A Conversation with Martin Bradbury Wilk
aa r X i v : . [ s t a t . O T ] N ov Statistical Science (cid:13)
Institute of Mathematical Statistics, 2010
A Conversation with Martin BradburyWilk
Christian Genest and Gordon Brackstone
Abstract.
Martin Bradbury Wilk was born on December 18, 1922, in Montr´eal,Qu´ebec, Canada. He completed a B.Eng. degree in Chemical Engineeringin 1945 at McGill University and worked as a Research Engineer on theAtomic Energy Project for the National Research Council of Canada from1945 to 1950. He then went to Iowa State College, where he completed aM.Sc. and a Ph.D. degree in Statistics in 1953 and 1955, respectively. Aftera one-year post-doc with John Tukey, he became Assistant Director of theStatistical Techniques Research Group at Princeton University in 1956–1957,and then served as Professor and Director of Research in Statistics at Rut-gers University from 1959 to 1963. In parallel, he also had a 14-year careerat Bell Laboratories, Murray Hill, New Jersey. From 1956 to 1969, he was inturn Member of Technical Staff, Head of the Statistical Models and Meth-ods Research Department, and Statistical Director in Management SciencesResearch. He wrote a number of influential papers in statistical methodologyduring that period, notably testing procedures for normality (the Shapiro–Wilk statistic) and probability plotting techniques for multivariate data. In1970, Martin moved into higher management levels of the American Telephoneand Telegraph (AT&T) Company. He occupied various positions culminatingas Assistant Vice-President and Director of Corporate Planning. In 1980, hereturned to Canada and became the first professional statistician to serve asChief Statistician. His accomplishments at Statistics Canada were numerousand contributed to a resurgence of the institution’s international standing.He played a crucial role in the reinstatement of the Cabinet-cancelled 1986Census. He remained active after his retirement, serving as a Senior Advisorto the Privy Council Office as well as on several national commissions. In ad-dition, he chaired the Canadian National Task Forces on Tourism Data andon Health Information. Martin is a former President of the Statistical Societyof Canada (SSC) and Vice-President of the American Statistical Association(ASA). He is an elected member of the International Statistical Institute andan honorary member of the SSC. He has received many honors, including theGeorge Snedecor Prize, the Jack Youden Prize, the F.G. Brander MemorialAward, the SSC Gold Medal, and a Distinguished Alumni Achievement Cita-tion from Iowa State University. He is a fellow of the Institute of MathematicalStatistics, the American Statistical Association, the Royal Statistical Society,the American Association for the Advancement of Science, and the New YorkAcademy of Science. He was made an Officer of the Order of Canada in 1999for his “insightful guidance on important matters related to our country’snational statistical system.”The following conversation took place at Martin Wilk’s home in Salem,Oregon, October 6–7, 2005.
Key words and phrases:
AT&T, Canadian census, probability plots,Shapiro–Wilk statistic, Statistical Society of Canada, Statistics Canada.
Christian Genest is Professor of Statistics,D´epartement de math´ematiques et de statistique,Universit´e Laval, 1045, avenue de la M´edecine, Qu´ebec,Qu´ebec, Canada G1V 0A6 e-mail: [email protected]. Gordon Brackstone isretired; he was formerly the Assistant Chief Statistician,Informatics and Methodology Field, Statistics Canadae-mail: [email protected]. C. GENEST AND G. BRACKSTONE
Fig. 1.
Official photograph of Martin Bradbury Wilk, circa1985.
INTRODUCTION
Christian : Martin, tell us something about youryouth.
Martin : I was born in Montr´eal in December 1922.I was the youngest in a family of three children. Ihave two sisters who are still alive today. We livedin a part of town where English was dominant, andI should immediately confess to you, Christian, thatalthough I had many years of training in French,I ended up with virtually no capacity in that lan-guage. I think I’m just one of those people, unfor-tunately, who inherited not much capability for a
This is an electronic reprint of the original articlepublished by the Institute of Mathematical Statistics in
Statistical Science , 2010, Vol. 25, No. 2, 258–273. Thisreprint differs from the original in pagination andtypographic detail. second tongue. But I dare say, it wasn’t much of ahandicap in Montr´eal, at least in those days.I got my secondary education at the now defunctStrathcona Academy in Outremont [now part of Mont-r´eal]. I had no burning academic interest at thetime. Mathematics and geometry were the topics Ithought were easiest. In Grade 11, which was thefinal year, my worst grades were in French, oral andwritten. Other than that, all topics seemed prettystraightforward, and I guess I learned some bad work-ing habits as a result.
ENGINEERING AT McGILL
Christian : After graduation from Strathcona in1940, you studied engineering at McGill University,in Montr´eal. What guided your choice?
Martin : When I was young, I was generally in-terested in the technical side of things, and McGillseemed like the place to go. It was one of the largestand most prestigious universities in Canada. It hada big reputation, especially in medicine. There wasa year of transition with English literature, history,mathematics, and so on. Then you could go for afour-year program in the Engineering Faculty.I found that first year really easy, but once I gotinto engineering that turned out to be quite a dif-ferent story. The first two years of the engineeringprogram were common to all. We were about 130people at the start, and whoever gave the welcom-ing speech made a remark to the effect that half ofus would be gone by the end of the first year. Andhe was right.The first year in engineering was really hard forme. My bad habits were a disservice to me, and Iended up being second to last in the list of peoplewho made the cut! Luckily, my performance beganto improve at that point, and I was able to completethe program.
Gordon : What got you interested in chemical en-gineering specifically?
Martin : I enjoyed chemistry in high school, al-though we never had a chance to use a laboratory,because that wasn’t part of the understanding atthe time. Once I got to McGill, I found out that Iwas very good hands-on. Also, there was an arrange-ment at that time that you had to be employed forthe summer and write a term paper with regard toyour experience. And as it turned out, all my sum-mer jobs had to do with chemistry.
CONVERSATION WITH MARTIN B. WILK SUMMER JOBS
Christian : Where did you work as a student?
Martin : At Howard Smith’s paper mill the firstsummer. They operated out of Cornwall, Ontario.Then a year later, I worked for Distillers Corpora-tion in Montr´eal. This is a company that SamuelBronfman had founded; it produced commercial al-cohol. It had enjoyed substantial growth in the 1920s,due to Prohibition in the States. But during the war,the company was busy making pure alcohol for thearmy.In my final summer before graduation, I workedfor the Montr´eal Coke and Manufacturing Company.It was the most interesting of my summer jobs. Itwas a dry distillation of coal to produce coke. It wasdone in drying ovens whose temperature had to bemonitored very closely, 24 hours a day. There weredata to be collected and I was involved in that. Iremember going up the chimney, wearing a sweaterto keep me from being burned, and gathering data.
Gordon : Did the war affect you in any way?
