aa r X i v : . [ s t a t . O T ] O c t Statistical Science (cid:13)
Institute of Mathematical Statistics, 2013
A Conversation with Stephen E. Fienberg
Miron L. Straf and Judith M. Tanur
Abstract.
Stephen E. Fienberg is Maurice Falk University Professor of Statistics andSocial Science at Carnegie Mellon University, with appointments in the Department ofStatistics, the Machine Learning Department and the Heinz College. He is the CarnegieMellon co-director of the Living Analytics Research Centre, a joint center betweenCarnegie Mellon University and Singapore Management University. Fienberg receivedhis hon. B.Sc. in Mathematics and Statistics from the University of Toronto (1964),and his A.M. (1965) and Ph.D. (1968) degrees in Statistics at Harvard University.He has served as Dean of the College of Humanities and Social Sciences at CarnegieMellon and as Vice President for Academic Affairs at York University in Toronto,Canada, as well as on the faculties of the University of Chicago and the University ofMinnesota. He was founding co-editor of
Chance and served as the Coordinating andApplications Editor of the
Journal of the American Statistical Association . He is one ofthe founding editors of the
Annals of Applied Statistics , co-founder and editor-in-chiefof the new online
Journal of Privacy and Confidentiality and founding editor of thenew
Annual Review of Statistics and its Application.
He has been Vice President ofthe American Statistical Association and President of the Institute of MathematicalStatistics and the International Society for Bayesian Analysis. His research includesthe development of statistical methods, especially tools for categorical data analysisand the analysis of network data, algebraic statistics, causal inference, statistics andthe law, machine learning and the history of statistics. His work on confidentiality anddisclosure limitation addresses issues related to respondent privacy in both surveysand censuses and especially to categorical data analysis. He is the author or editor ofover 20 books and 400 papers and related publications. His 1975 book on categoricaldata analysis with Bishop and Holland,
Discrete Multivariate Analysis : Theory andPractice , and his 1980 book on
The Analysis of Cross-Classified Categorical Data areboth citation classics. He served two terms as Chair of the Committee on NationalStatistics at the National Research Council (NRC) and is currently co-chair of theNAS-NRC Report Review Committee. He is a member of the U.S. National Academyof Sciences, and a fellow of the Royal Society of Canada, the American Academy ofArts and Sciences, and the American Academy of Political and Social Science, as wellas a fellow of the American Association for the Advancement of Science, the AmericanStatistical Association, the Institute of Mathematical Statistics, and an elected memberof the International Statistical Institute.The following conversation is based in part on a transcript of a 2009 interview fundedby Pfizer Global Research-Connecticut, the American Statistical Association and theDepartment of Statistics at the University of Connecticut-Storrs as part of the “Con-versations with Distinguished Statisticians in Memory of Professor Harry O. Posten.”
Miron L. Straf is Deputy Executive Director for SpecialProjects, Division of Behavioral and Social Sciences andEducation, The National Academy of Sciences, 500Fifth St. N.W., Washington DC 20001, USA e-mail:[email protected]. Judith M. Tanur is Distinguished Teaching Professor Emerita, Department of Sociology,State University of New York Stony Brook, PO Box280, Montauk, New York 11954, USA e-mail:[email protected]. M. L. STRAF AND J. M. TANUR
Fig. 1.
Miron Straf, Steve Fienberg and Judy Tanur at theUniversity of Connecticut, October, 2009.
MS:
So, Steve, how is it that you came to becomea statistician?
SF:
It’s actually a long story, because when I wasin high school and entering university, I didn’t evenknow that there was such a field. I was good at math-ematics and I went to the University of Toronto,which was in my hometown—that’s where the beststudents went if they could get in. I enrolled in acourse called
Mathematics , Physics , and Chemistry .It was one of the elite courses at U of T, and dur-ing the first year, as I went through my chemistrylabs, I never succeeded in getting the right resultwhen I mixed the chemicals up in the beakers; I re-alized chemistry wasn’t for me, and so the secondyear I did only math and physics. Then there werethe physics labs, and I could never quite get the ap-paratus to work properly to get what I knew wasthe correct answer. I still got an A in the physicslab, because I could start with the result and workbackward and figure out what the settings were andthings like that; but it was clear to me that physicswasn’t for me as a consequence. So that left me withmathematics, and it was in the second year that wehad a course in probability. So I was being gently in-troduced to statistical ideas. Then in my third yearthere was a course in statistics that was taught byDon Fraser, and he was terrific. His course was arevelation, because I didn’t know anything about This is an electronic reprint of the original articlepublished by the Institute of Mathematical Statistics in
Statistical Science , 2013, Vol. 28, No. 3, 447–463. Thisreprint differs from the original in pagination andtypographic detail.
Fig. 2.
Steve as a Toddler in 1940s in Toronto. statistics coming in. Don followed the material inhis
Introduction to Statistics book and he beganwith probability theory and he brought into playgeometric thinking throughout. When he got to in-ference, it was like magic. Of course, in those daysDon did what was called “fiducial inference”—hecalled it “invariance theory” and later “structuralinference”—where you went suddenly from proba-bility statements about potential observables givenparameters to probability statements about the data.I recall the old cartoon by Sydney Harris that peoplelike to reproduce of the two scientists pointing to ablackboard full of equations, and one of them pointsto an equal sign and says, “And a miracle suddenlyoccurs here.” That’s sort of what happened in Don’sclass. He was a great lecturer, he was friendly withthe students, and it was very clear that statistics wasa really neat thing to do. Thus, in my fourth year Itook three classes involving statistics and probabil-ity and then applied to graduate school in statistics.The rest, as they say, is history.
MS:
So it was mathematics by elimination andstatistics by revelation. Let’s go back a bit. Whendid you discover that you had an aptitude for mathe-matics and statistics? In elementary school? Or highschool?
SF:
Not at all. In those days statistics never showedits face in the K-12 curriculum—this was before
CONVERSATION WITH STEPHEN E. FIENBERG Fig. 3.
Steve at Camp Tamarack, near Bracebridge, Ontarioin 1952.
