problems with data. Aside from administration, his main interest was in computing, and this led to the 1962 emergence of BMDP as the first statistical software system distributed worldwide, followed by such as SPSS and SAS. Dixon would go on to be a major force in the School of Public Health and the School of Medicine at UCLA for more than thirty years.

Growth
In the Mathematics Department, the growth of Statistics began with the hiring of Tom Ferguson in 1956. Ferguson had done his Ph.D. degree at Berkeley under LeCam, and his coming to UCLA doubled the then teaching power of the nascent (Probability and) Statistics group. Ferguson’s statistical interests were broad, with a particular focus on game theory, and his active tenure at the school would come to exceed 50 years, and long past his formal retirement in 1994.

Ferguson had a large impact on the teaching aspect of the department. Thus his 1967 book, “Mathematical Statistics: A Decision Theoretic Approach” was routinely used in the beginning course for graduate students at UCLA, and widely used elsewhere. His 1996 book “A Course in Large Sample Theory” spelled out the material that had been used in the latter part of the introductory course for many years, providing what students in Statistics (and from Biostatistics) needed, and used very effectively, for the qualifying exam in probability and statistics (until degree requirements were changed in the late 1980’s, students with an interest in pursuing Statistics also had to pass qualifiers in real and complex analysis, and in an elective area). As to supervision of students, Hoel and Ferguson created a legacy of upwards of 30 Ph.D. students in the Mathematics Department.

Through the early 1960’s, the group in the department grew to include Leo Breiman, another Berkeley product, Bob Jennrich, Charles Stone, and Sid Port. The last two came as a package in 1964, Stone a Stanford grad and Port with a degree from Northwestern. The overall focus was very much on the theoretical side of the subject with Breiman, Port, and Stone all working mainly in probability; in a separate direction, Jennrich was involved with statistical computing and already offering courses in the subject. Breiman resigned after seven years on the faculty to go into private consulting, and was subsequently on the Berkeley Statistics faculty. In the late 1960’s Tom Liggett and Don Ylvisaker joined the Probability and Statistics group; they were Stanford products in Mathematics and Statistics, respectively.

There came to be a fairly stable period, beginning in the late 60’s, in which this group interacted on a consistent basis, watched over the interests of Probability and Statistics in the Math Department and, especially, its teaching mission. There were two tracks in the upper division Probability and Statistics course: a two-quarter course, based on Hoel’s book, and a year-long course for which Hoel, Port, and Stone wrote a three-volume set of introductions to probability, statistics, and stochastic processes. These last, along with Hoel’s earlier work, were widely adopted texts that lasted well.

There was stability at the graduate level as well during this time, enrollments were relatively healthy, and a good number of Ph.D.’s were graduated. There was, as well, a fair amount of cooperation with Biostatistics at the time. Students in that program were obliged to take the beginning graduate courses in Statistics, and to take the corresponding qualifying exams over this material. It was then the case that most doctoral candidates in Statistics or Biostatistics had committees that contained faculty from the other group, and to good effect. The Biostat faculty included, in the early days, Dixon and Massey, Abdelmonem Afifi, Jean Dunn (a student of Hoel’s from the mid-50’s), and Robert Elashoff.

Some significant publications of the research being done during those years are recounted here with a 2011 citation count in parentheses: the 1969 paper of Jennrich, “Asymptotic properties of non-linear least squares” (868); “A Bayesian analysis of some nonparametric problems” by Ferguson (1,820); Charles Antoniak’s thesis written under Ferguson, “Mixtures of Dirichlet Processes, with Applications to Bayesian Nonparametric Analysis” (889); Stone’s 1977 paper “Consistent nonparametric regression” (1,029). Aside from plentiful citations, a common feature of the works is that each portended important developments in statistical thinking: a growing call for non-linear modeling, a rapid increase in the use of Bayesian inference, and the rather sudden proliferation of smoothing problems (and methods).

Stone, whose interests had become more statistical over time, left the group in 1981 for a position in Berkeley’s Statistics Department. Efforts to replace his FTE foundered until Ker-Chau Li was enticed from Purdue in 1984; Li’s degree was in 1981, under Jack Kiefer at Berkeley, but an earlier attempt to recruit him to UCLA had not been successful.

The late 70’s and early 80’s marked an uneasy period at UCLA, if not elsewhere. Computing was fast becoming a serious rival to mathematics as a force in Statistics, but how this impacted the local scene was not at all evident, since resources were not available within the Mathematics Department if, indeed, clear goals could have been formulated. Thus, students were not provided with computer access and, aside from Jennrich, there was little in the way of classes or labs devoted to statistical computing. It would be a good fifteen years before a new statistics department could do a proper job of responding to the computing revolution.

Graduate enrollments went down during these times, reaching a nadir in the late 1980’s. The demanding set of qualifying exams in Mathematics began to look even more arcane for statistical purposes, and the pool of potential students shrank in the face of new opportunities to go into computer science.

