A Conversation with Ingram Olkin

Statistic al Scienc e 2007, V ol. 22, No. 3, 450– 475 DOI: 10.1214 /0883423 07000000122 c  Institute of Mathematical Statisti cs , 2007 A Conversation with Ingram Olk in Allan R. Sampson Abstr act. Ingram Olkin w as b orn on Ju ly 23, 1924 in W a terbur y , Conn ecticut. His family mo v ed to New Y ork in 1934 and h e graduated from DeWitt Clinton High School in 194 1. He serv ed three y ears in the Air F orce du r ing W orld W ar I I and obtained a B.S. in mathemat ics at the Cit y C ollege of Ne w Y ork in 194 7. After receiving an M.A. in mathematical statistics fr om Columbia in 1949, he completed his gradu ate studies in the Departmen t of Statistics at th e Univ ersit y of North Carolina in 1951. His d iss ertation wa s written un der the direction of S . N. Ro y and Harold Hotelli ng. He joined th e Departmen t of Mathematics at Michiga n S tate Univ ersit y in 1951 as an Assista nt Professor, subsequently b eing pr omoted to Professor. In 1960, he to ok a p osition as Ch air of the Department of Statistics at the Unive rsity of Minnesota. He mo v ed to Stanford Universit y in 1961 to tak e a join t p osition as Professor of Statistics and Professor of Education; he wa s also C hair of the Departmen t of S tatistics from 1973– 1976. In 2007, Ingram b ecame Pr ofessor Emeritus . Ingram w as Ed itor of the Annals of M athematic al Statistics (1971– 1972) and served as the ﬁr st editor of the Annals of Statistics from 1972–1 974. He was a primary f orce in the foun d ing of the Journal of Educ ational Statistics , for whic h h e w as also Associate Editor during 1977–198 5. In 1984, he wa s President of the Institute of Mathemati- cal Statistics. Among his many p rofessional activities, he has serve d as Chair of the Committee of Presidents of Statistical So cieties (COPSS), Chair of the Committee on Applied and Theoretical Statistics of the Natio nal Research C ouncil, Chair of the Managemen t Board of the American Education Researc h Association, and as T rustee for the National Institute of Statistical S ciences. He h as b een honored by the American Statistical Asso ciatio n (ASA) with a Wilks Medal (1992) and a F ound er’s Aw ard (1992) . T he American Psycholo gical Asso ciation ga ve h im a Lifetime Contribution Aw ard (1997) and he w as elected to the National Academ y of Education in 2005. He receiv ed the C O PSS Elizab eth L . Scott Aw ard in 1998 and deliv ered the R. A. Fisher Lecture in 2000. In 2003, the City Unive rsity of New Y ork ga v e him a T ownsend Harris Medal. An author of 5 b o oks, an editor of 10 b o oks, and an author of more than 200 publi- cations, Ingram h as made ma j or con tributions to statistics and education. His researc h has fo cused on m ultiv ariate analysis, ma jorization and inequalities, distribution theory , and meta-analysis. A volume in celebration of Ingram’s 65th b ir thda y conta ins a brief biograph y and an interview [Gleser, P erlman, Press and Sampson (1989)]. In gram was c h osen in 1997 to participate in the American Statistical Asso ciation Distinguish ed Statistician Video Series and a videotaped con versatio n and a lecture (Olkin, 1997) are av ailable fr om the ASA (1997, DS041, DS042). Key wor ds a nd phr ases: Educational statistics, m a j orization, meta-analysis, multi- v ariate analysis, pr obabilit y inequalities. Al lan R. Sampson is Pr ofessor, Dep artment of Statistics, Cathe dr al of L e arning 270 1, University of Pittsbur gh, Pittsbur gh, Pennsylvania 15260, USA e-mail: asampson@stat.pitt.e du . This is an e lectronic reprint of the origina l a rticle published by the Institute of Mathematica l Sta tistics in Statistic al Scienc e , 2007, V ol. 2 2, No. 3, 4 50–47 5 . This reprint diﬀer s from the origina l in pa g ination and t yp ogr aphic detail. 1 2 A. R. SA MPSON This con v ersation to ok place on Decem b er 9, 2005 at the Renaissance Hotel in W ashington, DC. INTRODUCTION A ND EAR L Y YEARS Sampson: Ingram, w e’v e b een trying to get to- gether to do this conv ersation for quite a while. It’s diﬃcult to ﬁnd an y gap in yo ur extremely hec- tic tra ve l sc hedu le. Y ou’re p erh aps more inv olv ed in statistics no w than y ou w ere wh en y ou b egan y our career 55 y ears ago. And the question I wan t to start with is: w hat is it ab out statistics that’s still so com- p elling to y ou? Olkin: That’s an in teresting qu estion. S tatistics has a r ole in so m an y diﬀerent applications, and what I alw ays ﬁ nd exciting is the fact that yo u’re confron ted with a new discipline and a new set of p eople, and they b ring diﬀerent and interesting sci- en tiﬁc questions in whic h the statistician can partic- ipate. Ju st to illustrate, I wa s most recent ly ask ed to discuss some asp ects of what is called “the v alue of a s tatistical life” with the Envi ronmental Pr otectio n Agency . This is a fascinating p roblem b ecause it co v- ers very large areas of the en vironment, and what’s in teresting is the p eople working on the p ro jects are economists, not statisticia ns, so it brings me in to con tact with a whole new area. Sampson: Y ou’v e b een in v olv ed w ith these v arious forms of app lied problems f or a long time. But what k eeps y ou still so ﬁred-u p? Olkin: This question ab out th e “ﬁre” is one I hav e trouble answering. Sampson: It ma y precede y our b eginn in gs in statis- tics—p erhaps s omething in y our upb ringing. Olkin: I s u sp ect that’s true. My mother had ﬁre unt il she was 98. Sampson: Amen! Olkin: The “ﬁre” was mostly addressed to me. And I think there ma y b e some genetics b ecause m y d augh ters h a v e a certain amount of that trans- mitted. Sampson: When y ou were a c hild, were you as in- tense and as passionate in whatev er you were d oing then as y ou are now? Olkin: Let me put it this wa y . I was b orn in W a- terbury , Connecticut and I lived ther e until age 10. I w en t to a public sc ho ol th at had v ery little heat, and w hen you misb eha v ed th ey sen t you to sit in the cloakroom, wh ic h was cold. What I can tell y ou is m y mother was a constan t visitor to the s chool b e- cause I w as in the cloakro om an inordinate amoun t of time. S o, I th ink I was a very activ e kid at that time. Sampson: It sound s more than activ e, but p erhaps reb ellious. Olkin: I don’t kn o w. Bu t I do kn o w, and I can’t tell you what grade, the teac her tap ed my mouth. Sampson: Some of yo ur colleag ues no w, I susp ect, wish they knew that teac h er’s secret! [Laugh ter] Olkin: No w, of course, this was abuse under cur - ren t deﬁnitions. But, it didn’t b other me at the time. I think I’v e alwa ys b een activ e and int erested in d if- feren t things. I think that certainly is a characte ris- tic. Sampson: Did y ou ha ve a lot of int erests as a c h ild? Olkin: Y es I did . And I think I w as sort of go o d in a lot of diﬀerent areas and so I w as int erested in sp orts, I collected stamps, I wen t to theaters, music, and s o on. I really av ailed myself. Sampson: Some of that sounds lik e it o ccurred when you we re older. Olkin: Th eater and m usic o ccurred during my early teens after I mo ved to New Y ork. There w ere not that many outside inte rests a v ailable in W aterbury , Connecticut. T he only place that w as av ailable wa s a library , and my mother took me to the libr ary on man y , man y o ccasions. It w as a ritual. That’s an in teresting p oint b ecause m y mother was an immi- gran t who really didn’t h a v e m u c h education. Sampson: Where wa s y our mother from? Olkin: My mother came from W arsa w, m y father came from V ilnius, but he w as in W arsa w at the time, and so they b oth came o v er. But m y mother knew that b o oks were goo d. Sh e dragged me to the library , but sh e didn ’t ha ve to drag v ery m uch. It w as just nice to go there eve ry Saturda y morning. By the time I got to New Y