Continental Classroom . Actually it was K-13 inToronto where I was born and raised. They got rid ofgrade 13 only decades after I was in school. At anyrate, although my mother thought I was genius—don’t all mothers think that about their children—I don’t have any memory of being anything otherthan just a good student. I was very good at whatpassed for mathematics, but even through highschool I don’t think I was truly exceptional, and, be-sides, we did pretty elementary stuff—algebra, Eu-clidean geometry, and then in grade 13 we had trigo-nometry. As I reflect on those days, I was goodat mathematics, but certainly not precocious andI only took standard high school math and witha heavy component of rote and repetition. By thetime I got to grade 13 I was at the top of my class,however, and in the province-wide exams at the endof the year I was No. 2 in my school. But I alsoplayed oboe in the orchestra and band, and drumsin the marching band, as well as participating in sev-eral other extra-curricular activities. So math wasn’tmuch of a preoccupation and I didn’t know whatstatistics and probability were all about at all.
JT:
So that explains your broad early work inmath, physics and chemistry as a kind of omnibuscourse rather than going directly into math or statis-tics. So after your undergraduate work at the Uni-versity of Toronto, you applied to graduate school;where did you apply and where did you end up go-ing? Continental Classroom was a series of television “course”broadcasts by NBC on a variety of college-level topics in theearly 1960s. Fred Mosteller taught the course on Probabilityand Statistics during 1960–1961
Fig. 4.
Steve with Don Fraser and Nancy Reid at a confer-ence on the occasion of Don’s 75th birthday, June 2000.
SF:
Well, at the University of Toronto there hadactually been many people to go into Statistics fromMP&C. Don Fraser was perhaps the first, but thenthere were Ralph Wormleighton, Art Dempster andDavid Brillinger—they all went, by the way, to Prince-ton. The year before me there was John Chambers,and John had gone to Harvard. I knew John prettywell, and I asked him how it was at Harvard. Heseemed pleased with what he was doing and I didapply to Harvard and was admitted. I also appliedto Princeton, and in their wisdom they didn’t thinkthat I should carry on the tradition from the Univer-sity of Toronto, and that made the decision easierfor me.
MS:
Were you disappointed about not being ad-mitted to Princeton?
SF:
Clearly at the time I was. This was my firstrejection, and it prepared me in a way for what wasto come when I submitted papers for publicationto major journals! But Sam Wilks, who was the keyperson at Princeton with whom I had hoped to work,died in the Spring of 1964, before I would have ar-rived.
JT:
By the time you went to Harvard you werealready married, is that right?
SF:
No, I had met my wife Joyce at the Univer-sity of Toronto when we were both undergraduates.I was actually working in the fall of 1963 in the reg-istrar’s office, and on the first day the office openedto enroll people, Joyce came through. And one ofthe benefits about working in the registrar’s office,besides earning some spending money, was meetingall these beautiful women students passing through.That first day I made a note to ask Joyce out ona date. The next day she came through again, this
M. L. STRAF AND J. M. TANUR
Fig. 5.
Graduation portrait from the University of Toronto,1964. time bringing through another young woman whoturned out to be the daughter of friends of her par-ents. And I thought this was a little suspicious, butauspicious in the sense that maybe I would succeedin getting a date when I asked her. And the next day,she came through again! This time with her cousin!Then I knew that this was really going to work out.And it did. We got engaged at the end of the sum-mer of 1964 after I graduated, but we weren’t mar-ried when I went away to graduate school. In fact,yesterday I was talking to one of the students at theUniversity of Connecticut who was a little concernedabout graduate school; it was wearing her down, andI told her I almost left after the first semester be-cause I wasn’t sure if I was going to make a go ofit, in part because I was lonely. But I did survive,and Joyce came at the end of the first year; we gotmarried right after classes ended, and we’ve beentogether ever since.
MS:
And where were your children born?
SF:
Ah, conceived in various places, born in oth-ers. We believe that Anthony, my older son, wasactually conceived in Scotland, on the vacation wetook just after I graduated from Harvard. He wasborn in Chicago, where I had my first academic ap-pointment, and, indeed, as we traveled across the
Fig. 6.
Joyce and Steve in Portugal for a conference on pri-vacy and confidentiality, 1998. country, from Boston to Chicago, Joyce began ex-periencing morning sickness (all day long), whichdidn’t make for such a great trip. Then Howard wasborn in Minnesota just after we had moved thereand I had joined the University of Minnesota fac-ulty.
JT:
Tell us more about what happened when youfirst arrived at Harvard.
SF:
Well, one of the reasons I went to Harvard isthat they not only gave me a fellowship, but also aresearch assistantship to work with Fred Mosteller.The day after I arrived, I went into the departmentbecause I didn’t quite know what a research assis-tant did, and I went to see Fred (at the time hewas Professor Mosteller, of course—I didn’t learn tocall him Fred until later). Fred was busy, but hisassistant, Cleo Youtz, said he would like to havelunch with me. So I came back for lunch, and wewent to the Harvard Faculty Club. Fred was beingvery courteous, and he suggested I order the horsesteak, a special item on the faculty club menu atthe time. And the horse steak came—I’m not sure ifyou’ve had horse steak—it’s not quite like the kindsof steaks we normally order, it’s a little bit tougher.I cut my first piece of horse steak, I put it in mymouth and started to chew. And then Fred beganto describe this problem to me. It was about as-sessing probability assessors. I didn’t understand athing, and he’s talking away, and I’m chewing away.Then Fred asked me a question, and I’m chewingaway. At this point, he pulled an envelope out ofhis pocket and on the back of it there were thesescribbles. He handed it to me, and I’m still chew-ing because you really can’t eat horse steak except
CONVERSATION WITH STEPHEN E. FIENBERG Fig. 7.
Steve dining with Fred Mosteller at ISI meetings inParis, 1989. in very small bites. It turned out that the scribbleswere notes from John Tukey about this problem. Infact, this was a problem that John and Fred wereworking on for some larger project, and my job wasto translate the chicken-scratches on the back of theenvelope into something intelligible, when I didn’tknow anything about what was going on. I workedat it for a while, and then Fred slowly told me whatJohn’s jottings meant, and the key idea was thatfor assessing probability forecasts, you have to looknot just at the equivalent of means, or the bias inthem (known technically as calibration), but alsoat the equivalent of variability (how spread out theforecasts are). Actually, that was a very importantlesson, although I didn’t have any clue about it inmy first months at Harvard.Over the course of my first fall at Harvard, I dis-covered a paperback book called
The Scientist Spec-ulates : An Anthology of Partially Baked Ideas , editedby Jack Good, with whose work I later became veryfamiliar. In it was a short essay by Bruno de Finettion assessing probability assessors, and de Finetti’sideas went into the technical report I wrote up onthe topic with Fred and John. Fifteen years later, atthe Valencia I Bayesian meeting, Morrie DeGrootand I began to work on the problem and ultimatelywrote three papers on the topic of calibration andrefinement of probability forecasters, heavily influ-enced by that first research exercise with Fred.