Martin : Not really. There was no conscription inCanada in the early 1940s. However, we did have tocome to school in uniform twice a week, and spendsome time at so-called training. I also learned howto identify a flying aircraft and to read Morse codealong the way. But more significantly, perhaps, isthat when I graduated in 1945, the war was still on.So I was offered a choice between doing a Ph.D. inchemistry at McGill or joining Canada’s NationalResearch Council (NRC).
Christian : You make it sound as though you hadno other option. Was that really the case?
Martin : Well, I mean I was told “That’s whereyou’re going.” And at the time it was part of theunderstanding that you do as you’re told. I mean,the alternative to that was you go in the army.
Gordon : And why didn’t you go for the Ph.D.?
Martin : It paid well: $75 a month, if I recall. Thiswas a substantial amount of money at the time. ButI didn’t have the patience. You see, I don’t thinkI’m a natural student. This is something I learnedsubsequently. And I found that listening to peo-ple telling me about things that they have done, orabout things that other people have done, was notpart of my nature.
CHALK RIVER LABORATORIES
Christian : What was your position at NRC?
Fig. 2.
Martin’s graduation photograph, McGill University,Montr´eal, 1945.
Martin : When I was hired in June, 1945, I movedto Ottawa and spent about six months there. ThenI joined the Chalk River Laboratories, located inDeep River, in the upper Ottawa valley. The firstnuclear reactor outside of the United States had justgone operational up there. There were about 250scientists on the site. Nuclear technology was thefocus of our activity.There were many people from abroad. One of themost remarkable figures was Bruno Pontecorvo, anItalian physicist who had worked with Enrico Fermi.I got to know him because each of us separately hada bad habit of missing the bus that drove us to workin the morning. Pontecorvo was probably a genius,and I am not using that term loosely. It came to meas a great surprise when he defected to Russia in1951, in the middle of the Cold War.
Gordon : And what were your responsibilities?
Martin : I was mostly in charge of testing an aircooling system for the rods that were used in the
C. GENEST AND G. BRACKSTONE
Fig. 3.
Martin’s Photo ID when he applied for graduateschool in statistics at Iowa State College. experimental heavy-water pile. I gathered a lot ofdata in this context and under the best of circum-stances, there was a great deal of variability asso-ciated with it. Radioactive behavior is by-and-largeunpredictable, except on average. So I began devel-oping a few techniques of my own to handle suchdata.
STATISTICS TRAINING IN IOWA
Christian : Does this explain why you chose toleave Chalk River in 1950 to do graduate work instatistics at Iowa State College?
Martin : Well, not quite. Let me explain. My timein Chalk River was very pleasant and productive.I even had a paper published there [23]. But afterfive years, I had a feeling that if I stayed longer,I’d probably be there for the rest of my life. And Ishould mention something here, which at the time Ihad no knowledge of, but it’s really turned out thatthroughout my career, I had a different job everyfive years. Not that I ever manipulated or arrangedfor it, but it just so happened.Anyway, it so happened that some friends of minehad gone to Ames, Iowa. They were microbiologistsand enjoyed it there. So they encouraged me to jointhem and even went as far as negotiating a contractfor me with a chap called R. G. Tischer, who op-erated in food technology. He offered me a positionas a research assistant and the salary was enough tomake me feel that I could go.
Gordon : So you went just like that?
Martin : To tell you the truth, it was sort of amindless decision, like many I took in my life. Ididn’t really think hard about it and don’t recallthat I needed to be particularly brave. And actu-ally, once I got there, I found out that ProfessorTischer’s research program was much too prosaic inmy outlook. I do recall spending countless hours in the lab, carrying out experiments for him, and read-ing instruments, etc. But I was quickly bored.Now Professor Tischer knew enough about his ownsubject to recognize that there was variability in theobservations he was dealing with, and given my priorexposure to data in Chalk River, I could providesome assistance with this. Eventually, he encouragedme to sign up for a statistics course, which I did, andI was interested to discover that there was a theorybehind much of what I had basically developed onmy own, in some form.
Christian : Do you remember who taught you thatcourse?
Martin : It was Bernie Ostle. He had me meet withvarious people in the Statistics Department, and heencouraged me to solidify my mathematical encoun-ters. I began doing that, and it all came very natu-rally. I mean, I really never had in mind that I wasgoing to leave chemical engineering, but before thefirst year in Ames was over, I decided that it wasn’tright for me to continue taking money from Tischer.I decided to pursue the mathematics I was engagedin and work toward a Master’s degree.
Gordon : Who was your advisor?
Martin : It was Oscar Kempthorne. He supervisedmy Master’s thesis, which I completed in 1953, andthen he persuaded me to stay on for the Ph.D.,which I got in 1955. My thesis was in the area oflinear models and the analysis of variance for ran-domized block designs. It led to several publications[24, 25, 45–47], some of them joint with Oscar.
Christian : What was it like to work with ProfessorKempthorne?
Martin : He was a man of substantial intellect andhe had very strong feelings about a variety of mat-ters, technical, and nontechnical. I shared an officewith him throughout most of my studies in Iowa,and we grew to be very good friends. He tendedto be quite outspoken and to express his opinionsemphatically. This may explain in part why the de-partment was polarized, or at least separated be-tween those people around Herman Hartley, whowere involved in survey sampling, and those like Os-car Kempthorne and I, who were more concernedwith analysis of variance and experimental design.
Gordon : Who else was there at the time?
Martin : People like Bernie Ostle, John Gurland,Ted Bancroft, and so on.
Christian : And you joined the group after youcompleted your thesis, right?
CONVERSATION WITH MARTIN B. WILK Martin : No, actually I was established as an As-sistant Professor there before I completed my Ph.D.But as it turned out, Oscar wanted me to gain ad-ditional experience before settling in at Iowa State.So he got in touch with John Tukey, who was work-ing at Princeton and for Bell Laboratories, and hemade arrangements for me to go there for a year asa post-doctoral fellow. I was actually quite pleased with this offer, be-cause I had already met briefly with Tukey at themeeting of the American Statistical Association inMontr´eal, in September 1954. I presented a joint pa-per with Oscar Kempthorne at that meeting, and Iencountered John Tukey at that time. He alreadyhad a reputation as the ultimate wise man, and myimpression was one of awe. He struck me as a very
Fig. 4.
Bernie Ostle teaching statistics students at Iowa State College, circa 1950.
Fig. 5.
Martin and his first wife, Thora Sugrue, at their home in Neshanic, New Jersey, circa 1957.
C. GENEST AND G. BRACKSTONE intelligent fellow, but it was only later that I foundout how smart he really was.
POST-DOC AT PRINCETON UNIVERSITY
Christian : Tell us about your post-doc year atPrinceton University then.
Martin : I arrived at Princeton in the spring of1955. Sam Wilks was about to leave for Texas andI inherited his office.
Christian : Wilks’ office became Wilk’s office, eh?