ork, I was indep enden t— ev en though I was only 10 y ears old. The thing ab out New Y ork is that children we re totally in dep end en t. Sampson: Y ou liv ed in the Bronx? Olkin: I lived in th e Bronx on Arth ur Aven ue and 179th Street. Sampson: Y ou ﬁn ished elemen tary s c ho ol in New Y ork City? Olkin: I ﬁn ished elemen tary sc h o ol in New Y ork Cit y and then wen t to middle sc ho ol follo wed by Dewitt Clin ton High S c ho ol. That wa s a v er y goo d high sc h o ol at the time. Dewitt Clin ton had an an- nex and it was th at annex w hic h b ecame the Bronx High Sc ho ol of Science. They had sup erior teac hers. A CONV ERSA TION WITH INGR AM OLKIN 3 Fig. 1. Ingr am Olkin, 2001. The math teac her actually published in The Am er- ic an Mathematic al M onthly . There w as a m ath club in high sc ho ol, and there was m usic. An d at that time I pla y ed trom b one, and I was in the orchestra. Sampson: Y ou w ere an only c hild? Olkin: I was th e only c hild. My mother actually lost sev eral c hildren , some in miscarriage, and one c h ild at a yea r old. I was really somewhat of a last c h ance b ecause th ey were getting older by the time I was b orn. My mother wa s 36. A t th at time, giving birth at that age wa s probably already precarious, Sampson: Y our dad w as in the jewe lry busin ess in W aterbu ry? Olkin: Y es. In those y ears, it w as v ery customary for immigrants t o ha ve a relativ e or f r iend i n the United States wh o would sort of set the stage for them. My father had a v ery close friend who had a jew elry business in W aterbur y , and my father wa s a jew eler in Europ e. So he had a job ready-made in W aterbu ry and th at’s ho w w e got there. I w as b orn in 19 24 and this is pre-depression. When I w as a c h ild, my father actually was out of wo rk for quite a bit of time thou gh m y mother didn’t kno w it. He nev er told her, and somehow w e m anaged. But it w as clear that W aterbur y was somewhat of a dead- end in the jew elry busin ess, which is one of the ﬁrst to go du ring any depr ession in any case. But there w as a second factor of wh y we mov ed to New Y ork. By this time, I w as getting close to 10 years old, and the question w as what would the fu tu re brin g in terms of college? C on n ecticut was n ot known at that time—w e’re talking ab out th e 30s—for lots of go o d state univ ersities. And, in fact, I h ad a cousin who w as teac hin g at Cit y C ollege of New Y ork, and the general view was that Cit y College wa s the place where I could get an edu cation. W e m o ved to New Y ork in 1934, and it was clear that I w ould go to Cit y College b ecause that w as the only p lace op en in the immediate p ost-depression era. NEW YORK Sampson: Did y our father ev entuall y ﬁnd jewe lry w ork in New Y ork? Olkin: Y es. In the jewelry trade, they’ll ha v e one big store with little stalls. And h e had a stall on Canal Street whic h had a jewe lry district and one on 47th Street. Throu gh ou t his life he con tin ued to w ork in the jew elry trade. Sampson: F rom th e wa y you describ e it, it sounds lik e your mother was more inﬂuential in y our up- bringing than y our father. I s that a fair statemen t? Olkin: Th ere’s no qu estion ab out that. My m other w as a very str ong woman. She had ﬁrm id eas, an d she was also a very activ e t yp e. My father was a v ery gent le kind p erson, whom eve ryone in the fam- ily and the extended family thou ght was great. No- b o d y ever had an un kind wo rd ab out m y f ather. My mother w ould generate diﬀerent reactions fr om dif- feren t p eople, bu t she w as a very p ositiv e force. Sampson: Did she h a v e any in ﬂuence in yo ur going in to the mathematical sciences? Olkin: No, I ha v e to sa y in thinking ab out my u p- bringing and the up bringing of ou r c h ildren, there’s a distinct diﬀerence. My mother did not kn o w wh at I was d oing in colle ge or in high school. But she b e- liev ed that wh atev er I did w as great, so I had com- plete full sup p ort for whatev er I wan ted to do. Sh e had faith that no matter what I wo uld do, it would b e ﬁne. And no w I think of our o wn actions w ith our c h ildren. My d augh ter wo uld come home and sa y , “Ok, I’m taking math.” I would say , “W ell, w hat 4 A. R. SA MPSON Fig. 2. Ingr am Olkin and his p ar ents, Julius and Kar ola Olkin, 1940s. math?” She w ould sa y , “Algebra.” I’d say , “What are yo u stud ying in algebra?” and s o on. There’s something v ery p ositiv e ab out u nequivocal supp ort that r eally is v er y healthy . Early on in ev ery Jewish family , there wa s alwa ys the qu estion of whether y ou’re going to b e a do c- tor. And it b ecame very clear at an early age that I did n ot like anything conn ected with blo o d. Thus I nev er had to go thr ough th e p erio d of “m y s on, the do ctor.” My mother alwa ys said, “W ell, he do esn’t lik e medicine” and that left me out of that. Sampson: As an aside, I am in terested in y our fam- ily’s inv olvemen t w ith Ju daism. Olkin: My parent s w ere what yo u would call mild ortho do x. Our house w as ko sher, but m y father, when times w ere bad, d id go to w ork on Satur da y . W e alw a ys observed all the holida ys , and he did go to the syn agogue on Saturdays a lot of th e time. And I wo uld accompan y him. I wa s certainly Bar Mitzv ahed , and marr ied in the syn agogue, and our house was what y ou might call med iu m orth o do x. Sampson: Let me come b ac k to y our h igh sc ho ol again. I kn o w y ou once told me that when they asked y ou in y our high sc ho ol y ear b o ok w hat y ou in tended to b e, yo u said, “a statistician.” Olkin: Th at is correct. It’s clear n o w, to b oth of us, that I didn’t kno w anything of what b eing a statisticia n meant. Sampson: Are y ou an y wiser 65 ye ars later?! [Laugh ter] Olkin: That’s a go o d question! I ha ve a feeling that the math club had lots of diﬀeren t problems, and I susp ect th at there were some problems in statistics that we wen t through th at must ha ve cap- tured my imagination. But I cannot no w repro du ce ho w I ev er c hose th e term statistician. I ma y b e the only one in the profession who in high school said that they w an ted to b e a statistician, and it came true. No w, it’s interesti ng to go bac k to the high sc ho ol yearbo ok, which I did. Sev eral p eople were w ell kno wn. F or example, James Baldwin, the au- thor, w as at Dewitt Clinton High School. I c hec k ed to see what he said he wa nte d to b e. He s aid he w an ted to b e a writer. Th ere w ere sev eral others whose stated goals in high sc ho ol came tru e in later life. S omehow, there ma y h a v e b een some germs of what p eople thought th ey m igh t b e intereste d in, and I susp ect that s tatisticia n just meant some p art of mathematics without my r eally knowing what it w as. Sampson: Y ou started college at CCNY and yo u had to tak e a break for the war, b efore return ing to CCNY? Olkin: I graduated Dewitt C lin ton High S c ho ol in 1941. I started in CCNY and then in late 1942, the go v ernment—and I don’t kno w ho w this came to me—said they were interested in students in mathe- matics, p h ysics and p ossibly engineering, going into radar, meteorology , and languages. T here was some- thing called the AST P , the Armed Services T r ainin g Program. Ev en though math ma j ors w ere d eferred, I though t th at ev en tually I wo uld b e drafted. Ev ery- b o d y was drafted at the time, and so I enlisted in the Meteorolo gy Program. I w as indu cted in F ebru - ary of 1943. I w as a w a y from 1943 to 1946, and then I was disc harged. My disc harge w as in C alifornia, but I return ed to New Y ork and ﬁnish ed Cit y Col- lege in 1947. After that I wen t to Colum bia for a master’s de- gree which wa s a one-y ear program. I started in A CONV ERSA TION WITH INGR AM OLKIN 5 Fig. 3. Mr. and Mrs. Selby R obinson (b