MS:
I wanted you to talk about Fred. Fred hasbeen a very influential person in your career, andnot just during your thesis. Maybe you want to tellus a little bit more about how he influenced yourlifek and also how you came to go from Harvard toChicago.
SF:
Well, during that first year I worked on severalproblems with Fred and I wrote up some memos, butthey never quite moved into papers at the time. Fredwas pretty busy, and I got interested in Bayesianinference and multivariate analysis. I had begun totake an interest in Bayesian methods, having partic-ipated as a first year student in a seminar across theriver at the business school run by Howard Raiffaand Bob Schlaiffer. At the time, Art Dempster wasthe person who seemed to be most involved in theseBayesian things and multivariate analysis, so I be-gan to meet with him. In the process of workingwith Art, I met George Tiao, who was visiting theBusiness School with George Box for the year. As aconsequence, George and I wrote a paper togetheron Bayesian estimation of latent roots and vectorsbut it just didn’t look like it was going to be a thesisproblem.The next summer, Fred ran into me in the halland said he had some problems that I might liketo work on. Fred had become deeply involved inthe National Halothane Study at the NRC and, un-like most NRC studies, he and others—Tukey, JohnGilbert, Lincoln Moses, Yvonne Bishop, to name afew—were actually analyzing data and creating newmethods as they went along. The data essentiallyformed a giant contingency table and Fred got meworking on a few different problems that ultimatelycame together as the core of my thesis. In the pro-cess I collaborated on separate aspects of the workwith John Gilbert, Yvonne Bishop and Paul Hol-land. I did most of the work in 1967 and that wasthe summer of “The Impossible Dream,” when theBoston Red Sox won the pennant. I would work intothe wee hours and go to Fenway Park and sit inthe bleachers for the afternoon games. Professionalsports where cheap in those days. We also used togo to Boston Gardens for Bruins and Celtics games.Fred was also a Red Sox fan and he actually got tick-ets for some of the 1967 World Series games. I wasenvious, but when I returned to Boston in 1975 onsabbatical we both were able to get World Seriestickets. I got tickets for game 6 and Fred got themfor game 7!Fred introduced me to lots of other statistical prob-lems. I was also his TA one year, working with Fredand Kim Romney who was in the Social Relationsdepartment at the time. Then the time came to geta job, and Fred said to me, “Where would you like togo?” Things were different in those days, as you willrecall from your days at Chicago. We went through
M. L. STRAF AND J. M. TANUR the list of the best places in the field, at every one ofwhich Fred had a friend. He called up John Tukeyat Princeton, he called up Erich Lehmann at Berke-ley, Lincoln Moses at Stanford and Bill Kruskal atthe University of Chicago. I either got offers withoutshowing up for different kinds of jobs at these placesor I got invited out for an interview. When I was in-vited to interview at the University of Chicago, itjust seemed like a really neat place. All the facultymembers were friendly. The temperature in Januarywas really cold, but I liked everything about the uni-versity from the people to the architecture; it lookedlike a university. Leo Goodman was there on the fac-ulty and he had done work that was directly tied tocontingency table topics in my thesis. Chicago justseemed like a great place to go to, so I did.
JT:
It was there that you first met Bill Kruskaland started being influenced by him?
SF:
Bill Kruskal was the department chair at thetime, and I barely got in the door before he begantalking to me about a slew of different statisticalproblems. . .
JT:
Without horse steak?
SF:
Yes, without horse steak. Bill would just comeand say, “What do you know about this?” And oneof the first topics we actually discussed was politicalpolls. This was the summer of 1968; there was a lotgoing on politically in the U.S., and the
Sun TimesStraw Poll was showing up in the newspaper regu-larly. Two of the key questions were: What was theirreal methodology? How accurate were their predic-tions? I began to save the data from the newspaperreports and work on the question of variability andaccuracy. Then Bill got me to do a trio of televi-sion programs with Ken Prewitt and Norman Brad-burn on a special series that aired at 6 o’clock in themorning when nobody ever watched. But right fromthe beginning, Bill and I interacted; he introducedme to Hans Zeisel in the law school, to people inthe business school, in sociology. It was really hardto trail after Bill, because he was interested in ev-erything in the university and outside, and almosteverything we discussed seemed pretty neat. So, asI launched my professional career at Chicago, I triedto do something similar—not precisely the same asthe way Bill did things—but similar.
MS:
Bill was a real Renaissance man, and I pre-sume you were a recipient of his many clippings fromnewspapers.
SF:
Well, the clippings started when I was in myfirst year—he’s the one that started to give me the
Sun Times Straw Poll clippings. But it wasn’t justclippings. Bill would leave library books for me inmy box; he would go to the library, which was onthe second floor of Eckhart Hall, the building wewere in, and he would browse—people don’t do thattoday—the stacks are closed. He would come back,armed with books, and he would share them withhis colleagues and get Xeroxes of pages. And thiscontinued up through the 1980s. I would always getpackets of different materials from Bill, includingcopies of letters to somebody else that would say:“I hope you don’t mind my sharing this with a few ofmy closest friends and colleagues.” I had this imagethat he was making hundreds of Xeroxes to sendaround the world.
MS:
And before that, carbon paper. So, tell us abit about your life after Chicago.
SF:
The University of Chicago really was a greatplace for me to work. I had a second appointmentin theoretical biology, which was interesting becauseI had never taken a course in biology as a student.And actually it was a very formative experience, be-cause it taught me that I could go into an area thatI had never studied, never learned anything about,and learn enough for me to make a difference in theapplication of statistics. I wrote papers on neuralmodeling, and I wrote papers on ecology; I didn’tdo a lot of genetics, but I read genetics papers andbooks because I included that material in the courseon stochastic processes that I taught. Unfortunately,Chicago wasn’t the safest of places in those days,and Joyce made it pretty clear that she wanted tolive in a place where our children could play in thebackyard by themselves, not under adult supervision100 percent of the time. So I began to be receptiveto conversations with people from the outside, andsoon I was approached by one of my former students,Kinley Larntz, who had just joined the Universityof Minnesota. They were looking for a chair for thenewly created Department of Applied Statistics, aspart of a School of Statistics. So after four years atChicago, I became an administrator as well as re-searcher and teacher.