Martin : That was a good start! David Cox wasthere at the time; he and I worked under Tukey’sleadership that summer. Tukey had a grand planand lots of problems he wanted us to work on. Hewas the smartest man I’ve ever associated with upclose. His speed of thinking was quite incredible,and his ability in mathematics was overwhelming.He was always so far ahead that he could lectureme on what I was trying to do far beyond what Icould ever manage to accomplish. The whole processwas very depressing and irritating to me. David Coxseemed to cope better; I guess he had an interest anda capacity that I didn’t have. But until David leftin early fall, the two of us would often commiserateon the fact that Tukey was as smart as he actuallywas.Anyway, my propensity is not to want to learnthings unless I have a motivation, so I thought myyear in Tukey’s environment at Princeton was mis-erable. Maybe I didn’t do as badly as I felt at thetime, but in any event, I was in pretty bad shapeby the end of the year. At the same time, my homelife was hectic too. While doing my Master’s thesisin Ames, I had gotten married to Thora Sugrue andby that time we had four young children. That drewa lot of energy!
RESEARCH AT BELL LABS
Christian : So how come you didn’t go back to Iowaat the end of the year?
Martin : Well, in the spring of 1956, John Tukey,who seemed to know more about me than I knewabout myself, said how would you like to spend thesummer at Bell Labs? I hadn’t a clue what Bell Labswas, but it sounded intriguing. So, at John’s insti-gation, I went there and gave a talk in front of thestatistics group, which was under the direction ofMilton Terry at the time. I must have made a goodimpression because at the end, they made me anoffer and after some cogitation, I accepted.
Christian : Did they offer you a regular contractright then?
Martin : No, it was just a post-doc at first. Butthen John Tukey arranged for me to work two days aweek at Bell Labs and the rest as an Assistant Direc-tor for the Statistical Techniques Research Group atPrinceton. So I got in touch with Oscar Kempthorneand Ted Bancroft, who was the Department Chair,to let them know that I would not return after all.
Gordon : What prompted your decision?
Martin : The first summer I spent at Bell Labs wasa very exciting time. To begin with, I encounteredcomputer technology in the guise of an IBM 650,which was an incredibly fast machine at the time.Also, the friendly atmosphere and the freedom youenjoyed as a researcher working for Bell Labs wereexceptional. The staff was about 20,000 overall, andmaybe 10% of these were scientists doing researchin physics, chemistry, materials engineering, mathe-matics and statistics, etc. The Research Departmenthad a distinctive character, and I was able to get in-volved in many projects as a statistical consultant,and to some extent, as a chemical engineer, too. Acouple of publications ensued [12, 49].
Christian : And what were your responsibilities atPrinceton?
Martin : I was involved in research there too. ButI still found that association to be disappointing.
Gordon : And why was that?
Martin : Chiefly because of my inability to sup-port the work of George Box, who had just beenappointed Director of the Statistical Techniques Re-search Group at Princeton. He was a very kind per-son, and he had a quick and relatively creative mind,but our styles just didn’t match. He tended to bevery focused in his research, and he was good atit too, but I guess because of my background andthrough my contacts with people at Bell Labs, I hada much broader view of things. Besides, I had prettywell fallen in love with Bell Labs at that time. So Iquit Princeton after only a few months, but throughBell Labs, my association with John Tukey lasted along time.
Christian : Did your responsibilities at Princetoninclude any teaching?
Martin : I had done a little bit of teaching in Iowa,but I did none at Princeton. Where I did teach wasat Rutgers, where I was a professor of statistics from1959 to 1963.
CONVERSATION WITH MARTIN B. WILK PROFESSOR AT RUTGERS
Gordon : How did that come about?
Martin : I was approached in 1958 by Ellis Ott,who was heading the Department of Statistics there.He wanted to establish a Ph.D. program. He per-suaded me to come down and teach an evening classin mathematical statistics. I agreed, and then in Jan-uary 1959 he came up with the notion that I couldbe a full professor there and have the responsibilityfor research. I agreed, on the provision that I couldstill do consulting and other work with Bell Labs,one day a week.The arrangement appealed to me on several ac-counts. First, Rutgers paid well and I needed thatkind of money, given that I had a large family. Sec-ond, I thought maybe I’d like to work in a universityenvironment and lecture, as opposed to listening toother people. But I didn’t want to give up on BellLabs completely. The colleagues at Bell Labs turnedout to be very encouraging in this regard.Finally, an additional reason why Rutgers appealedto me was that Marion Johnson, who was Dean ofthe Graduate School at the time, was close to retire-ment and the indication was that I would be a goodcandidate for replacement.
Christian : And did your expectations materialize?
Martin : I was on faculty at Rutgers four yearsin total, and I certainly did my best to serve theirinterests. But to tell the truth, I had a miserabletime throughout that period, in that my wife Thoracame down with cancer. This episode started almostimmediately after I joined the faculty at Rutgers. Iwas at home almost round the clock, caring for herand looking after the children. In the end, we lostthe battle against cancer and she died on April 15,1965. Needless to say, I never applied for the Dean’sposition at Rutgers.Somehow, despite the hardships, I managed to bereasonably productive throughout that period. Nodoubt, this was due in part to the great collaboratorsI had. I am thinking especially of Ram Gnanade-sikan and Sam Shapiro. Another more prosaic factoris that throughout my entire life as an adult, I neverslept more than about four hours a night. Of course,I would also “cat nap” on occasion during the day.In fact, I did it even while I was Chief Statisticianof Canada, but I’m getting ahead of myself now.
CONTRIBUTIONS TO STATISTICALMETHODOLOGY
Christian : You made a large number of contribu-tions to statistical methodology in the 1960s. Howdid that develop?
Martin : Much of the research I did in that periodwas dictated or inspired by questions of consultancyat Bell Labs. If you look at my publications from1960 to 1970 [2, 6–17, 19–21, 26, 37–44, 48], you’llsee that a prime concern of mine was diagnostic pro-cedures for classical distributions, e.g., the normalor the exponential and the gamma. A fair portionof this work was carried out with Ram Gnanade-sikan, who was a colleague at Bell Labs, and SamuelShapiro.Ram and I formed a highly compatible team. Ithink we wrote 12 papers together. One thing aboutRam is that because of his Indian origins, he hada cultural bias about seniority and although eachof us would do our fair share of our joint work, heinsisted that my name be listed first; this continuedfor many years [37–44]. As for Shapiro, he was astudent in one of my classes at Rutgers. He was agood guy, and I agreed to supervise his dissertation.Our association continued for a while [16–19, 48].At one point in time, it became clear to me thatthe behavior of order statistics, in some sense, wouldhave to reflect the nature of a distribution. Ram,Sam, and I used the properties of these statistics todesign QQ-plots and goodness-of-fit tests.
Christian : The 1965
Biometrika paper [16] thatintroduces the Shapiro–Wilk test statistic is certainlya classic.
Martin : I am obviously pleased with that, but tobe truthful I don’t regard this work as such a greataccomplishment. As I am fond of saying, significancetests are things to do while one is trying to think ofsomething sensible to do. This being said, while itis true that the idea behind the test was mine, Samcarried it considerably further with power compar-isons and approximations to the null distribution ofthe W statistic, as he kept calling it. Christian : This was truly seminal work. It is just abit unfortunate that many people think of your con-tributions and those of Sam Wilks as coming fromthe same person!