ottom) with his former City Col le ge of New Y ork students: Ingr am Olkin, Herman Chernoﬀ, Herb Solomon (top), Stanfor d, 1971. the summer of 1947, and then ﬁnished the follo w in g y ear. The summer of 1947 wa s one of the ﬁr s t p ost W orld W ar I I classes and it was an exciting time at Colum bia. Many of the students who were there b e- came close life-long friend s includ ing Bob Bec hh ofer, Milton S ob el and Rosedith S itgrea ves. Also that w as a time when Hotelling had m o v ed from Columbia to Chap el Hill (in 1946). While at Columbia, I applied to Chap el Hill for the Ph.D. program and for a fel- lo w s hip, and we then mov ed to Chap el Hill in 1948. GRADUA TE S CHOOL A T U. NORTH CAROLINA Sampson: T ell me a bit ab out y our exp eriences at Chap el Hill. Olkin: I think one h as to thin k somewhat in terms of the history of s tatistics. Nineteen forty-six was the time when statistics w as really b eginnin g in s ev- eral p laces. C olum bia had a department in 1946, with Ab raham W ald, Jac k W olfo witz, T ed And erson and How ard Lev ene, and I think Henry Scheﬀ ´ e w as there at the time. Princeton did not h a v e a Depart- men t of Statistics, but it had S am Wilks and John T uk ey . Berk eley , of course, had Jerzy Neyman and Eric h Lehmann . Th ey did n ot h av e a d epartmen t in 1946—that came later. But they h ad a statis- tical lab oratory and man y eminent facult y . Chicago w as another p lace; Io wa State; North Carolina State started; but Chap el Hill was a galaxy of facult y . Harold Hotelling w as the leader. Herb Robbins w as there, W assily Ho eﬀding was ther e, as were R. C. Bose, S. N. Ro y and P . L. Hsu, although Hsu wa s on lea v e and it was not clear whether he was com- ing bac k. William Co c hr an w as at Raleigh, but came to Chap el Hill on man y o ccasions. Bill Mado w was there. Sampson: Gertru d e Cox? Olkin: Gertrud e Cox w as the head of the en tire In- stitute, and sh e wa s h oused more at Raleigh. In any case, Chap el Hill was a ph enomenal place. The stu- den ts we re great. At that time, Ind ia h ad sup p orted a lot of studen ts to s tu dy in the Un ited States. This w as really the “heyda y” of Indian statistics in the United S tates. Ra j Bahadur wa s there as a student. S. S. Shrikh ande, Gopinath Kallianpur and D. N. Nanda were students, and later on, K. C. S. Pillai, Shant i Gu p ta, Ram Gnandesik an and Go vind Mud - holk ar. I can go on, but I don’t remem b er them all. Sampson: Su dhish Ghryr e was someone th at you w ork ed with? Olkin: He came a ye ar later. But he wa s there. It w as really a ﬁrst rate group. And then there w ere American stud ents: Ralph Bradley was a student, Mey er Dwass was there, Joan Rosenblatt, Morris Skibinsky , Su tton Monro, and many others. There w as a lot of camaraderie among th e students, and there were v ery few barr iers b et ween facult y and student s. Part of it wa s that the faculty w as young. 6 A. R. SA MPSON Fig. 4. Ingr am Olkin and Bil l C o chr an, 1973. They were only a few y ears older than some of the student s. W e would gather across the street from the department in mid-afterno on. There w as a little store where p eople could get ice cream. In terms of classes, Robbins’ batting a ve rage wa s alwa ys some- where in the order of 75% to 80%, by whic h I mean 75% or 80% of his lectures w ere r eally sup erb. And Ho eﬀding wa s a diﬀerent t yp e of lecturer. Ho eﬀd- ing’s lectures were b uilt-up from the small to the b ig, whereas Robb ins tr ied to in s till some excitemen t, and to comm u nicate a lot of the ke y ideas. Ho eﬀd- ing was m uc h m ore m etho dical, as w as Bose. Th ey co vered the int ricacies, and th eir lectures were not really exciting, bu t the material was exciting. It w as all new. One of the things that p eople d on’t quite realize is that there were no b o oks. This was 1948. Cram ´ er ( 1946 ) wa s a v ailable. The Kendall b o oks w ere a v ailable (Kendall, 1944 , 1946 ). William F eller’s b o ok in probabilit y did not come out u n til 1950 (F eller, 1950 ). Robbins h ad a cop y of the man uscript, I su sp ect, and taught some parts of it, but added a lot. S o eve rythin g wa s new. Hoeﬀdin g’s lectures in nonparametric we re all h is o w n which then app eared subsequently as his pap ers, and the same with Bose and Ro y . Hotelling was n ot a great lecturer. But his lectures were pleasant to hear at times b ecause he w as very erudite, an d I’d lik e to sa y he sp ok e in pr ose. Th er e w ere alw ays complete sentences and they were b eautifully put together. Sampson: Th at’s th e newsp ap er man in h im. Olkin: T h at’s right. He was a jou r nalist ma j or at the Universit y of W ash ington, and his wr iting is clear, and b eautiful, and this is true in the pa- p ers th at he publish ed. Hotelling was an in tellectual force. He was in terested in push ing student s in ev ery p ossible w a y and fostering th em. Sampson: What do y ou mean b y p ushing s tu den ts? Olkin: W ell, let me giv e y ou an illustration in my o w n case. I w ant ed to tak e a course in multi v ariate and it happ ened th at at the time that I w ant ed to tak e the course, th e course w as not giv en. I had nev er tak en multiv ariate. I had no idea w h y I wan ted to tak e multiv ariate, but I did. I w ent to Hotelling and I told h im that I wan ted to tak e m ultiv ariate, and he said, “W ell, why don’t you get P . L. Hsu ’s notes from last year, and study them on your o wn ?” “And then we’ll h a v e an oral—y ou can lecture on it.” I’v e forgotten exactly how he w as going to grade it. As it turned ou t, one of the other student s, W alter Deemer, h ad the same desire. The t w o of us got Al Bo wker’s copy of P . L. Hsu ’s notes, whic h w e wen t thr ough, and we work ed on it. T hat’s where W alter and I recognized that s ome of th e re- sults on Jacobians were things that we could im- pro v e and expand on , which w e did. But I w ant to come bac k to the role Hotelling pla y ed. A t th e end of the quarter, W alter and I ga ve a lecture on wh at w e had r ead and accomplished on Jacobians. Hotelling said, “Y ou m ust p u blish this.” W e nev er th ough t ab out p ublishin g this. W e we re actually v ery yo ung, and timid, and so we left it more or less in Hotelling’s hands. Both W alter and I said this is b ased on Hsu’s lectures, and mayb e Hsu should b ecome a coauthor or even author it himself. In any case, Hotelling said he w ould write to Hsu. A t this time the Un ited States d id not ha v e in ternational relations with Ch ina, and you could not send a letter fr om the U.S. to China. Hote lling w rote a lette r to Egon P earson and ask ed Egon Pe arson to transmit the letter to Hsu. Later, Hsu r eplied in the same w a y and indi- cated th at he d id not w ant to b ecome a coauthor. He thou ght w e had extended the method ology , and A CONV ERSA TION WITH INGR AM OLKIN 7 that we should pub lish it. How ev er, h e had a few though ts ab out one of the theorems at the end, and he asked if we wo uld includ e sev eral paragraphs that he su p plied. W e did, and our pap er came ou t in Biometrika (Deemer and Olkin, 1951 ). In general at ev ery lecture, at ev ery public meet- ing, Hotelling would pu blicly men tion w hat students at C h ap el Hill had d one, so that the profession w ould kno w what th ey h ad accomplished and in this w a y he would try to h elp their careers. Sampson: Did he hav e so cial ev en ts at his hous e to encourage in teractions among stu d en ts and facult y? Olkin: Hotelling had an afterno on tea, th e s econd Sund a y of every m on th, wh ic h, if I recall correctly , the stud en ts lab eled “Hotelling’s T .” F acult y and student s w ould meet at his house. Hotelling had an encyclop edic memory . His conv ersation w as nev er what you migh t call small stuﬀ. Y ou would ask him some sim p le q u estion s uc h as “I see that