MS:
Did you work with Seymour Geisser there?
SF:
The School of Statistics was an interestingidea. Minnesota had had a statistics department,and it had run into some problems over the years.The university came up with this plan to reinvigo-rate statistics, and they created the School of Statis-tics. Seymour was the director, and the School wassupposed to have three departments. There was the
CONVERSATION WITH STEPHEN E. FIENBERG Fig. 8.
Judy Tanur, John Bailar, Steve, Henry Block and Jim Press at a conference in Bejing, 1987. old statistics department, renamed as the Depart-ment of Theoretical Statistics, there was the newapplied department that I was chairing, and therewas the Biometry department in the School of Pub-lic Health. But the biometry faculty didn’t reallyseem to want any part in this, and so they resisted,and ultimately the school had two departments plusthe Statistical Center—the consulting center thatwas associated with our department on the St. Paulpart of the Twin Cities campus. Seymour and I in-teracted throughout my eight years at Minnesota,but we never wrote a paper together.
JT:
I want to take you back a little more. Youtalked about these two giant figures who were col-leagues and mentors—Fred Mosteller and Bill Krus-kal. How do you see how they shaped your career,your interests—not only technical, but practical?
SF:
One of the things I didn’t know as a gradu-ate student was how easy it would be to work onand contribute to new problems and new areas ofapplication. The worst fear of a graduate student—well, the worst fear is that they won’t finish theirthesis—the second fear is they won’t have a newidea, and, in fact, 80% of students never publishanything other than their thesis. But Fred was go-ing from area to area: when I arrived at Harvard hehad just published
The Federalist Papers with DavidWallace; while I was there he was leading the effort on the Halothane report; I worked with him evaluat-ing television rating surveys from Nielsen and othercompanies for a national network (that was a con-sulting problem). He just seemed to work aroundthe clock on all sorts of different topics, and so I fig-ured that’s just what a statistician did. It’s funnybecause, in some senses, clearly, everyone didn’t be-have like Fred, as we all know. But that was mymodel! So when I got to Chicago and Bill actedin the same way, and Paul Meier in addition, thatseemed like a natural way for me to do work as astatistician. They seemed to work around the clockon statistics, so I did too.Now Fred liked art; in later years he actually tookup reproducing art and it showed up in his office.When I was a graduate student I went into his of-fice one day and there was a picture by Escher, theDutch artist, called “The Waterfall” and I was verysurprised because I had been introduced to Escheras an undergraduate. Escher’s work showed up onthe cover of a book called,
Introduction to Geome-try , written by Donald Coxeter—the great geome-ter at the University of Toronto. I had three courseson different aspects of geometry from Coxeter. Thisinfluenced some of my thesis research—and I stilldo some geometry—but I also learned about Escherfrom Coxeter! And there was this Escher print inFred’s office which I recognized immediately. Fred
M. L. STRAF AND J. M. TANUR
Fig. 9.
Steve and Seymour Geisser, attending a BayesianWorkshop in Rio de Janeiro, Brazil, summer 1990. told me where he had purchased it, and shortly af-terwards I went off to the store. I still own two Es-cher prints as a consequence, ones that I couldn’tafford to buy today, all because of Fred. Fred andI would occasionally go off to museums, and whilewe looked at the art we would talk about statistics,art and other topics.Both Fred and Bill were Renaissance men andI didn’t know how I would do things in the sameway they did, but it became very clear to me thatjust doing papers in the
Annals and in
JASA wasn’tenough. While I had colleagues whose careers lookedlike that, I thought I should be doing something dif-ferent with my career. I was easily seduced into allthese other activities—and everything was so muchfun. For example, Dudley Duncan, the sociologist,called me one day and asked me if I would join anadvisory committee set up by the Social Science Re-search Council on social indicators in Washington.I hadn’t been to Washington since I was 7 years oldand I went off to this meeting and then spent eightyears interacting with giants in the field of sociol-ogy and survey methods! That experience just rein-forced the way I was using my statistical knowledgein diverse applications.And of course Bill and Fred would just sort ofnudge me once in a while to get things done thatthey cared about deeply. In particular, Fred wantedto see the log-linear model work that his studentshad done for the Halothane study appear in a book.Fred was big on books. And as I left Harvard, hegathered together all the different students who hadworked on different aspects of contingency table anal-ysis—Yvonne Bishop, Dick Light, myself and Paul Holland, who was a junior faculty member, for ameeting at his house. There were also a couple ofother faculty members who sort of disappeared bythe wayside in this enterprise, there were a few moregraduate students—Gudmund Iversen who ended upat Swarthmore, for example—and Fred said, “Weneed to have a book on this.”But we didn’t have Fred’s grand picture in mindand the book didn’t begin to take shape until longafter I had joined the faculty at the University ofChicago. I taught a contingency table course in myfirst year there and it included the first three Ph.D.students I worked with—Tar (Tim) Chen, ShelbyHaberman and Kinley Larntz. Shelby extendedYvonne’s code for multi-way tables and this inspiredhis thesis. I began to use iterative proportional fit-ting on new problems and this triggered a paper onmulti-way incomplete tables and a draft of the firstbook chapter. But then everything progressed ratherslowly, and the book took a full six years to produce.Fred kept pushing the book behind the scenes.One of the things I learned is the time to pro-duce a book goes up as the power of the numberof authors. It would have taken less time if I hadwritten the book myself instead of with Yvonne andPaul. But while we worked at the core of the en-terprise, the three of us had different conceptionsof some materials, and this slowed us down. Fredwas a full partner, pushing us to “get the job done.”He edited draft chapters over and over again, andDick Light contributed big chunks to the chapteron measures of association, which Paul and I re-did and integrated with the asymptotics chapter.If everyone who had come to Fred’s house back in1968 had become involved, we might still be work-ing on the book today! Fred didn’t want his nameon the cover of the book. So we had this back-and-forth. The book ended up with five names on thetitle page; it’s Yvonne Bishop, Stephen Fienberg,Paul Holland, with the collaboration of FrederickMosteller and Dick Light; Dick had contributed toa chapter in the book and Fred had contributed tothe whole enterprise.
JT:
The book, which many have called the “JollyGreen Giant” because of its cover, really put you onthe map. In fact, that’s how we met, when I tookthe short course the three of you gave based on thebook in 1976 at the Joint Statistical Meetings.