Change
In 1984, the Committee for Advanced Quantitative Methods in the Social Science Division of the University recommended to the dean of that division that, among other things, funding be obtained to appoint at least six faculty members in the various departments of the division to serve in an interdepartmental program in Statistics. The program would pay direct attention to individual needs for quantitative training in such departments as Economics, Sociology, and Geography. In the face of such a development, and news to them, the statisticians in Mathematics felt they should have a say in any hiring process, at the very least.

The thought of forming a separate department, ever in the background, thus came to the fore. Some discussion of the new discipline had taken place in the 1950’s but Hoel, never an administrative type, was not the strong force in this regard that he would had to have been. The rapid spread of statistics departments across the country, an expansion dating to the 1930’s and accelerating after WWII, had reached a state that Cornell and UCLA were commonly cited as the remaining major holdouts. Even aside from Hoel, others in the group had not been moved to rebel against their treatment in Mathematics, individually or collectively, but with the subject now moving in the direction of computing and applications and away from mathematics, the time had evidently come to take concrete action.

A letter setting out the need for a separate statistics department, and signed by a several statisticians on campus, was sent to Ray Ohrbach as Provost of the College and, subsequently, Jennrich and Ylvisaker met with the Dean of Physical Sciences, Clarence Hall. This meeting lasted just long enough to determine that the thought of doing anything about Statistics was a new and foreign one to him – he could not afford to do anything, and saw no reason for it if he could. He would come later to grant some grudging administrative support for a cost-free Statistics Division within the Mathematics Department, and later still to a supportive position on Statistics.

In the Social Sciences Division, Dean Sears took early note of the committee proposal of his faculty, and quantitative people were recruited to the campus, beginning with Kazuo Yamaguchi to Sociology in 1986 and Douglas Rivers to Political Science in 1987. A search for someone to lead a Social Statistics program, begun in 1986, culminated in the 1987 appointment of Jan de Leeuw from the University of Leiden to a position as Professor of Psychology and Mathematics. Other additions to the program followed and, of special note, Richard Berk came to Sociology in1988. Starting in 1987 the Advisory Committee for the Program in Social Statistics, led by de Leeuw, held meetings that included statisticians from around the campus. The committee looked at the quantitative training of social science graduate students, then the undergraduate teaching of Statistics across the campus, with one of its conclusion being that these issues should be dealt with in a separate statistics department.

In the Mathematics Department the Division of Statistics, with Ylvisaker as Head, was administratively blessed by Dean Hall in 1986. The advantages of this came in the separate listing of undergraduate courses in Statistics, and in the ability to recruit and admit graduate students to a program in which degree requirements could be made distinct from those in Mathematics; the downside was that resources would still be located in, and controlled by, that department. Thus, the group had achieved non-financial freedom. Recruitment of students to the Division was, by 1990, being actively pursued, and a newly configured graduate program had been put in place for them. The research program of the Statistics faculty in the new division remained active through this period despite administrative goings-on. De Leeuw, while soon acquainted with all college goings-on, was actively involved with his own research and engaged with a bevy of Dutch visitors, students and colleagues, in this. Li’s sliced inverse regression approach to dimension reduction and Ylvisaker’s work on designs for computer experiments were influential during this period in areas freshly renewed by increases in computing power.

Thanks in no small measure to de Leeuw’s leadership and his ability to operate effectively across departmental lines, communication between the statisticians in the College had become commonplace by the late 1980’s where, for some time, this had been sporadic at best. In 1991, space was found in Bunche Hall that allowed faculty of the Social Statistics program some cohesion, independent of originating department. The major tasks faced were to to consider the teaching of introductory statistics as it was being practiced around the campus and how it might be shaped for the future, and to ensure that the administration did not relegate statisticians to the teaching role, that the graduate program and faculty research were in every sense worthy of departmental status in the University. Toward the first of these tasks, Dean Sears appointed the Advisory Committee for Introductory Statistics Teaching, the report of which reached the then Dean of Social Sciences, Scott Waugh, in the Spring of 1993. The Administration was witness to the number of introductory statistics courses across the campus, and could count the students with Statistics requirements, but the importance of Statistics as a discipline was not so apparent by the early 1990’s if one had a background in such as History, Physics, or the Geological Sciences.

A landmark in the recognition of Statistics at UCLA came in 1991 in the form of a large NSF grant in support of its reorganization and the coordination of its teaching and research. These funds also supported graduate students in Economics, Mathematics, Biomathematics, and Sociology, and allowed the acquisition of workstations and servers to support graduate teaching and distributed statistical computing. The Statistical Consulting Service was begun in 1992 with principal consultants Berk, de Leeuw, and Ylvisaker, and ties with other faculty; it provided students with some financial help as well as first-hand experience in working with clients, and it financed computing equipment. As space became available for consulting, there was substantial work done for the LA Taxicab Companies, the VA Hospital, the California Prison System, Access Services, and the California Lottery, among others. Rob Gould, a 1994 hire in Mathematics, was active in these efforts, as was Vivian Lew for her computing expertise; these two were fine teachers as well, and would go on to long-term positions in the Statistics group.