they’re tearing up F ranklin S treet?” And Hotelling would sa y , “It wa s in 1824, that F ranklin Street was ﬁrst dev elop ed,” and he w ould giv e y ou an en tire history of w hatev er it is yo u w ere discussing. Hotelling’s wife, Susann a, wa s a viv acious p erson w ho w as in - terested in the s tu den ts, h elp ed everyb o dy , and wa s extremely so cial. Sampson: O n e has the sense that p erh ap s Hotelling migh t b e a bit in timidating to h a v e con v ersations with? Olkin: I don’t kno w if in timidating is right, b ut y ou wouldn’t talk to Hotelli ng without recognizing that you we re talking to someb o dy wh o had ac hieve d a lot. There w as a resp ectful tone that on e alwa ys main tained with Hotelling. He could generate a re- sp ect for the ﬁ eld, and an inspir ation for working in the ﬁeld. Sampson: Did y ou k eep in con tact with Hotelling after you left North Carolina? Olkin: Afterwa rds b ecause of m y sense of app reci- ation for him, when Hotelling turn ed 65, I suggested to sev eral p eople from Chap el Hill that we h av e a festsc h rift in his honor. Gh urye, Ho eﬀding, Mado w, Mann and I we re the editors and w e invited man y p eople to sub m it a pap er (Olkin et al., 1960 ). I re- call v ivid ly the r esp onse from Joseph Do ob. W e had ask ed Do ob if he w ould submit a p ap er in h onor of Hotelling, and Do ob said, “Of cour s e I will subm it a pap er, Hotelling sa v ed m y life.” As it turn ed out, in 1939, I b eliev e, Do ob w as out of a job and Hotelling recognized his talen t and ga ve him a p osition at Colum bia. He ga v e Do ob a job and h e br ough t W ald to Colum bia. Not man y p eople would b ring some- one as go o d as W ald wh o m ight b e a comp etitor. Hotelling w as ab o ve that in the sen s e that he was only interested in your s cholarly and intel lectual af- fairs. He nev er en tered into any kind of gossip—he w as alw ays just basically on an in tellectual lev el. Sampson: Hotelling was your th esis advisor? Olkin: My thesis advisors we re Ro y and Hotelling. Sampson: What w as Ho eﬀd ing lik e? Olkin: Ho eﬀdin g wa s v ery , v ery quiet. I’m n ot sure if h e came from Germany or Russia in the 30s, but he w as there with his mother. Ho eﬀding had a sibling. I thin k h e wa s a journalist w ho also came o v er at the time. Hoeﬀdin g w as fo cused on the mathematics of the ﬁ eld. He was not a p erson for small talk, so it w as ve ry hard to ha v e a con versation. I mean, you migh t giv e a whole sentence, and his reply w ould b e “y es.” Then yo u would ha v e to thin k of an en tire new sentence. Y ou did n ot get too muc h resp on s e. Sampson: Y ou later h ad translated from German one of Ho eﬀding’s works that was multiv ariate in nature. Olkin: It w as Ho eﬀding’s 1940 thesis that was on correlations, ﬁ xed marginals and b iv ariate d istr ibu- tions (Ho eﬀding, 1940 ). What we call the F rech ´ et b ound s are really Ho eﬀd ing–F rech ´ et b ound s. It’s hard to know who did w hat ﬁrst, b u t Ho eﬀdin g def- initely h ad the result there. RESEARCH A ND COLLABORA TIONS Sampson: I’d lik e to no w talk ab out your researc h. Y our pu blished researc h is quite div erse, but there seems to b e one constant. And with v ery few excep- tions, the v ast ma jority of it is p u blished with coau- thors. And this b egan relativ ely early in yo ur career. It wasn’t something that y ou pic k ed up later. What is it ab out join t r esearc h in statistics that s o ent ices y ou? Olkin: I think ther e are pr ob ab ly t wo factors. One factor is, I think, I enjo y wo rking with p eople, and y ou h a ve to enjo y working with p eople if you’re go- ing to coauthor pap ers. The other is that I foun d that with coauthors, the sum is r eally more than what eac h ind ivid ual pro vides. This is b ecause in some w a y eac h one complemen ts the other, and so y ou mov e ahead muc h faster. It av oids doldrums. If one p erson is do wn, the other p erson hop efu lly is up and th en you con tin ue pro d ucing. Sampson: When I lo ok thr ough y ou r list of coau- thors, there is not a whole lot of common threads among th em. Th ey are a tremendously div erse group. 8 A. R. SA MPSON Fig. 5. Al Marshal l and I ngr am Olkin, 1973. Olkin: That’s right. Most of the coauthorships were started in an u ndesigned wa y . That is, they w ere not though t of that I wo uld work with someone. Some- ho w an idea came up. I ma y ha v e talk ed to some- b o d y ab out it, or they talk ed to me ab out it. Then suddenly it b ecame really an inte resting problem that the t wo or three of us w ould work on. There’s only one pap er that I recall that w as d one b y dif- feren t groups. There’s a pap er wh ic h Seymou r Sh er- man calle d the “Chicago Six,” and that pap er did ha v e contributions from diﬀerent p eople without our starting to w ork together (Das Gupta et al., 1972 ). This m igh t b e a goo d p oint to tell you ab out a particular collab oration. Y ou mentioned that I had a lot of coa uthors , which is correct. But I also ha ve one coauthor with whom I’v e b een in v olv ed for 50 y ears. Sampson: Al Marshall! Olkin: Y es and that h app ened by acciden t. I w as on sabb atical lea ve at Stanf ord from Mic higan State in 1958, and I h ad just pub lished a pap er with John Pratt on a multiv ariate Chebyshev inequalit y (Pratt and O lkin, 1958 ). I’m not ev en sur e the p ap er had app eared. In any case, Al Marshall’s thesis, at the Univ ersit y of W ashington with Bill Birnbaum wa s on a multiv ariate Cheb yshev inequalit y . I b eliev e that I sa w an abstract of Al’s, and I think I wr ote to him ab out the pap er that Pratt and I had published —they did not signiﬁ can tly o v erlap b y the w a y . In an y case, Al, w h en h e graduated, w as ei- ther a p ost-do c or a visitor at Stanford, and as it turned out by c h ance our oﬃces w ere next d o or to eac h other. (As an aside, I s till remem b er the p eople in the ﬁve oﬃces adjacen t to me. T here w as George F orsythe, Bill Mado w, myself, Al Marshall, Ben Epstein, and Bob Bec hh ofer. I could talk to George F orsyth e ab out computing, with Bill Mado w on samp ling, with Ben E p stein on the exp onen tial distribution, and with Bob Bec hhofer on r anking and selection, but mainly with Al.) So Al and I qu i- etly s tarted w orking together and w e wrote 1 or 2 pap ers, an d then I r eturned to Mic higan S tate. I w as bac k at Stanford in 1961 and we started a collab- oration at that time that mo ve d fr om Ch eb yshev inequalities, o ccasionally to other inequalities, and then later to S ch u r f u nctions and ma j orization. W e w ere able to cont inue in a serious w a y b ecause w e b oth sp ent a year at Cam bridge in 1967–196 8. By that time, the s eeds of our ma j orization researc h had b een established. At Cam b ridge, Al and I ga v e lec- tures on m a jorization and b egan the work that led to our b o ok (Marshall and Olkin, 1979 ). F r om 1967 to when the b o ok app eared in 1979—that’s o ve r a 12 or 13 yea r p erio d—we were collecting resu lts, work- ing together, and th en u ltimately wrote the b o ok. Sampson: Y ou also s p ent a lot of time at Bo eing. Olkin: Oh ye s. I wa s a consultan t at Bo eing whic h means that they made the opp ortunit y av ailable for me to come up wheneve r Al and I wan ted to. A CONV ERSA TION WITH INGR AM OLKIN 9 The Bo eing Scientiﬁc Researc h Labs w as really a great place. It had a galaxy of go o d p eople. In addition to Al, F rank Prosc h an w as there, as w ere Jim Esary , George Marsaglia, S am S aunders, Roger W ets, Da ve W alkup and sev eral others. And the visitors w ere Dic k Barlo w, Ron Pyk e, Victor Klee, Bill Birn baum and m yself. It w as alw a ys an exciting place. Sampson: How m uch time did yo u sp end there? Olkin: My recollection is that w e‘d try to get to- gether eve ry few wee ks. Al w ould come do wn to Stanford for a