SF:
We actually met earlier, when Fred organizeda meeting in Cambridge to discuss the ASA-NCTMbook projects that ultimately produced
Statistics by
CONVERSATION WITH STEPHEN E. FIENBERG Example and
Statistics : A Guide to the Unknown ,your first magnum opus . I was a bit intimidatedsince you seemed to be the organizer for
Statistics : A Guide to the Unknown , and so we just didn’t talkmuch.
MS:
Steve and I met around the same time as well.I remember his coming to Chicago to interview andtalking about the geometry of 2 × SF:
But when I got to Chicago you were one ofthe few good students who didn’t take my contin-gency table course. You were too busy campaign-ing for Hubert Humphrey and worrying about weakconvergence!
MS:
Well, one of the things that you have ad-vanced in that book and elsewhere derives from thegeometric structure that gave you so much insightinto what’s going on in these tables. Now, you men-tioned taking geometry at Toronto, and we know R.A. Fisher was influenced by this, so how did thatplay out in the later research?
SF:
It’s come into play in an amazing sort of way.If you look at the cover of
Discrete MultivariateAnalysis , there is an artist’s depiction of the sur-face of independence for a 2 × × JASA and published a generalization in the
Annals ,and I always thought about contingency tables andother statistical objects geometrically. Don Fraserthought geometrically, and so you’re always up here“waving arms” in some abstract space, and he wouldalways wave with his arms. And I think in high-dimensional space in some sense, although obviouslywe don’t see in high-dimensional space. But a lot ofstatistics is projecting down into lower-dimensionalspaces. I had left the geometry stuff behind, exceptfor motivation, until I got into confidentiality re-search in the 1990s.In the 1990s there was a paper, unpublished forfive years by Persi Diaconis and Bernd Sturmfels.Persi was at Cornell and Bernd had been at Cornellbut moved to Berkeley. In the paper, they talked
Fig. 10.
Steve, Judy Tanur and Morrie DeGroot, JointStatistics Meetings, 1978. about the algebraic geometry structure associatedwith contingency tables. This turned out to be rightat the heart of what I needed for my problem, and soI learned algebraic geometry, which I had not reallystudied carefully before. I learned at least enough tobring my problems to Bernd for help. And one of thethings I realized is that figure on the cover of Bishop,Fienberg and Holland was being used by algebraicgeometers in a different context; it’s called a SegreVariety, named after Corrado Segre who was oneof the fathers of algebraic geometry. That work isnow reflected in the theses of a couple of my formerPh.D. students and lies at the heart of a lot of whatI’ve been doing over the last several years, includ-ing recent work on algebraic statistics and networkmodels.
JT:
I think I derailed you sometime back whereyou were talking about the trajectory of your career.And we’ve left you at Minnesota. Can you tell uswhy you left?
SF:
Minnesota was a giant bureaucracy. It wasa big, big university, and one of the moments thatconvinced me of this was after I had presented areport, prepared with colleagues from around theuniversity, to the president and the vice-presidentson the teachings of statistics at the university, whereI had pointed out that 40 different departments orunits where teaching statistics or courses in whichstatistics represented a serious part of the activity.Virtually all of this was going on with little or nocoordination with the School of Statistics. And thenI met him [the president of the university] about amonth later at a reception. Joyce and I were goingthrough the reception line, and I shook his hand, andhe asked what department I was from. I said applied M. L. STRAF AND J. M. TANUR statistics, and he said, “
Do we have a statistics de-partment at the University of Minnesota? ” At thatpoint I said to myself, “Oh my goodness!” and I un-derstood where the School of Statistics and my de-partment stood in the big picture of the university.A year or two later, I was wooed by friends at an-other Big Ten university, but the right offer didn’tquite come to pass. In the mid-70s I was working asan associate editor for the
Journal of the AmericanStatistical Association , initially with Brad Efron astheory and methods editor, and then with MorrieDeGroot. Later I became Applications and Coordi-nating Editor of
JASA , and so Morrie and I workedtogether. We had become friends a number of yearsearlier, drinking in a bar together at an IMS regionalmeeting. Morrie and Jay Kadane, who had joinedthe Department of Statistics at Carnegie Mellon inthe early 70s, and I would interact at the Bayesianmeetings that Arnold Zellner organized twice a year.They both knew that I had flirted with the possibil-ity of leaving the University of Minnesota, and theysaid, “You should just come to Carnegie Mellon; youcould bring the rest of
JASA over and we’d have thewhole journal. Besides, it’s a great place.” So theyworked on the possibility of an appointment for me.When I came to interview, it wasn’t just to meetwith the Dean, and with Jay and Morrie and thepeople in the department that I knew. They tookme to see the president of Carnegie Mellon (CMU),who at the time was Richard (Dick) Cyert. Dick wasan economist but also a statistician! He took coursesfrom Hotelling and Cochran at Columbia as a grad-uate student, and although his degree was in eco-nomics, he always thought that he was a statisticianas well. In particular, he was a member and Fellowof ASA. Dick helped found the CMU Department ofStatistics in the mid-1960s when he was the dean ofthe Graduate School of Industrial Administration.He was actually the acting chair at the outset untilMorrie took over. So the staff ushered me into his of-fice. I had never met Dick before, but that afternoonI spent two hours with the president of CarnegieMellon. And I told you about my interaction withthe president of the University of Minnesota! HereI am sitting with the president of Carnegie Mellon,this great university, and he’s telling me how im-portant it is for me to come to Carnegie Mellon andwhat I’m going to do for the field of statistics. Hesaid, “If you come here, everything you do will becalled statistics. You will get to change the field.”So I came. And I hope that I’ve changed parts ofthe field.
Fig. 11.
Richard Cyert, Dennis Gillings and Steve, at a Na-tional Institute of Statistical Sciences Board of Trustees Meet-ing, 1993.
MS:
Cyert was a visionary, and really led the Grad-uate School of Industrial Administration to a highplace among business schools and understood thathe needed quantitative strength, and so he influ-enced you and supported you. I wanted to ask aboutone of your greatest honors, and that is your elec-tion into the National Academy of Sciences. Wherewere you and how did you get the word?