Martin : Well, the difference between Wilk andWilks was always clear to me at least! But you areright that my work with Ram and Sam led to ahost of publications, much of them quite sophisti-cated mathematically, too. To an extent, however,
C. GENEST AND G. BRACKSTONE this flurry of activity illustrates a problem I see withmathematical statistics and more generally with sci-ence as I perceive it today. The problem is that thereare a lot of facets to science, and it is now quitefragmented and being pursued in an opportunisticfashion by individuals in ever more specialized cat-egories: mathematics, statistics, multivariate analy-sis, and so on down the line. I think this is regret-table.We need people who look at problems much morebroadly, and certainly the work that Ram Gnanade-sikan, Marylin Becker, and I did on the problem ofspeech recognition was much broader in nature andof much greater importance potentially, althoughour efforts in this direction did not translate intorefereed publications. Ram has given a very goodaccount of the atmosphere in our research group,and the issues we contended with, when Jon Ket-tenring interviewed him for
Statistical Science [Vol.16 (2001), pp. 295–309].
Gordon : It sounds as though you became grad-ually disenchanted with mathematical statistics, ifnot more. Is that why you moved on to managementat AT&T in 1970?
Martin : I certainly became disenchanted. Throughthe 1960s, I acquired a certain sense of the wholeorganization: I mean Bell Labs, of course, but alsomore generally AT&T. And if I self-examine, I wouldsay that by 1969, I was waiting for an opportunityto go into administration.
GETTING INTO MANAGEMENT AT AT&T
Christian : How did that actually happen?
Martin : Through hearings for a rate case thatAT&T had to submit to in front of the FederalCommunications Commission (FCC). At the time,AT&T was a huge organization. It was effectivelycontrolling the local and long-distance telephone net-work throughout the US, and because of its quasi-monopolistic position, the FCC had something tosay about its rate of return to equity. One impor-tant ingredient in the hearings held at the end of the1960s was the so-called Gordon model. It is a vari-ant of the discounted cash flow model or a methodfor valuing a stock or business, if you wish. It wasoften used at the time to provide difficult-to-resolvevaluation issues for litigation. It was named afterMyron Gordon, who was a finance professor at theUniversity of Toronto.Now at the time, John Tukey had been asked byAT&T to criticize some conclusions that had been derived by the FCC using this model. And John be-ing the irritating person that he was (because hecould figure out everything so much more quicklythan anybody else), I was curious to see how myhero would fare in the formal and constrained envi-ronment of a federal commission hearing. So I askedpermission to attend that part of the hearings. Andat the time, I had no sense of the global issue athand, but John talked about this model, and I wasa bit amazed to see that everybody treated him withsuch great respect. Not that he didn’t deserve it, butlitigation tends to be rather merciless.
Gordon : It sure can be. But how did you person-ally get involved?
Martin : John’s testimony sparked my interest inthe Gordon model, so upon my return to New York,I started to investigate it on my own, and then I ranquite a few computer simulations. Marilyn Huyett,who was a close collaborator [12–14, 41–43], helpedme check the computations. Within about a week,I got a pretty good sense of what was wrong withthe proposed application of this simplistic model. Iwrote up my findings as a technical note and sentit to John Tukey, who suggested that I pass it onto the upper management at AT&T. So I did, andthen I got a phone call from Mark Garlinghouse,AT&T’s General Counsel, who asked me whether Iwould mind briefing him about the Gordon model.Amazingly, he had read my report, which was def-initely not written from a business point of view.In fact, part of it was later published in conferenceproceedings [27].At any rate, Mark Garlinghouse and I met for awhole day shortly thereafter. We got on extremelywell, though we had important political differences.He listened to me quite carefully and ended up mak-ing representations to the FCC, based on my report.I attended his presentation and found the experiencequite stimulating.The bottom line is that through that event, the se-nior people at AT&T developed the realization thatthey needed more attention paid to what was re-ferred to by them as “management science.” Subse-quently, I was invited to spend some time there andhelp them out with some of their difficulties. It wasa lateral move, shall we say, and at first I did it as acourtesy to them. But after I had a few interactions,I just got captured by the effort and started takingan interest in the general character of AT&T as anorganization.
CONVERSATION WITH MARTIN B. WILK GOING UP THE RANKS AT AT&T
Christian : From 1970 to 1975, you worked atAT&T in various capacities. You were successivelyDirector of Corporate Modeling Research (1970),Director of Corporate Research (1971), Director ofPlanning (1972), and then Director of CorporatePlanning (1973–1975). What was the bulk of yourresponsibilities?
Martin : I was initially part of a unit headed byHenry Boettinger. The function of that small group,less than 20 people in total, was to look into man-agerial and financial problems. The group was ac-tually unique at AT&T, which had no mathemati-cian, no economist, or statistician at the time. Theorganization had basically three functions: raisingmoney, operating interstate transmissions, and pro-viding common services for the 22 telephone com-panies in the AT&T group that ran within-statecommunications. Our group was set up to formu-late plans and policies, to ensure coordination withthe developers at Bell Labs and the implementers atWestern Electric.
Gordon : With such responsibilities, was it hardfor you to remain active on the statistical front?
Martin : Well, I didn’t actually drop technical workcompletely, at least in the early 1970s. For instance,I wrote a paper with Sam Shapiro on an analysis ofvariance test for the exponential distribution [18].We were subsequently awarded the Jack Youden Prizefor the best expository paper that appeared in
Tech-nometrics in 1972. I also managed to get some jointwork done with Jane Gentleman [3, 4]. But as timeunfolded, the bulk of my responsibilities became moreand more managerial. And of course even more sowhen I was Assistant Vice-President and Director ofCorporate Planning, from 1976 to 1980.
Gordon : Can you summarize what you accom-plished in that period?
Martin : Generally speaking, I carried out workon many large and small tasks in economics, finan-cial and technological areas, including cost analysis,registration, and what was then referred to as “pic-turephones.”One major thing that was dumped on me was theManagement Research Information System (MRIS).It had been initially designed to be the ultimatepath-finding operation, a crucial ingredient in op-erating an interrelationship between the thousandsof pieces and parts that a telephone system involves.As it turned out, this MRIS was a colossal mess,and I made it legal to say so. I found that it was not possible to reorganize it in any sense or fash-ion. And the embryonic computing system that wasthere to make it alive, supposedly, consisted of partsthat could never fit together. In part, this was be-cause every operating telephone company had a dif-ferent computing environment at the time. Also,sadly, some of the people did not have the exper-tise and the knowledge required to pull it together.There just weren’t many people around then, whoknew how to operate these systems in a coordinatedfashion. Yet the notion of an integrated system cameup every time AT&T wanted to make an upgradewith what was already in place! In the end, I gotthis MRIS closed down.