few da ys, or I would go up to S eattle for a few da ys. Th e visits were not extensiv e visits, but w e would often work on th e germ of the idea, and then tr y to d evelo p it a little bit individually , and th en get toget her again. Sampson: And th at was b efore email bac k and teleconference made comm unications so m uc h eas- ier. Olkin: Ev er y th ing was hand written, of course. Ho w- ev er, y ou do hav e to recognize that it’s very diﬃcult to write a b o ok just by going b ac k and f orth this w a y . So w e actually sp ent time toget her a wa y . Beside the yea r in Cambridge, w e were also in En gland for a ye ar and we w ere in Augsbur g and Zuric h f or thr ee mon ths together. W e were in a num b er of places for more extended p erio d s then just these visits. Sampson: Did yo u sp end time in British C olum bia? Olkin: I sp ent a quarter ther e, when Al w as on the facult y at the Un iv ersit y of British Columbia, Also I d id visit on o ccasion for shorter p erio ds. What we tried to do w as wo rk in short sp urts and then ha v e culmination b y ha ving longer p erio ds together. Sampson: Y ou also did with Al a fair n umb er of pap ers on d istributions, and some more recent w ork has b een on families of d istribution. And of course there’s the w ell-kno w n Marshall–Olkin biv ariate and m ultiv ariate exp onentia l distributions (Marshall and Olkin, 1967 ) that are w idely used. Olkin: W e started w orking on the question of how to generate b iv ariate distrib utions that h a v e certain kinds of nice pr op erties, and our ﬁr s t instance of that w as wh at’s no w called the Marshall–Olkin bi- v ariate exp onentia l. More recen tly we plann ed to try to write a b o ok on nonnorm al biv ariate distri- butions. W e started, and then w e said w e’d b etter ﬁrst wr ite a c hapter on univ ariate nonnormal dis- tributions. W ell w e started that and we found th at w e wr ote one c hapter, and then another c hapter, and then another c hapter, and we found we nev er got to m ultiv ariate distributions! W e are no w ab out to pu blish a b o ok ent itled Life Distributions: N on- p ar ametric, Semip ar ametir c and Par ametric F ami- lies. The b o ok is almost complete. W e’re in the pro- cess of pr eparing the graphs for the b o ok, and wh en that h app ens , the b o ok w ill b e ﬁnished . So m y guess is within the next three mon ths, w e will b e s en ding it oﬀ for publication. Sampson: T o go bac k to y our work on the exp o- nen tial, my recollection is th at b efore y ours and Al’s w ork, p eople did not do a lot of work lo oking at prop erties of th e exp onen tial distrib u tion in the un i- v ariate case and tryin g to extend them. W as y ours the ﬁr st that lo ok ed at the memoryless prop ert y and tried to extend it to t wo-dimensions? Olkin: I th ink in th at resp ect, ours w as the ﬁrst. Ho wev er, I think y our p oin t is really an imp ortant one. What w e tried to do is look at a prop ert y of a u n iv ariate distr ibution and see in w hat wa y one could extend that particular prop ert y . Not all u ni- v ariate prop erties h a v e nice extensions to the biv ari- ate case. The memoryless p r op erty was one that h ad a n ice extension. But there are lots of other examples where we extended th e ideas of the c haracteristics of the univ ariate distribution to the b iv ariate. Sampson: More generally , y ou ha v e a r eal inte rest in solving functional equations. Olkin: F un ctional equations h a v e sto o d us very w ell. I’m really a prop onen t of Acz ´ el’s ( 1966 ) b o ok on fun ctional equ ations. Al and I j ust submitted a pap er in honor of S. N. Ro y that’s coming out in a sp ecial issue of the Journal of Statistic al Plan- ning and Infer enc e. And there w e solv e a n umb er of functional equations. Y ou might ask the question, wh y is th e W eibull distribution so p opular and ho w did W eibull come ab out it? W ell, it turns out that W eibu ll actually came to it from data analysis. How- ev er, there is a nice rationale in terms of the distribu- tion as th e solution of a functional equ ation. When y ou ha ve s cale-parameter families and prop ortional hazard families, the W eibu ll is the coincidence of the t w o families. This u ses a fu nctional equation s olu- tion again. Sampson: Ingram, what do y ou see as the secret of y our collab oration w ith Al in terms of w orking st yle? I think yo u told me h e is a cr aftsman b y na- ture and for instance lov es to build fur niture. Is that st yle something th at carried ov er int o ho w you t w o in teract? Olkin: W ell, Al is deﬁnitely more of a mathemati- cian than I am, b u t I don’t thin k that’s the issue. 10 A. R. SA MPSON Fig. 6. Cy Derman, Ingr am Olkin and L e on Gleser, 1986. I think that Al h as a un ique w a y of lo oking at prob- lems, and it happ ens to b e sligh tly d iﬀeren t from mine. And so Al b rings an exp ertise that I would not h a ve, bu t I bring a diﬀeren t view of some of the problems that Al w ouldn’t ha v e. I think it’s fair to sa y that Al pr obably could write th e b o ok without me, and that I probably could not wr ite the b o ok without him. But that the end result is v ery diﬀer- en t fr om what it w ould h a ve b een had h e ju s t writ- ten the b o ok alone. I think the wid e applicabilit y is something that comes ab out f r om our collaboration. If you thin k ab out the ma jorization b o ok, there are a lot of tec hnical asp ects, but there’s a lot of b readth and scop e—applications in probabilit y and statis- tics, in matrix theory and in com binatorics. I think a lot of that is the kind of thing that I w ould bring to the b o ok, and a lot of the results w ould b e things that Al deve lop ed on his o w n or we did toget her. Sampson: My sense is, in add ition, that the t wo of y ou w ere goo d p erson al friends. Y ou s hare a lot in terests in common. Olkin: W e do. W e go to concerts together, and our wiv es ha v e gone toget her on diﬀeren t trips, and w e’v e gone toget her on diﬀeren t trip s. W e used to do this a lot more than no w b ecause we eac h h av e large families that we ’re inv olv ed with. W e ha v en’t visited eac h other as muc h as we u sed to, but in the early da ys w e certainly sp ent a lot of time together. Our collab oration has b een uniqu e in m y life in that it has extended o ve r such a long p erio d of time and o ver so many diﬀerent p ap ers and t wo b o oks. It’s b een very , ve ry fruitfu l. Indeed, some p eople think we’re one p erson wh ose name is h yphenated as Marshall–Olkin. I ha ve to tell y ou that Al once sen t me a CD that he h ad foun d. The cov er featured m usic b y a comp oser whose name is Ingram Marshall. Sampson: Y ou’v e had a num b er of other coauthors that y ou’ve done a lot of pap ers with in cluding Leon Gleser and Larry Hedges, b oth of wh om were y our Ph.D. stud en ts. Olkin: L eon and I o verlap a lot in inte rests and skills, and that’s a v ery n ice collab oration b ecause w e’re b oth attuned to the same kind of orien tation. Leon is really a ve ry go o d problem solver and to- gether we ’v e work ed on quite a num b er of diﬀerent problems. I’v e alw a ys enjoy ed that collab oration. I’v e probably published more with Leon, after Al, than with an yb o d y else. MET A-ANAL YSIS Olkin: Larry and I started w orking together as a resu lt of my app oin tmen t in the S c ho ol of Edu- cation at Stanford and my early in vo lv ement with meta-analysis. Dur in g my life I’ve tried ev ery ten, ﬁfteen y ears, to b ecome in v olv ed in something a lit- tle diﬀerent, a little orthogonal from what I h ad b e- ing doing. Beca use once yo u con tinue it’s v ery hard A CONV ERSA TION WITH INGR AM OLKIN 11 to ke ep up the excitement in the ﬁeld after yo u’ve b een pub lish ing, and it’s n ice to start on something new. And meta-analysis w as one of the areas that w as very diﬀeren t. And that