SF:
Most people don’t know what goes on at theNational Academy—it’s like a secret society—andit’s selection process is Byzantine, running over thecourse of one or more years. At the end, the NASmembers meet in Washington at the annual meet-ing in a business meeting and they elect the newmembers. That happens between 8:30 and 9 in themorning; then they take a break in the meeting andeverybody rushes out to find a telephone and theycall their friends and the newly elected members tothe section to congratulate them. This was in thespring of 1999, and I was teaching—actually thatyear I was teaching an introductory statistics class,so I had to be there relatively early—it was just at 9o’clock, I was opening the door to my office, and thephone rang. I answered and it was several friends,mainly demographers—Jane Menken, Doug Massey,a couple of others—and there was a chorus on thephone saying “Congratulations, you’ve been electedto the National Academy!” I was floored, becauseI’m not quite sure whether they knew, a year or soearlier I wouldn’t have been eligible, because I wasborn and raised in Canada, and I hadn’t becomean American citizen until January 1998. Thus beingelected the next year was a special honor.
CONVERSATION WITH STEPHEN E. FIENBERG JT:
You have received many other awards andhonors; that must be very exciting.
SF:
Well I would be lying if I said that receivinghonors and awards is not fun, and each is alwaysvery special. But I am reminded about somethingthat Fred taught me. He said that awards and hon-ors are really not for the people who get them, butthey are for the field. Of course the person gettingthe honor benefits, but the field benefits more, forexample, when statisticians get elected to the Na-tional Academy of Sciences. In that sense we don’thave enough big awards.
MS:
There are some of our colleagues who arehappy that there isn’t a Nobel Prize in Statistics,and as a consequence statisticians cooperate morewith one another than scientists in other fields. Doyou agree?
SF:
Well, I think if we follow Fred’s reasoning wewould all be better off with a Nobel Prize in Statis-tics because once a year all of the newspapers andmedia in the world would focus on our field andthe accomplishments in it. What most statisticiansdon’t know is that there almost was a Nobel Prize!The story goes back several decades when Pet-ter Jacob Bjerve, who was the director of Statis-tics Norway, began to raise funds for a Nobel Prizein Statistics. He was off to a good start when heran into a political obstacle. Those in charge of theprize in Economic Sciences objected because, theyargued, their prize encompassed a large amount ofwhat was important in statistics. In the end Bjerveabandoned his quest, and the money he raised wasleft in a special account in Statistics Norway. Finally,the government auditors forced Statistics Norwayto close this account and our colleagues there de-cided, among other things, to use the funds to hosta special international seminar, to which they in-vited statisticians such as Fred Smith from the UK,Jon Rao from Canada, Wayne Fuller, me and a fewothers. They paid for our spouses to come as welland we got the royal (small R) treatment, with rel-atively fancy hotel rooms and outstanding dinners.So in this sense you could say that I ate the NobelPrize in Statistics, although there is no public recordand it doesn’t show up on my CV.
JT:
You’ve been active in several committees andpanels and so forth, including at the National Acad-emies before and after your election as a member—what stands out particularly from those?
SF:
Well, of course this is Bill Kruskal at work—most statisticians who are going to read this inter-view don’t know the history—Bill Kruskal founded the Committee on National Statistics (CNSTAT) atthe NAS. It was an outgrowth of the 1971 Report ofthe President’s Commission, chaired by Allen Wal-lis and co-chaired by Fred Mosteller; and Bill talkedthe people at the National Academies, and the Na-tional Research Council (NRC, its operating wing),into creating a committee although there was no ex-ternal funding, and the NAS really had to put upresources. Bill ultimately got some money from theRussell Sage Foundation to tide the committee overwith a part-time staffer—Margaret Martin, who wasand is absolutely fabulous and with whom the threeof us have worked—and the committee slowly gotgoing. Bill was succeeded by Con Taeuber. At thattime I actually was on another committee, on therehabilitation of criminal offenders, but Miron wasworking for CNSTAT and I would run into him onoccasion. I got to join CNSTAT a year or so laterwhile I was still doing the work on criminal justice.Getting involved in CSTAT was like all these otheractivities I have been describing—I was exposed tolots of new ideas and problems to work on. I was likea kid in a candy shop! The committee didn’t havea lot of projects then, but I just got to look aroundthe Academy and the Federal government, and therewere possibilities everywhere. I could only do somuch, but I pushed the staff to do other things andgot my friends on the committee to lead panels. Bythe mid-80s the committee was humming and therewere all these neat activities on census methodology,on cognitive aspects of survey methodology, statis-tical assessments as evidence in the courts, sharingresearch data—there was just no end.
MS:
I wanted to ask about one of them in particu-lar, which Judy chaired and which you were instru-mental in creating, and that is Cognitive Aspects ofSurvey Methodology. When you were inducted intothe American Academy of Political and Social Sci-ences, you referred to that in your speech as one ofthe most important activities that you had partici-pated in. Why was this and how did it affect yourwork?
SF:
Well, sample surveys is a very strange part ofstatistics. In my department, nobody else really doesit, in the research sense. People think the theory issettled. But doing surveys is really hard. The mea-surement problems are enormous. Designing ques-tionnaires is a big, big problem. In the 1970s I gotinterested in the National Crime Survey on Victim-ization through the SSRC committee on social in-dicators in Washington on which I served. I learnedabout the difficulties in counting victimization events. M. L. STRAF AND J. M. TANUR
Fig. 12.
Participants at 1983 CNSTAT Workshop on Cognitive Aspects of Survey Methodology watching a survey interviewvideo, from left to right: Kent Marquis, Judy Tanur, Phil Converse, Lee Ross, Steve (in upholstered chair), Miron Straf.
In 1980 Al Biderman, who was involved in the re-design effort for the victimization survey, broughttogether a few people from the re-design projectwith cognitive psychologists to ask if we could learnsomething from cognitive science. I thought this wasjust terrific because I could see ways that I couldtake methodological statistical ideas and really in-tertwine them with the theoretical ideas that cameout of cognitive psychology. As a consequence,I pushed for that CNSTAT activity even thoughothers thought it made no sense. I was part of theCNSTAT workshop that you and Judy organized—Judy and Beth Loftus and I wrote a series of 4 pa-pers on cognitive aspects of surveys afterward. I wasalso on the SSRC council, and we created a commit-tee that followed up on those activities. It brought innew people to the enterprise, and it helped get theseideas embedded in the statistical agencies. JanetNorwood ran with the idea at BLS. It was part of theculture at NCHS at that time because Monroe Sirkinwas at the CNSTAT workshop and a moving spiritin establishing a cognitive laboratory at NCHS. TheBureau of the Census was actually the last of thebig three agencies to create a separate laboratoryfacility—but they did—and the influence spread be-cause the associated ideas changed research at the boundaries of survey methods and psychology in avariety of different ways. The reason I am especiallyproud of this activity is because you’d hardly knowthat there was any statistical theory or methodologylurking behind it, but there really was.