In the early part of the 1990’s the reality of the situation was that statisticians were active as group, were supporting themselves and the university’s mission rather well, but basically running in place in their quest for administrative recognition and departmental status.

And then a miracle happened.
In 1993, Dean Hall initiated an Interdivisional Program in Statistics, with support from the Deans of Physical, Life, and Social Sciences, as well as the Provost of the College. It was through such standing that space in the Math Sciences Building, not far from present-day Statistics Division faculty offices, became available in 1994. This space housed de Leeuw, Berk, and statistical consulting, and shortly was a hubbub of activity. At long last, the proximity of these offices and those of the statisticians in Mathematics provided the coalition of statisticians in the College something akin to a departmental atmosphere.

Gestation
An early retirement program advanced by the University in 1993-94 that brought about the retirements of Ferguson, Jennrich, Port, and Ylvisaker at the end of the school year, turned out to be a mixed blessing for Statistics. The fact that no new FTE in Mathematics were to go to Statistics meant that the Division of Statistics would consist of Li plus one-quarter of de Leeuw. Statistics would thus be relegated to the sidelines, its teaching at the introductory level would be done by mathematicians and, at the graduate level, would be dependent on faculty recalls.

In late September of 1994, Berk, de Leeuw, and Ylvisaker met with the Brian Copenhaver, recently Provost of the College, and Roberto Peccei, the new Dean of Physical Sciences, to argue in the light of these developments for resources, and preferably through a statistics department on campus. The Provost and Dean were unprepared for such a request (famously likening the departmental thought to the idea of, say, three physicists coming to them to express a need to set themselves apart from their own department). What did result from the meeting was an agreement, often considered but never before agreed to, to hold an outside review of Statistics at UCLA. Larry Brown, who had recently spent a visiting year at UCLA, agreed to serve as Chair.

Letters on the status of Statistics written by Provost Brian Copenhaven to his deans, dated 8/23/94, and to Dean Peccei on 6/29/95 had markedly different tones, reflecting the intervening events. The first of them referred to a plan for “Introductory Statistics Teaching” in the College that, to the Provost, had a considerable cost and no clear benefit to instruction; as well, it gave no indication that the Provost understood that Statistics had broader goals than undergraduate teaching. The second letter gave a clear call for a Center for Interdisciplinary Instruction that, in its other aspects, fell in line with a visiting committee’s recommendation to establish a Statistics Department in the Division of Social Sciences. Importantly, the Provost made a commitment to present and future FTE in the unit, and otherwise looked to ensure its funding and growth.

The committee report, filed in April of 1995, had recommended that a new statistics department be formed in the Division of Social Sciences. The Provost agreed with this in his 1995 letter, being especially mindful of the associated teaching mission, but argued that it should begin, if not end, as a Center for Interdivisional Instruction. This would involve setting up an organization in the Division of Social Sciences with designated, and growing, faculty, and appropriate funding from the Divisions of Life, Physical, and Social Sciences. The structure set out by the Provost would not survive, but Statistics was no longer to be viewed as a stepchild by the administration.

By 1996, the appropriate location for the Statistics group had come to be thought of as the Division of Physical Sciences rather than the Social Sciences. In some part, the Social Sciences Dean may have lost interest in being the center of these developments and, in large part, it was due to the unexpectedly enthusiastic support of Dean Peccei. Thus, for example, he was the one who set up meetings with the faculty of the various departments in the College over their introductory courses, meetings that left little room for the status quo in the presence of the wide array of courses to be offered by the Statistics group. Indeed, considerable work was needed during this period of time in order to frame the statistics courses that would be the foundation of the new curriculum, but this could be done in the belief that a full-fledged department in Physical Sciences, while never guaranteed, would likely result.

These were eventful years in other respects for the statistics group. Wing Wong came to UCLA in January of 1997, already heavily engaged in biology and genetics research. Despite administration overtures, he left for Harvard in 2000 and subsequently went to Stanford. Jianquin Fan was another faculty addition, from 1997-2000, this between sojourns in Hong-Kong and prior to going on to Princeton. Considerable activity was evident during the late period of the Statistics Division and, in particular, a number of graduate students finished their degree programs in Mathematics.

The Proposal for a UCLA Department of Statistics, under the signatures of Berk, de Leeuw, and Ylvisaker, carried a date of April 16. 1997. On April 1, 1998, following external reviews, the Coordinating Committee on Graduate Affairs approved proposed M.A. and Ph.D. degrees in Statistics. The CCGA added recommendations that four faculty be added in the following three years, and three more within five years, and that increased funding be made available to the program. In effect, the Statistics Department had been brought to life.

In this manner, the era of Statistics in the Mathematics Department came to an end after close to 60 years. During that period, the subject had gone from academic infancy to a mature and fast-moving discipline, enrollments now growing rapidly in response to the data explosion and the lure of the new computer age. The Mathematics group’s legacy included some 65 Ph.D. graduates, alongside countless others, given a training that would leave a sizable impact on their careers.