started in a v ery inno- cen t wa y . It b egan b ecause a colleague in the S c ho ol of Edu cation said to m e, “Ingram, there hav e b een literally hundreds of pap ers on th e eﬀects of cer- tain teac hing asp ects, bu t these studies are sm all and th ey’re not s igniﬁcan t. Is there any wa y to put together this mass of in d ividual studies?” S o that started me on the ﬁeld of meta-analysis. Th at w as in the early 70s, and I wrote one pap er in the 70s b y m yself, and then I started w ith stud en ts. And Larry was the ﬁrst s tu den t. He wrote a dissertation on meta-analysis. Sampson: He w as a stud en t in Education? Olkin: Y es, at that time w e h ad a program in edu - cation. It was comparable to wh at y ou m ight call biostatistics, only it was edu cational s tatistics in whic h th e students would tak e all the courses in the Statistics Departmen t, more or less thr ough w hat w ould b e a normal Ph.D. p rogram, bu t the disserta- tion w ould b e on an educational topic. Th ey ma y not ha v e tak en ev ery single course, but they to ok most of the courses. Larry w as in that situation. And we started writing this b o ok on statistical metho d s for meta-analysis. Sampson: W as that while he w as still a studen t? Olkin: No. The b o ok w as p ublished in 1985 (Hedges and Olkin, 1985 ), and m y recollection is he may h a v e ﬁnished h is Ph .D. in 1981. Bu t we started working on it almost immediately thereafter. And the reason w e s tarted working on it w as we r ead many of the pap ers that were b eing published, and we recognized that th ere w as not muc h statistica l metho dology b e- ing used in th em. There w as a b o ok earlier than ours b y Glass, McGra w and Smith ( 1981 ), and it had some statistics. But it didn’t hav e a systematic statistica l dev elopmen t. So that’s wh at we pr o vided in our b o ok. Our b o ok seems to ha v e had a catalytic eﬀect on meta-analysis. Afterwa rds m y increased in- v olv emen t w ith meta-a nalysis brough t me in to the sphere of the medical p r ofession. These were really the p eople who are doing medical researc h and try- ing to come up with conclusions ab out the state of their ﬁeld, so it has b een very exciting for m e. Sampson: Y ou’v e giv en and still cont inue to giv e short cours es on meta-a nalysis. Olkin: The sh ort courses really started with one of the medical originators of meta-analysis, a p h ysi- cian by the name of T om Ch almers. Ch almers was a deligh tful p erson, and he and I m et when he was at Mt. Sinai Hospital. He w as the h ead of wh at I think w as called a T ec hnical Assessmen t Division. And they had a site visit, and I w as on that s ite visit committee. I enunciated a num b er of factors in fa- v or of meta-analysis th at the hospital, at the time, hadn’t recognized. As a result of that, Chalmers and I b ecame friends , and we started giving some sh ort courses. It w as a m arvelous collab oration b ecause he w ould sp eak for an hour and a h alf on medicine and then I w ould sp eak for an hour and a half on the statistica l asp ects of the studies. And then we w ould go bac k and forth, and we did this for t w o da ys, at seve ral diﬀerent places. Chalmers started his lecture in a w ay that I couldn ’t b ecause I’m not a physician. His ﬁrst few sent ences w ere something like, “I got int o the ﬁeld of m eta- analysis b ecause I r ealized that I was killing pa- tien ts.” W ell, of course, when a physici an sa ys that it creates qu ite a stir in the aud ience. T om h ad a v ery wo nderf ul wa y of presen ting the material. Un- fortunately , h e died a few y ears ago of prostrate can- cer. One of his prot ´ eg ´ ees is a physician that I w ork with a lot no w, who’s very go o d, and a lo vely col- league. That’s Joseph Lau, who’s at the New En g- land Medical C en ter. Joseph h as a group w orking there, and they’re one of th e pro du cers of a lot of meta-analyses. Sampson: Where ha v e you give n some of y our short cours es—p articularly the in ternational ones? Olkin: T o start with, several were for the Ameri- can S tatistical Asso ciation. I’v e give n sh ort courses in S in gap ore and in Hong Kong and I’v e giv en short courses throughout Europ e, including Sw itzerland, Spain, Croatia, Holland, and Austria. I ma y hav e left out a f ew places—I hav e not given a s hort cours e in F rance. Some of these courses I taugh t with J oseph Lau. EDITORIAL CONTRIB UTIONS Sampson: I ngram, let’s talk ab out y ou r editorial w ork. Y ou’v e had a life-long in volv emen t in v ari- ous editorial capacities, in statistics journ als, edu- cational jour nals, and mathematics j ou r nals. Why ha v e you chosen to dev ote so m uc h ener gy to these purp oses? Olkin: W ell, earlier on I mentio ned m y mother to ok m e to the libr ary . Sampson: Y ou are blaming it on yo ur mother?! [Laugh ter] 12 A. R. SA MPSON Fig. 7. Ingr am Olkin sp e aking on meta-analysis, University of Pittsbur gh, Mar ch, 1998. Olkin: No, I’m not going to blame it on my mother. I d on’t kn o w wh y I w as so inv olv ed in journals. When I was at Chap el Hill in 1946, it w as not a ma- jor enterprise to r ead all the journ als. They would come to the library and I wo uld b r o wse through the journals. I found that a lot of times br o wsing through an article w ould ring a b ell ab out something that wa s similar to what I was doing. Th at wa s the b eginning of my in terest in j ournals. A t Stanford, I wa s on the Libr ary Committee and I w as alw a y s in v olv ed in trying to b uild up the libr ary . But that do esn’t relate to the editorial w ork. I w as an asso- ciate editor for JASA , but that also wasn’t the real catalyst. It w as when I b ecame editor of the An- nals that I started th inking a lot more ab out jour - nals. What b ecame clear to me is that the gro wth in statistics from the time I graduated in 1951 to the time I w as an editor of th e Annals in 1972, had resulted in the A nnals ’ pu b lishing ov er 2,000 pages a y ear. No editor could really review all the pap ers . The editor w as really a manager. Y ou just shuﬄed the p ap ers to diﬀerent asso ciate editors and you in terv ened in questionable cases. Otherw ise, when things were clear-cut, y ou merely accepted what the asso ciate ed itor su ggested. Probabilit y and statis- tics w ere b oth gro wing and it was ju st unreasonable A CONV ERSA TION WITH INGR AM OLKIN 13 to think of trying to kee p the t wo together. Now, there wa s a con tro v ersy at this time. The question w as should we b egin to splinte r into sev eral separate groups? There w ere camps on b oth sides. There we re clearly p eople who though t that the Annals of Math- ematic al Statistics should in clude b oth statistics and probabilit y . I thought from a practical p oint of view, the gro wth wa s to o big in b oth ﬁelds, so I prop osed splitting th e Journal. W e w en t to the Coun cil. After discussion, w e did sp lit the Jour nal. I r emained as editor of the Annals of Statistics and Ron Py ke b e- came the ﬁrst editor of the Annals of Pr ob ability . Of course, in the year 2005, n o one would ev er question going bac k to a single jour nal. The t wo ﬁelds hav e gro w n and, in fact, they’v e splintered even more into the Annal s of Applie d P r ob ability and now into the Anna ls of Applie d Statistics . Sampson: I’m curious, was op p osition to the split more from the statisti cal types or th e probabilists? Olkin: The statisticia ns d id n ot w an t pr obabilit y split oﬀ. But more than that, they also felt that splin tering of the ﬁeld was not go o d. I wa s not of that opinion for a v ariet y of reasons. What I sa w w as that w h en the so ciet y d ecided not to pu blish a topic journal suc h as th e Journal of Multivariate Analysis , or the Journal of Time Series , or the J ournal of Se quential Analysis , the commercial publishers w ent in to that ﬁeld. An d