CHIEF STATISTICIAN OF CANADA
Gordon : How about the transition from AT&T toStatistics Canada? This occurred in 1980, right?
Martin : Again, this is not something that I sought.It came up on its own about five years into my exec-utive position at AT&T. As it happens, things werein a bad shape at Statistics Canada at that time,and the Secretary of the Treasury Board had setup a group under the headship of Claus Moser tolook into the technical competence of the federalagency. And there was a second, thicker documentthat had been prepared by management consultants.At any rate, some member of the Executive Recruit-ing Committee phoned me in New York on a Fridayafternoon, either in May or June 1980, to tell methis long story.
Gordon : Sorry. Were you being approached for theposition of Chief Statistician of Canada?
Martin : They probably had that in mind, but theguy was not explicit about it and it didn’t occur tome at first, for two reasons. First, it was not un-usual for senior people at AT&T to be approachedfor counsel, as a service to the public, and given mybackground, it would not be completely unexpectedfor Canadians to seek my opinion in a period of tur-moil at the national statistical agency. Second, I was57 at the time, so I considered myself too old or tooclose to retirement to be perceived as a candidatefor the job.Nevertheless, I was sufficiently interested to lookat the reports that were sent to me after the call, andthen to travel to Ottawa to share my impressions.
Gordon : Who did you meet in Ottawa?
Martin : The meeting was chaired by Jack Manion,who was Secretary of the Treasury Board of Canada C. GENEST AND G. BRACKSTONE at the time. Other participants were Harry Rogers,Fred Drummie, and Larry Fry, who was then interimChief Statistician. They summarized the two reportsbriefly, and then Manion asked me point blank howI would handle the situation, and why I felt I coulddo the job! That was a real eye-opener to me, andI responded by saying that I had come to give myreactions to the reports, and that I would like tostick to what I came up there to do. He seemedrather puzzled by my reaction, but we stuck to myagenda.After I returned to New York, we had an exchangeof correspondence and eventually, I went back toOttawa for further discussions. The Moser reportwas well done, though possibly a little too lenient onthe quality of the surveys and the publications, butit was a fair assessment. The other report, however,was a terrible document with all kinds of criticismsthat made no sense to me, even though I knew littleabout the organization. But what I knew and what Ihad heard did not correspond at all to that report’sassessment of the situation.
Gordon : So in the end, what convinced you tomeet that new challenge?
Martin : As I reflected upon the subject, I be-came quite apprehensive as to what might happento Statistics Canada if nothing was done. And thething that attracted me to take the job is that con-trary to the US, official statistics in Canada was setup as an integrated system in which data collection,analysis, and so on are coordinated parts of a grand plan. Anyway, I think I was half decided alreadywhen I was invited for brunch by Michael Pitfield,who was Clerk of the Privy Council and Secretaryto the Cabinet at the time.
Gordon : For the benefit of our readers, it shouldprobably be said here that the Clerk of the PrivyCouncil is the most senior nonpolitical official in theGovernment of Canada. He provides professional,nonpartisan support to the Prime Minister on allpolicy and operational issues that may affect thegovernment.
Martin : Right. And Michael Pitfield, who occu-pied the position at the time, was a most obligingperson. He gave me a very honest briefing about therealities of what was going on at Statistics Canada.He was quite frank, in particular, about the existingfracture lines within the organization.
Gordon : Before we get to that Martin, could youtell us how soon after that meeting you accepted theposition?
Martin : It took a few more months. You see, mysecond wife, Dorothy Barrett, had problems withher eyes at the time. She had to undergo surgery,and I was not in a good disposition to make a deci-sion right then.However, I kept making enquiries about StatisticsCanada through the fall of 1980, while attending tomy regular duties at AT&T. In those days, there wasalso a certain amount of reorganization taking placein the New York head office, and this influenced myhook up.
Fig. 6.
Martin brushing up his French soon after joining Statistics Canada in 1980.
CONVERSATION WITH MARTIN B. WILK Fig. 7. “Uncle Martin” in a poster designed by StatisticsCanada and adopted widely in the Canadian Government topublicize a program launched in 1984–1985 to enable civil ser-vants to change assignments.
Gordon : Had you had any real involvement withofficial statistics prior to becoming Chief Statisticianof Canada?
Martin : I was on the US Census Advisory Com-mittee in the early 1970s. I remember urging themto develop some sort of strategic plan for evolutionand commit it to paper. And I remember makingan argument (which I subsequently made at AT&Ttoo) that although the plan may not be fully imple-mented, it would at least force them to think wherethey were, where they wanted to go, and how toget there. Also, it is important to put a calendar toit. At a later stage, I remember making a numberof concrete suggestions to the US Census Director,Vince Barabba, and they were well received.
MAIN CHALLENGES AT STATISTICSCANADA
Christian : Can you summarize the issues that youfaced when you arrived at Statistics Canada?
Martin : One major organizational problem wasthe division of social statistics and economic statis-tics into two almost completely separate units. Thesesections were headed by Ivan Fellegi and Guy Leclerc, respectively. Each of them had a title of AssistantChief Statistician. The survey methodology peoplewere also split between these two sections.In addition, there was a major problem in humanrelations and in the image that the organization wasprojecting. It affected its credibility in the eyes ofthe public. There were allegations of nepotism in thepress and many good people ended up leaving theorganization because of the extremely negative cli-mate that prevailed. Jacob Ryten was one of them.And you left too, Gordon, isn’t that right?
Gordon : Yes. I went to work for the British Colum-bia statistical agency for a while.
Martin : The morale at Statistics Canada was verylow, and this was at a time when the 1981 Censushad to be carried out. Mind you, the planning forthat operation was well underway. At the heart of itwas Ed Pryor. It was a major undertaking, but wewere globally satisfied with it.
Gordon : Luckily, by the end of your term, staffrelations had improved considerably.
Martin : I am glad it worked out okay. I certainlytried to assist every way I could. One person whohelped me a great deal in sorting out the staff prob-lems was Jean-Jacques Blais, who was the Minis-ter responsible for Statistics Canada when I tookoffice. We developed a very good rapport. In addi-tion, the Treasury Board was prepared to give mesome slack, and that helped me considerably in deal-ing with some individuals. There was a fair amountto be done in this regard, but there is no pointtalking about this, as it would involve people byname and so forth. Suffice it to say that there wasa certain amount of unpleasantness about it, butnot much beyond that. And ultimately, staff move-ments helped resolve the separation between socialand economic statistics.
OTHER ISSUES
Christian : Besides human relations, could you giveus an idea of the issues you had to deal with?