MS:
It’s really had a profound effect on the surveyfield, and now in many places it’s commonplace—concepts of cognitive interviewing and all that.You’ve been especially close to your students, fos-tering them personally as well as professionally. Pic-tures of you attending weddings of your students ap-pear frequently on websites in your honor. So couldyou tell us a little about your personal interactionswith your students.
SF:
Well, in the early years the students were mycontemporaries. In fact, I had a couple of studentswho were older than I was. Kinley Larntz was notonly my Ph.D. student and collaborator, but wewere good friends, and remain so. Over the yearsI got a little older than my students, and whenI moved to Carnegie Mellon I really had the op-portunity to have a different kind of student, andwith them different kinds of interactions. We werea small department in those days and I interactedwith lots of students, not just those whose researchI supervised. Each of the students I worked with
CONVERSATION WITH STEPHEN E. FIENBERG Fig. 13.
Steve with friends at the Objective Bayesian Analysis meeting in Rome, June, 2007. From left to right: Steve, LarryWasserman, Jim Berger, Susie Bayarri, Robert Wolpert, Isa Verdinelli. then was interested in a somewhat different topic;they went in different directions, and we remainedclose in most instances.But then, something happened—first, I becamea dean, and then four years later I left CarnegieMellon, as you know. I had a second administrativecareer going on the side—actually, I had three ca-reers, or four. There was also the committee workat the National Academy, which was a full-time jobfor awhile, there was the methodology I worked onin part with students in the Department of Statis-tics, and I was also an administrator—I was De-partment Head for three years and then I was Deanof the College of Humanities and Social Sciences.I was on an administrative track in the late 1980sand early 1990s, and my contact with graduate stu-dents actually tailed off toward the end of my timeas Dean. I was also teaching, but there are only somany hours in the day and days in the week. In 1991,I left and went to the York University in Torontoas Academic Vice-President (that’s like a provost—they don’t have that title at York) and so my regularties with graduate students were severed. I resignedfrom Carnegie Mellon to go to York, although wedidn’t sell our Pittsburgh house, and I returned to Carnegie Mellon a few years later and re-joined thedepartment.I like to describe the move back to Carnegie Mel-lon as a promotion to the best position in the univer-sity—as a tenured professor with no administrativeobligations. I slowly began to work with graduatestudents again. Somewhere along the way I thinkI had learned something, which is you can’t neces-sarily get graduate students to do what you want,and thus what you have to do is get them to do whatthey want to do in the best possible way. You haveto get them to complete a thesis, but you have to beable to get them through and have them gain confi-dence in what they’re doing so that they think theycan make a difference. And I was lucky—I just hadfabulous students; they were terrific people and allthe rest of the stuff just sort of happened. I hadthe opportunity to give away in marriage one ofmy students, Stella Salvatierra, who was workingin Spain, at a ceremony in the mayor’s office inBilbao, because her father had a heart attack andcouldn’t come to the wedding. And there have beenseveral other weddings since! Because my studentshave been so great, the best thing I can do in somesense is to get them to do the things that they do M. L. STRAF AND J. M. TANUR
Fig. 14.
Steve with his wife Joyce and many of his former graduate students at a 65th birthday celebration at CarnegieMellon, October, 2007. From left to right: Ellie Kaizer, Edo Airoldi, Elena Erosheva, Jason Connor, Sesa Slavkovi´c, MikeMeyer, Joyce, Steve, Alessandro Rinaldo, Justin Gross, Russ Steele, Adrian Dobra, Amelia Haviland, Elizabeth Stasny.
Fig. 15.
The longtime members of the Carnegie Mellon Department of Statistics in the DeGroot Library, 2011. Back row: RobKass, Mark Schervish, Steve, Joel Greenhouse; middle: Margie Smykla; bottom row: Jay Kadane, Bill Eddy, John Lehoczky.
CONVERSATION WITH STEPHEN E. FIENBERG best. That’s in many ways a serious part of mylegacy. JT:
I was going to ask you what advice you wouldhave for graduate students in statistics, or under-graduates for that matter. Clearly, the best adviceI could give would be for them to come to be yourstudents, but since you can’t spread yourself totallythin, failing that, what alternative advice would youoffer?
SF:
Well, I really can’t work with them all! It’sreally bad because now we’ve got this undergradu-ate program with upwards of 150 majors. I can dealwith one or two graduate students at a time. But myadvice to budding statisticians is simple: statisticsis an exciting field. There are all these neat prob-lems. There are neat theories, neat methods, neatapplications; we’re in a new world. Big, big datasets. My joint appointments are now in the MachineLearning Department and in the Heinz College (ofPublic Policy and Management). I’m working withdata sets that people couldn’t conceive of dealingwith a few years ago. And the students I’m workingwith have the ability to go and do things with thosedata sets that were unimaginable a decade ago. Somy advice is simple. Work with data, take problemsseriously, but you have to learn the mathematicsand statistical theory if you want to do things right.And then you need to take seriously teaching peo-ple what you’ve done, not just doing the research.You need to get the descriptions of your work intoa form that other people can understand—that’s areally important part of what we do. That’s whatNational Academy reports are all about. Academyreports don’t have impact if they’re badly written.Enormous effort goes into the executive summariesof reports, into the review process, and everythingup the line. Learning how to do that as a studentis time well spent. It’s too late when you’re a fullprofessor and you still haven’t learned how to writearticles so that other people can understand whatyou’ve done.
MS:
So, of your vast experiences, what are youthe most proud of?
SF:
I’m actually proud of a number of things. Bythe way, I didn’t tell you what my fourth careerwas. I play ice hockey—I still play, that’s numberone, although the one for which I have the fewestskills or accomplishments.
MS:
All right, let me interrupt you. . .
SF:
Ha ha, no-no, as I left the locker room lastSaturday night, one of the guys across the dressing
Fig. 16.
Steve and Bill Eddy celebrating the 20th anniver-sary of Chance, a magazine they co-founded in 1988, wearingtheir original Chance t-shirts. room said to me, “So how many years have you beenplaying?” And I said, “62.” He then said, “ ?” andsilence ensued. But maybe hockey is really numbertwo; number one is my children and my grandchil-dren. They’re really amazing. They’re another partof my life. Joyce and I were really fortunate; I havetwo very smart sons, Anthony and Howard. Theyhave independent careers, they have lovely wives. . . MS:
Where are they now?