I felt it was b etter for the so ciet y to sp onsor these topic journals, b ut others felt that this sp lin tered the ﬁeld. My p osition was that it was going to b e splintered by the commercial publish ers in any case, and that I w ould r ather it b e sp lin tered within the So ciet y’s con trol. I lost that b attle. Also the b eginnings of splint ering led me to start to thin k ab out j ou r nals that the So ciet y m ight feel comfortable with. Let me go back in time for a mo- men t. There was a jour nal in engineering, T e c ho- metrics ; there were sev eral journals in biology , Bio- metrics , and to some degree Biometrika ; there we re journals in psy cholog y , Psychometrika ; and journals in economics, Ec onometric a . There w as nothing in education, so Mel Novic k wh o was in psychomet- rics at Stanford and I talk ed a lot ab out p ossibly starting a jou r nal in educatio n. He and I agreed on t w o p rinciples; one was that it should b e pu b- lished by a so ciet y , ev en though commercial pub - lishers had app roac h ed u s; and the second wa s that w e agreed that it would b e b etter for th is to b e a join t ven tur e b et w een education and statistics, rather than either one taking it on separately . Mel’s task was to go to the council of the American Ed- ucational Researc h Asso ciation; m y task w as to go to the b oard of the American Statistical Asso ciation and to see if we could prop ose a join t ven tur e. That w as ac hieve d and Mel No vick w as the ﬁr st editor— I had not w ant ed to b e inv olv ed as an editor—and the Journal of Educ ational Statistics (now called the J ournal of Educ ational and Behavior al Statis- tics ) was launched in 197 6. Sampson: What w ere the issues inv olv ed in the creation of Statistic al Scienc e and how did it get started? Olkin: Th at w as very in teresting in a v ariet y of w a ys. A group of statistic ians, without an y author- it y f rom so cieties, w ould get together. The group in- cluded National S cience program directors, b ecause they were on top of the problems in the ﬁeld, and also p eople who were inte rested in d oing something for the so ciet y . I hate to men tion members b ecause I’m going to lea v e out some, but I ’ll men tion Jerry Sac ks, Morrie DeGro ot, Stev e Fien b erg, P eter Bic kel , P aul Shaman and myself, and p eople suc h as Bruce T rumbull and Nancy Flourno y . Both Bruce and Nancy w ere NSF program directors as had b een Jerry Sac ks and Paul S haman. On e th eme that ke pt surf acing w as the need for some kind of generalist jour nal. Morrie and I, in p articular, w ere the leads in that. Morrie wa s in terested in b eing editor, and that was great b ecause h e wa s a really a sup erb editor. W e b oth h ad in mind doing some h istory , doing some in- terviews, and doing some general pap ers. What we ﬁrmly b eliev ed was that a pap er that was s u itable for JASA or the Annals would n ot b e suitable for Statistic al Sc ienc e . W e wan ted these to b e generalist t yp e article s with a w ider r eadership Sampson: Wh y did it end b eing published by the IMS? Olkin: I think th at was a decision that we made. It w as the pr actical it y of the situation. W e could get it through the IMS and w e could not get it th r ough the ASA. It w ould in v olv e a d iﬀeren t type of pr op osal to get it through the ASA. W e w ere all mem b ers of IMS. A lot of us were on the council; one of us was a treasur er. So we could get a p ositiv e r eaction f r om the Council of th e I MS , and as it turned out we did. Sampson: Statistic al Sc ienc e is almost t wen ty y ears old. Has it successfully fulﬁlled wh at was the original am bition for it? Olkin: When I talk to p eople, they sa y th ey lov e the inte rviews. They read other articles on o ccasion. 14 A. R. SA MPSON Fig. 8. ISI Me eting, Pari s, 1961. L eft side: Unknown, Elizab eth Sc ott, Jerzy Neyman, A nne Durbin, Jim Durbin, Miriam Chernoﬀ, Herman Chernoﬀ, Unknown; R ight si de: Jack Y ouden, Ingr am Oklin, Dor othy Gilfor d, Manny Parzen, Car ol Parzen, El len Chernoﬀ, Judy Chernoﬀ. It’s not a journ al that p eople read from co v er-to- co ver. But almost ev eryb o dy lik es something, and I hav e found v ery few p eople wh o sa y that they dis- lik e th e j ournal. A testament to its su ccess, I think, is th at a lot of members of the American Statistical Asso ciation are sorry that it’s n ot a journal that’s join t. Sampson: Do y ou think Statistic al Scienc e has k ept the mathematical con ten t far enough a wa y from JASA and the Annals ? Olkin: I think that it is diﬃcult in a lot of sub- jects to not b ecome to o tec hnical. Bu t I think, by and large, that th ere are enough articles that are not mathematic al, so that p eople con tinue to enjo y reading it, and keep referrin g to it. Sampson: Th ese days, one has a sense that there’s going to b e more of a mo v ement to electronic jour- nals, e-journals, that may nev er exist on pap er. Do y ou ha v e an y thoughts on th is? Olkin: W ell, it’s clear that the pr ofessions hav e to face the issue of electronic journals. One of th e questions, of course, for statistic s, is that statistics is an arc hiv al science. Not ev ery ﬁeld is arc hiv al. W e w an t to b e able to preserv e what the journ als pro- duce. I think one of the problems that will ha ve to b e faced is who is going to preserv e them and in what condition will they b e preserved? But certainly elec- tronic journ als will con tin ue to b ecome the natur al format. Sampson: Y ou mentio ned earlier that the for-proﬁt publisher s are pu tting out the journals in more fo- cused areas of statistics. W ould y ou b e advocating that the so cieties to d ay try to pic k u p that role, to o? Olkin: I ha v e tried to get so cieties to fo cus a little bit on some of these splint er topics, bu t b y-and-large A CONV ERSA TION WITH INGR AM OLKIN 15 so cieties are conserv ativ e. I thin k they fear that if they pu b lish splinter journals, the n umber of suc h splin ters could b e large. And they can’t quite face wh y , for example, the IMS should p ublish a journal in design of exp eriments, and n ot on multiv ariate, and not on sequenti al, etc. I think they are faced with the plethora of p ossible sp lin tered topics. It w as p erhap s somewhat tolerable when diﬀeren t pub lish- ers pick ed u p d iﬀeren t parts of the splintering, b ut as time has ev olv ed, these sep arate companies ha v e merged. So now, all of a sudden, in s tead of, for in- stance, ten journals f rom ten publishers , we ha v e ten journals from one pu b lisher—a sup er pub lisher. All of these s p lin tered j ournals are no w creating a r eally serious ﬁ nancial p roblem for an y lib r ary . Sampson: Th ese sup er publishers hav e v ery strong price leve rage. Olkin: That’s, in eﬀect, what is happ ening. And w e did not foresee the mergers. What we sa w w ere the splint ers. But we d id n ot see th e fact that these w ould all coalesce under one or t wo k ey pub lishers. The pr oblem is already very acute in medicine b e- cause the splinte rs hav e fu rther s p ecialized. There migh t b e a highly sp ecialized jour nal, and if there is a single do ctor stu dying that p articular area at an institution, their library may b e forced to purc hase a b u ndle of journals b y that publisher in order to obtain th e one sp ecialized j ournal. ST A TISTIC S ON THE NA TIONAL SCENE: NAEP , NISS AND NIST Sampson: Let’s talk ab out y our w ork in statis- tics and its inﬂ u ence on n ational p olicy issues—in particular, ab out y our wo rk with the NCES (Na- tional Cen ter for Education Statistics), NAEP (Na- tional Assessment of E ducational Pr ogress), NISS (National I n stitute of Statistical S cience) and NIST (National In s titute of Standards and T ec hn ology). Olkin: I w ould lik e to talk ﬁr s t ab out the Na- tional Cente r for Ed ucation Statistics. Th ere