Martin : There were many on the methodologicalside. And to be truthful, I didn’t have much of aclue at first as to what was going on in terms ofproduction, i.e., what people did, how they did it,and why they did it.One major problem for me was that I couldn’t af-ford to take the time to find out in depth, because Ineeded speed of reaction that was far greater thanmy speed of learning. For example, there were ma-jor discrepancies in employment figures at the time C. GENEST AND G. BRACKSTONE between the Labour Force Survey and the Survey ofEmployment Payroll and Hours. The problem was ofconsiderable importance and urgency because theseparticular surveys attract a lot of attention in thepublic.Another major issue that troubled me very earlyon was the range of publications that was turnedout by Statistics Canada. There seemed to be norhyme or reason to much of it. Much time, effortand money was being wasted in producing publica-tions that very few people would read. The publica-tion issue was eventually resolved; many significantchanges were made. It was a long battle, and manycompromises needed to be struck in order to makepublications financially self-sufficient.
Gordon : These efforts resulted in fewer publica-tions, and more importance to those retained.
Martin : Right. On another front, I promoted aform of operational integration within headquartersin Ottawa, and also regionalization of the data col-lection operations. I tried to give more importanceto the regional offices in Halifax, Montr´eal, Toronto,Winnipeg, and so on. The process has expandedsince, but setting it up involved a lot of fightinginitially. I ended up having to move a lot of peo-ple around, but in so doing I always tried to avoiddisrupting their career goals.And yet another important battle I recall was get-ting access to cabinet documents. There were strate-gic reasons behind this action, but it was part ofa concerted effort to bring Statistics Canada intoa closer relationship with other departments. Thisis one thing that struck me when I first joined theagency: it seemed that Statistics Canada operatedalmost entirely on its own, and that it had veryfew contacts with other branches of government. Ineeded to make the personal effort of contactingthe Deputy Ministers in charge of the various de-partments to set up working committees that couldoversee the information and data collection needs ofthese folks, and establish mechanisms to meet theirdemands.
Christian : By setting up external advisory groups,you also tried to help the agency in its efforts to im-prove its effectiveness and coverage and it ultimatelyled, in the early 1990s, to international marks ofrecognition for the world-class quality of its statis-tics.
Martin : Yes, that was the long-term plan, and itwas satisfying to see it materialize. To this end, I es-tablished quite a few of these advisory committees, maybe a dozen. And many of them have survived tothis day. In fact, were you not a member of the Ad-visory Committee on Statistical Methods with mein the late 1990s?
Christian : That’s right. I was on the committeefrom 1994 to 1999, I think.
Martin : I should take this opportunity to say thatthe most prominent of all these committees was ac-tually set up by my successor, Ivan Fellegi.
Gordon : You mean the National Statistics Coun-cil, right?
Martin : Absolutely.
THE 1986 CENSUS
Gordon : Maybe we should move on to the 1986Census now? I know you had just retired when itwas run but as I recall, it was nonetheless a majorissue for you.
Martin : Oh yes, indeed. When Brian Mulroneybecame Prime Minister of Canada, in September1984, he decided to make a clean sweep and under-took a number of initiatives. As you may remember,his party won the largest majority government inCanadian history. One of Mulroney’s priorities wasto control the deficit, which was running into thebillions of dollars. He started seeking contributionsfrom every department that would result in mon-etary savings. I was not initially consulted on thisissue but it turned out that the Minister responsi-ble for Statistics Canada volunteered on his own tocancel the 1986 Census!
Christian : Who was this?
Martin : It was Harvie Andre, a pleasant enoughfellow from Alberta who, incidentally, had a Ph.D. inChemical Engineering. But he was quite opinionatedand didn’t have a high regard for statisticians. Hecould see substantial savings in cancelling the 1986Census, and so he had simply gone ahead with it!When I first heard about this, it was underway.When I found out, I really had a rush of anger. I felthe had absolutely no right to make such a judgmentwithout consulting me at least. As I recall, I reallyblew him.
Gordon : What happened next?
Martin : Given the circumstances, my first respon-sibility was to establish a high-level committee in-ternally to review what would need to be done tomeet the requirement if it came to it. Second, inparallel to that, I encouraged people from outsidethe agency to make their feelings known as to how
CONVERSATION WITH MARTIN B. WILK much their operations would be affected if the cen-sus were cancelled.Pretty soon, we found out that by law, the Prairieprovinces had to have a census in 1986. And also,that the lack of census would affect seriously a hostof programs within and outside the agency. Harvie Andre and his staff came to recognize their mistake,and so the census was reinstated. But in the mean-time, the commitment had been made that Statis-tics Canada would cut off 100 million dollars fromits budget. That was an awful lot of money; maybea third of the agency’s entire budget! Fig. 8.
Martin Wilk at his retirement party, January 1986, receiving from his successor, Ivan Fellegi, the keys of a motorscooter, a gift from the employees of Statistics Canada.
Fig. 9.
Martin addressing SSC members at the banquet held during the 1987 SSC Annual Meeting at Universit´e Laval,Qu´ebec. Guests at the table of honor in the background (from left to right): Sir David Cox, Terry Smith, Marc Moore and hiswife, and Michael Stephens. C. GENEST AND G. BRACKSTONE
Christian : So it looks like the politician reachedhis goal anyway. How did you get around that one?
Martin : Through intense rounds of discussions withgovernment representatives like Jack Manion andHarry Rogers, it was agreed that the money we hadalready saved by integrating our operations and ra-tionalizing our publications would be included in the100 million. At the end of the day, that left me look-ing for some 40 million only!
Gordon : That’s still a hefty sum of money.
Martin : Through additional negotiations and re-flection, we came up with a big package in whichStatistics Canada absorbed maybe 10 million dol-lars on its own. Another chunk came through con-tributions from some departments that had vestedinterests in the census. And then the rest came as acredit for student hirings. You see, we estimated thatover 40,000 people would have to be hired to run thecensus, but the money was not actually needed until1986!
Gordon : All this happened within a very shortperiod of time too. It must have been quite a stress.
Martin : It sure was, and I even got physically illover it. So once the 1986 Census was officially rein-stated, I thought it was a reasonable time for me toleave. I mean it was a big victory, and at the time Ifelt there was little more that I could do, at least inthe short run, to establish Statistics Canada on anon-going basis. Bear in mind also that I was 63 atthe time.Now after Guy Leclerc transferred to the Secre-tariat of the Treasury Board, back in 1983, I hadarranged for Ivan Fellegi to be promoted to a newlycreated position of Deputy Chief Statistician. I chosethat designation to make it clear that he was a nat-ural candidate for my succession, and when the timecame for me to retire, I pretty well made it a condi-tion that Ivan would be my replacement.As Deputy Chief Statistician, he had begun tothink of the organization as a whole, including eco-nomics, and he had done very well, making fastprogress. So he was pretty well the natural candi-date, and Prime Minister Mulroney followed the rec-ommendation made by Privy Council. This way, Icould retire in September 1985, and Ivan had enoughtime to get established before the 1986 Census wasrun. He has done very well since and is now, I be-lieve, in his twentieth year as Chief Statistician ofCanada. [Note: Dr Fellegi retired from that positionon June 12, 2008.]