SF:
Anthony lives in Paris, and I have five grand-children in Paris, four granddaughters and a grand-son. And Howard lives in the DC area and I havea lovely granddaughter in Vienna, Virginia. Howardactually has come very close to statistics, as govern-ment liaison for a consortium dealing with surveysand marketing. The grandchildren are terrific. I lovebeing with them. We get to look after them everyonce in a while.Then there are my students. They’re really thepeople who are going to do the things that I canonly imagine. As I look back over what I’ve done,I see a changed field of statistics. Fred Mosteller andBill Kruskal were fabulous—and we’ve talked abouthow they shaped all three of our careers, not just M. L. STRAF AND J. M. TANUR
Fig. 17.
Steve (on the right) playing for the Division C na-tional championship as a member of the Leiden Beaver BeerTeam, in Eindhoven, March, 1997. my career. And they launched the Statistics Depart-ments at their respective universities. I was partof both departments and their programs in retro-spect look “traditional.” They emphasized mathe-matical statistics and probability. I like to think thatwhen I left Chicago and went to Minnesota, I startedto change what statistics did and how we thoughtabout it. And applications today sit at the core ofmuch of statistical theory and methods, and in mydepartment at Carnegie Mellon our students comeout having worked on multiple applied projects, andthey’re in demand, because that’s the future of ourfield. People recognize that advances in statisticalmethods—and theory—are intertwined with realproblems, major applications. I like to think thatI contributed to the change that we’ve seen over thepast 40 years.
MS:
Very nice, Steve. What you talk about is alegacy, not the individual research that may wanein importance over the years. . .
SF:
And it’s not just my work, it’s a collective. . .
MS:
But it’s the influence of your students, aswell as your children. I wanted to interrupt, because
Fig. 18.
Steve with twin granddaughters, Tiffany and Se-lena, trying out their new bikes, Paris, 2006.
I never thought you had four careers, I thought youhad dozens of careers. You talked about these pro-fessors that, you know, worked 24 /
7, so that wasyour model. As long as I’ve known you, you’re al-ways multi-tasking, and you were doing that beforethe word was even in vogue. You’re fielding ques-tions at a seminar or flying a hockey puck acrossthe ice. Did any of that rub off on your sons, onyour students?
SF:
I don’t think that either Anthony or Howardis quite as obsessed as I am with doing so manythings simultaneously.
MS:
How fortunate. . .
SF:
That’s right! But Anthony did play hockeyin Paris for many years, and both Anthony andHoward have these terrific kids—since Anthony hasfive they take up more of his time than mine did.Actually, Anthony has inherited some of this multi-tasking, at least at some level. He’s created his ownbusiness in France—a subsidiary of a Dutch insur-ance company. His job went from finding the lo-
CONVERSATION WITH STEPHEN E. FIENBERG Fig. 19.
Steve, buried amidst files, in his CMU office, 2005. cation to organizing the offices, to hiring the staff,to inventing the insurance policies and making surethat they were consistent with the ones of the parentcompany.My students also develop multiple facets of theircareers and lives. I tell them when they come inand ask if they can work with me that there area couple of things that are going to happen if thearrangement is going to succeed. One is they’re go-ing to live and breathe statistics. I see it everywhere.One of my favorite examples in my little contingencytable book came out of the program from the sym-phony at the Minneapolis Orchestra one night whenwe were there in the 1970s. It didn’t quite look likea contingency table, but I made it into one, as Table2.4. Then in my book, I described why you shouldn’tanalyze it the way you would have otherwise be-cause the units of observation are not independent.At any rate, I tell the students that I expect themto live and breathe statistics. They’ll get their ideasin the shower. . . they’ll play hard too, but when allis said and done, if they’re not into what they’re do-ing, they should find another advisor, because otherpeople have different attitudes about work and howto get your inspiration! Students of course have theirown lives, and as I’ve said, you don’t tell studentswhat to do, they tell you what they want to do.
JT:
What’s next? For you?
SF:
Wow. I’m too busy to stop at the moment tofind out! I still have more than one job. I’m editing, with some others, the
Annals of Applied Statistics ,I have launched the
Journal of Privacy and Confi-dentiality , I’m co-chair of the Report Review Com-mittee at the Academy. I have a whole bunch ofnew Ph.D. students and post-docs. We’ve got someabsolutely fantastic projects going on: research onconfidentiality problems and on network modeling,which by the way, links to confidentiality. Judy andI also have a book on surveys and experiments topolish up for publication, as Fred Mosteller wouldsay. I have six chapters that were, I had thought,pretty polished at one stage, but they are still ina drawer in my office. At least I know where thedrawer is.
JT:
And I know where my copies are. . .
SF:
And so, I’ve got more books to write too—with good collaborators.
MS:
Well, we’re almost out of time, but I haveone final question. How would you like to be remem-bered, Steve?
SF:
Unfortunately not as a great hockey player.As long as my teammates just let me on the ice, I’mhappy to be able to skate around and get off safely.I guess I’d like to be remembered as somebodywho produced really good students and who helped Steve took over as editor-in-chief of the
Annals of AppliedStatistics on January 1, 2013, and is simultaneously servingas the founding editor of yet another publication,
The AnnualReview of Statistics and its Application , scheduled to launchwithin the year. M. L. STRAF AND J. M. TANUR change the image of statistics in the sense that lotsof people now work on serious applied problems andhelp solve them. And that’s not just about statis-tics, that’s real interdisciplinary scientific work, andthat’s the legacy I inherited from Fred and Bill Krus-kal and Paul Meier, and all those other great peoplethat I had a chance to work with, like Bill Cochran.I would just like for people to think of me in theirkind of company, in some way or another. I suspectthat a couple of decades from now, if anybody everlooks at the video we’re making or reads this in-terview, they may not remember log-linear modelsfor contingency tables and other forms of counteddata because there will be new methodology, like the mixed membership and related models I now workwith. What I know from students today is that, if itwasn’t in the journals in the last three years, they’renot sure it’s worth their attention. So, if I am to havea legacy it needs to be something larger. I have notheorems, well, I do have theorems, but none of themare named
Fienberg’s Theorem . And even if therewere a Fienberg’s Theorem, it probably wouldn’tbe important—what’s important is the attitude, forwhat statistics is and how it’s recognized by otherpeople outside of our field.