PRESIDENCY OF THE SSC
Christian : Martin, when you had your official re-tirement party in January 1986, you were just start-ing your term as President of the Statistical Soci-ety of Canada (SSC).You had already made quitean impression on its members at the 1985 AnnualMeeting in Winnipeg. Many of us remember vividlyyour after-dinner speech on “blue-collar and white-collar statisticians.” There was a lot of wisdom inthat talk which, luckily, was later published in
Sur-vey Methodology [31].
Martin : In fact, it is even available in French [32]because this journal, also known as
Techniquesd’enquˆete , is published in both languages by Statis-tics Canada.
Christian : Right. And because of a change in theSSC By-Laws, you were in fact the only person toserve as President for more than a year. Could youtell us about your term?
Martin : When I was voted in, I found that theSSC was a pretty messy organization. It was still suf-fering, I think, from the unfortunate circumstancesthat led to its creation, in 1978. Of course, you al-ready know that story, Christian, having related itas you did with David Bellhouse in some past issueof
Statistical Science [Vol. 14 (1999), pp. 80–125].Anyway, I had an orientation at the time of mak-ing statistics better known and more visible inCanada. One way of accomplishing that was to ele-vate the level of the
SSC Newsletter to a full-fledgedpublication. I was fortunate that Nicole Gendreauagreed to take on that mandate, and to involve theresources of the Bureau de la statistique du Qu´ebecthat she was head of at the time. That’s how SSC’s
Liaison emerged as the quarterly, fully bilingual mag-azine that it continues to be to this day.Another major effort was the in-depth revision ofthe SSC By-Laws, which Peter Macdonald and you,Christian, put a lot of effort into.
Christian : When these By-Laws were approved bythe Board of Directors, I remember you congratu-lating me on this effort and telling me “And now,young man, on to greater things!” I was 30 then. Itmade a big impression on me. I’ve been trying tolive up to the challenge since!
Martin : These volunteer jobs are essential for theprofession, but they can be really time-consuming. Ialways tried to do my share, having served as Presi-dent of my local chapter of the American StatisticalAssociation, and then as Vice-President at the na-tional level, from 1980 to 1982. I was also a member
CONVERSATION WITH MARTIN B. WILK of the founding Editorial Board for Technometrics (1959–1963).
AFTER RETIREMENT
Gordon : You certainly had a very full professionallife, and your achievements have been underscoredwith many honors. From what I know, you didn’treally slow down after retirement. Could you sum-marize very briefly?
Martin : At first, I served as Senior Advisor to thePrivy Council Office, say for a period of six months.Afterward, I did quite a bit of consulting for Statis-tics Canada, Revenue Canada, the US Bureau of theCensus, the Ontario Premier’s Council, and so on. Ialso served as a member, and sometimes as a Chair,on various bodies both in Canada and in the USA,where my wife and I chose to go back in recent years.One of my major assignments was on the Cana-dian Institute for Advanced Research, whose mis-sion is to orient and promote research in Canada.The Institute might identify nanometrics, say, as anarea of opportunity for the country, and then therewould be a task force set up to explore that possi-bility, and so on. The Population Health Programwas one of the most visible initiatives launched bythis institute.
Christian : But your involvement with the TaskForce on Health Information was even greater. Right?
Martin : Indeed! At the time there was a publicsense that statistical information about the perfor-mance of Canada’s health system was either nonex-istent or fragmented, health being a matter of provin-cial jurisdiction. The Task Force report, that I wrotemuch of myself, was well received and led to a re-vamping of the organization and management ofhealth statistics. This included, for example, the cre-ation of the Canadian Institute for Health Informa-tion to complement the survey work of StatisticsCanada by gathering and analyzing operational in-formation from health institutions across the land.One other important committee I got involvedwith had to do with science. Curiously, StatisticsCanada had never gotten really interested in thearea of science. Our objective, or at least one ofthem, was to report on the characteristics and char-acter of science as being pursued in Canada. Mem-bers of the steering committee included Jacob Ry-ten, Steven Fienberg, as well as Mike Sheridan,Michael Wolfson, Scott Murray, Ray Ryan, and pos-sibly others.
Christian : And for quite a while, of course, youremained active with Statistics Canada as a memberof its Advisory Committee on Statistical Methods!
Martin : I enjoyed that a lot. But I am now over 80and I have pretty well given up on all these things.These days, I just try to enjoy myself with my wifeDorothy, my children and my grand children.
Fig. 10.
Martin Wilk and his second wife, Dorothy Barrett, at the Schreckhorn peek in Grindelwald (Switzerland), circa1993. C. GENEST AND G. BRACKSTONE
Christian : You certainly deserve it! Is there any-thing you would like to add before we close?
Martin : With your permission, there are severalpersons that I would like to thank here becauseof the important roles that they played in my ca-reer. First and foremost, I would like to mentionmy wife, Dorothy and my children, Rebecca, Carol,David, Teresa, and Kathryn, who saw me throughboth difficult and very pleasant times. Then from aprofessional point of view, I was also quite fortunateto find many collaborators and friends. I am espe-cially grateful to Oscar Kempthorne, John Tukey,Ram Gnanadesikan, Henry Boettinger, Paul Reed,Harry Rogers, Ivan Fellegi, Jacob Ryten, and FraserMustard. Needless to say, the list is not exhaustive.
Gordon : Thank you very much, Martin, for all thetime you have given us.
Martin : And thanks to you guys for traveling toOregon and for the time you devoted to me.
ACKNOWLEDGMENTS
Thanks are due to Marianne Genest for transcrib-ing the interview. Funding in partial support of thiswork was provided by the Natural Sciences and En-gineering Research Council of Canada, the Fondsqu´eb´ecois de la recherche sur la nature et les tech-nologies, and the Institut de finance math´ematiquede Montr´eal. REFERENCES [1]
Fellegi, I. P. and
Wilk, M. B. (1988). Is statistics sin-gular or plural?
Canad. J. Statist. Supplement,1–8.[2]
Gabbe, J. D., Wilk, M. B. and
Brown, W. L. (1967).Statistical analysis and modeling of the high-energyproton data from the
Telstar 1 satellite.
Bell SystemTech. J. Gentleman, J. F. and
Wilk, M. B. (1975). Detectingoutliers in a two-way table. I. Statistical behaviorof residuals.
Technometrics Gentleman, J. F. and
Wilk, M. B. (1975). Detect-ing outliers. II. Supplementing the direct analysisof residuals.
Biometrics Gnanadesikan, R., Kempthorne, O., Nicholson,W. L., Wilk, M. B. and
Zelen, M. (1979). Sta-tistical consulting. Transcript of panel discussion.In
Proc. of Computer Science and Statistics: 12thAnnual Symposium on the Interface
Gnanadesikan, R. and
Wilk, M. B. (1969). Data an-alytic methods in multivariate statistical analysis.In
Multivariate Analysis II (Proc. Second Internat.Sympos., Dayton, Ohio, 1968)
Gnanadesikan, R. and
Wilk, M. B. (1970). Useof maximum likelihood for estimating error vari-ance from a collection of analysis of variancemean squares.
Ann. Math. Statist. Gnanadesikan, R. and
Wilk, M. B. (1970). A prob-ability plotting procedure for general analysis ofvariance.
J. Roy. Statist. Soc. Ser. B Hogben, D., Pinkham, R. S. and
Wilk, M. B. (1961).The moments of the non-central t -distribution. Biometrika Hogben, D., Pinkham, R. S. and
Wilk, M. B. (1964).An approximation to the distribution of Q (a vari-ate related to the non-central t ). Ann. Math. Statist. Hogben, D., Pinkham, R. S. and
Wilk, M. B. (1964).The moments of a variate related to the non-central t . Ann. Math. Statist. Lundberg, J. L., Wilk, M. B. and
Huyett, M. J. (1960). Solubilities and diffusivities of nitrogen inpolyethylene.
J. Appl. Phys. Lundberg, J. L., Wilk, M. B. and
Huyett, M. J. (1962). Estimation of diffusivities and solubilitiesfrom sorption studies.
J. Polymer. Sci. Lundberg, J. L., Wilk, M. B. and
Huyett, M. J. (1963). Sorption studies using automation and com-putation.
Ind. Eng. Chem. Fundam. Pinkham, R. S. and
Wilk, M. B. (1963). Tail areas ofthe t -distribution from a Mills’-ratio-like expansion. Ann. Math. Statist. Shapiro, S. S. and
Wilk, M. B. (1965). An analysisof variance test for normality: Complete samples.
Biometrika Shapiro, S. S. and
Wilk, M. B. (1968). Approxima-tions for the null distribution of the W statistic. Technometrics Shapiro, S. S. and
Wilk, M. B. (1972). An analysis ofvariance test for the exponential distribution (com-plete samples).
Technometrics Shapiro, S. S., Wilk, M. B. and
Chen, H. J. (1968).A comparative study of various tests for normality.
J. Amer. Statist. Assoc. Tukey, J. W. and
Wilk, M. B. (1965). Data analysisand statistics: Techniques and approaches. In
Proc.Cal. Tech. Symposium on Information Processingin Sight Sensory Systems
TheCollected Works of John W. Tukey , Vol. 5, Graphics1965–1985, pp. 1–22 (1988).[21]
Tukey, J. W. and
Wilk, M. B. (1966). Data analy-sis and statistics: An expository overview.
AFIPSConf. Proc. 1966 Fall Joint Comp. Conf. The Collected Works of John W.Tukey , Vol. 4, Philosophy and Principles of DataAnalysis 1965–1986, pp. 549–578 (1988).[22]
Tukey, J. W. and
Wilk, M. B. (1988). Data analysisand statistics: Principles and practice. In
The Col-lected Works of John W. Tukey , Graphics 1965–1985, 23–29. CONVERSATION WITH MARTIN B. WILK [23] Wilk, M. B. (1949). Preparation and extraction of S . Canad. J. Res. Wilk, M. B. (1955). The randomization analysis of ageneralized randomized block design.
Biometrika Wilk, M. B. (1956). Linear models in the analysis ofvariance. In
Proc. 2nd Conference on Design of Ex-periments in Army Research Development and Test-ing
Wilk, M. B. (1970). The corporation and the universityas social engineers. In
Proc. of University of Mary-land Seminar on “University, The Corporation andAmerican Priorities in the Seventies.” [27]
Wilk, M. B. (1972). Utility prices and allowed rate ofreturn. In
Proc. Second Annual Regulatory Infor-mation Systems Conference . St. Louis, Missouri.[28]
Wilk, M. B. (1973). The study of complex systems. In
Computer Science and Statistics: Proceedings of theSeventh Annual Symposium on the Interface (W. J.Kennedy, ed.) 2–10. Iowa State Univ., Ames, Iowa.[29]
Wilk, M. B. (1975). Planning and uncertainty. In
Proc.Computer Science and Statistics: 8th Annual Sym-posium on the Interface
Wilk, M. B. (1975). Uncertainties in analysis of com-plex systems. In
Proceedings of the Symposium onStatistics and Related Topics (A. K. Md. E. Saleh,ed.) 14.01–14.17. Carleton Univ., Ottawa, Ontario,Canada. MR0365785[31]
Wilk, M. B. (1985). The relationship between statisti-cians and statisticians.
Survey Methodology Wilk, M. B. (1985). Statisticiens et statisticiens.
Tech-niques d’enquˆete Wilk, M. B. (1986). Summary of comments at 1986Annual Meeting of SSC.
SSC Newsletter (2) 7–11.[34] Wilk, M. B. (1986). A message from the President. /Le billet du pr´esident.
Liaison (1) 8–11.[35] Wilk, M. B. (1987). A message from the President. /Le billet du pr´esident.
Liaison (2) 8–9.[36] Wilk, M. B. (1987). The concept of error in statisticaland scientific work. In
Proc. Third Annual ResearchConf.
Wilk, M. B. and
Gnanadesikan, R. (1961). Graphicalanalysis of multi-response experimental data using ordered distances.
Proc. Nat. Acad. Sci. USA Wilk, M. B. and
Gnanadesikan, R. (1964). Graph-ical methods for internal comparisons in multi-response experiments.
Ann. Math. Statist. Wilk, M. B. and
Gnanadesikan, R. (1968). Prob-ability plotting methods for the analysis of data.
Biometrika Wilk, M. B., Gnanadesikan, R. and
Freeny, A. E. (1963). Estimation of error variance from smallestordered contrasts.
J. Amer. Statist. Assoc . Wilk, M. B., Gnanadesikan, R. and
Huyett, M. J. (1962). Probability plots for the gamma distribu-tion.
Technometrics Wilk, M. B., Gnanadesikan, R. and
Huyett, M. J. (1962). Estimation of parameters of the gamma dis-tribution using order statistics.
Biometrika Wilk, M. B., Gnanadesikan, R. and
Huyett, M. J. (1963). Separate maximum likelihood estimation ofscale or shape parameters of the gamma distribu-tion using order statistics.
Biometrika Wilk, M. B., Gnanadesikan, R. and
Lauh, E. (1966).Scale parameter estimation from the order statisticsof unequal gamma components.
Ann. Math. Statist. Wilk, M. B. and
Kempthorne, O. (1955). Fixed,mixed, and random models.
J. Amer. Statist. As-soc. Wilk, M. B. and
Kempthorne, O. (1956). Some as-pects of the analysis of factorial experiments in acompletely randomized design.
Ann. Math. Statist. Wilk, M. B. and
Kempthorne, O. (1957). Non-additivities in a Latin square design.
J. Amer.Statist. Assoc. Wilk, M. B. and
Shapiro, S. S. (1968). The joint as-sessment of normality of several independent sam-ples.
Technometrics Wilk, M. B., Torrey, M. N. and
Gohn, G. R. (1958).A study of the variability in the mechanical proper-ties of alloy A phosphor bronze strip.
Proc. Amer.Soc. Test. Mater.58