are a n umb er of go vernmen tal agencies c h arged with the collect ion of data relev an t to their particular area. There’s the National C en ter f or Education S tatis- tics, National Cen ter for Health Statistics, th e Bu- reau of Lab or Statistics, Agriculture, and so forth. One of th e th ings that b ecame clear early on w as that ev en though the National Cent er for Ed u cation Statistics has the w ord “statistics” in its title, the n umb er of whom w e would call professional statis- ticians in NCES is v ery small. Some of the other agencies ha ve man y m ore statisticians f or a v ariet y of historical reasons. In any case, in the late 1980s I had conta ct with Emerson Elliot, the C ommissioner of the National Cen ter f or Education S tatistics. W e talk ed a lot, and I comment ed to h im on the fact that not that man y statisticia ns mo ved in the sphere of education, and I though t that it wo uld b e go o d if that could b e c h anged. A t that time, the American Statistica l As- so ciation already had a fello wship p rogram—I b e- liev e it w as in existence with the Bureau of Lab or Statistics. In any case, I suggested to Emerson that there b e a f ello wsh ip p r ogram at NC ES. And Emer- son w as a v ery p ositiv e p erson in try in g to do all kinds of go o d things for NCE S , and h e th ough t he w ould b e willing to put up s ome ﬁ nancial supp ort. The ultimate conclusion to our discussion w as that a f ello wsh ip p rogram was started. He ask ed me if I wo uld b e willing to help start it, and I said I wo uld. I w as in strumenta l in getting Larr y Hedges and Ed Haertel to b e F ello ws dur ing the early p erio d. S ub- sequen tly , Julie Shaﬀer was a F ello w, Jeremy Finn w as a F ello w, and there hav e b een many others. And I thin k this b rought a b it more connection b et w een statistics and education. But it was also at NCES that I b ecame inv olve d in some of the tec hnical issues in the Natio nal As- sessmen t of Ed ucational Progress (NAEP). NAEP has b een in existence for many years, and is really one of the fun damen tal b arometers of the state of education in the United S tates. Lyle Jones and I re- cen tly edited a h istory of NAEP , often called the “Nation’s Rep ort Card” (Jones and Olkin, 2004 ). It is a fascinating story . The story starts in the early 1960s with F rancis Kepp el, then U.S. Com- missioner of Education, who r ecognized the n eed for a n ational assessmen t of edu cation. Kepp el wa s a friend of Ralph Tyler, then director of the Cent er for Adv an ced Study at S tanford, and of John Gard - ner, then president of the Carn egie Corp oration of New Y ork. The three had talk ed ab out the idea of a national assessmen t. K epp el ask ed Tyle r to suggest a wa y to ev aluate education, and T yler conv ened a committee consisting of John T uk ey (Chair), Rob ert Ab elson, Lee Cr on bac h and Lyle Jones to develo p a plan for a p erio dic national assessmen t. Gardner, via the Carnegie Corp oration provided fu nding for t wo conferences. F rom this b eginning, thr ough a series of conferences, committees, and partial assessments, an assessmen t of 17- ye ar-olds in citizenship, science and writing to ok place in 1969. W e hav e come a long 16 A. R. SA MPSON w a y fr om that p oin t, and it is a credit to NCES that NAEP has main tained credib ilit y through ou t its history . T o da y w e ha ve in ad d ition to th e national NAEP , a state NAEP , b ecause states are in terested in ho w well students in the state are doing. What all of this sho ws is that it take s the conﬂuence of many forces to accomplish a pr ogram of this magnitude. As a result of m y ed u cation conta cts, I b ecame a mem b er of what was then called the T ec hnical Advisory Committee to NAEP , and is no w called the Design and Analysis Committee of NAEP . I’v e b een on that C ommittee for probably 15 or 20 ye ars. It’s had somewhere b et ween 10 and 14 mem b ers, mostly statisticia ns and psyc hometricians—all ve ry go o d p eople. W e meet three times a y ear, and our task is to help in some of the tec hn ical intricacie s in doing a national assessment. I’ll giv e yo u an example of one of the tec h nical problems. When Congress mandates that a change b e made, this aﬀects the wa y the testing w ill take p lace. F or example, Congress mandated th at testing give ac- commo dation to children w ho h av e d isabilities. Th is mean t that NAEP must decide how to deal w ith c h ildren who h av e hearing problems, or ey esigh t prob- lems, or d y s lexia, or a v ariet y of other problems. But the p oint is that now if y ou’re doing a trend and sud - denly at a certain p oin t Congress mak es a c hange, ho w do you maintai n the trend giv en this c h ange? W e’ve had many discussions on that. Another exam- ple of a tec hn ical pr ob lem which is of interest is the state NAEP . W e ha ve ﬁfty states, and no w if y ou’re going to mak e comparisons b et wee n states, we might ha v e more than 1,200 comparisons. So, the qu estion is h o w should we make these m ultiple comparisons? The real p roblem is how to reconcile a tec h n ical correctness with an interpretiv e correctness. If you do multiple comparisons yo u might ﬁnd that state A an d state B are not statistically diﬀeren t. Bu t if y ou don’t do m ultiple comparisons, yo u migh t ﬁnd out there’s a signiﬁcan t diﬀerence. The issu e is no w that a legislator in some state will sa y , “Ho w can there b e t w o answers—w e’re diﬀerent or we’re not diﬀeren t?” W ell, as y ou kno w, th e m ultiplicit y of ac- ceptable statistical analyses is a standard problem for statisticians in man y con texts. It’s not just ba- sically a commen tary ab out multiple comparisons v ersus none. Someb o dy d o es a t -test and someb o dy do es a nonparametric test, and you could get diﬀer- en t answers. W e ha v e a problem in interpretati on, and th e National C en ter of Educational Statistics has th e task of telling th e nation its r esults. S o, if y ou lo ok at the rep orts, y ou’ll ﬁ nd a v ariet y of sug- gestions that ha ve b een implemented to try to b e clear to th e public as to what’s going on. W e’v e u sed fo otnotes to try to explain, and w e hav e con tin uing discussions on this p oin t. Sampson: NISS is another statistic al ent erprise that’s f o cused on issues in pub lic p olicy . Olkin: I’d like to go back to the h istory of how NISS actually b ecame NIS S. The group in vo lv ed with starting Statistic al Sci enc e also r ecognized that there wa s a deﬁciency in cross-disciplinary researc h . Ultimately , there was a prop osal th at Jerry Sacks and I submitted through the I MS to th e National Science F oundation to ha ve a panel to discuss and study cross-disciplinary researc h. When th at panel’s rep ort w as completed, a little b o oklet ab out its ﬁnd- ings w as issu ed (Olkin and Sac ks, 1988 ). That b o ok- let has b een used b y many c hairs of d epartmen ts when talking to deans. It’s b een u sed in a v ariet y of other con texts. In an y case, there w as a long discussion in th e b o oklet ab out the need for cross- disciplinary researc h. A conﬂuence of many ev ents are needed for an organization lik e NISS to b e form ed. The P anel’s rep ort pr ovided a rationale for an In- stitute of Statistics, whic h, I th ink, it is fair to sa y , w as my idea. Jerry Sac ks help ed with the plannin g and conceptualiz ation and Nancy Flourno y , the NSF Statistics Program Director, m an aged to pr o vide money for a feasibilit y study , w hic h resulted in a call for p rop osals. All in all, it to ok several ye ars b efore NISS came to fr u ition. Sampson: There w as a heated comp etition for NISS. Pittsbur gh was one of the ﬁnalists, as y ou kno w, along with North Carolina. Our state legisla- ture wasn’t as generous as the State of North Car- olina. Th e North C arolina group did a wo nderf ul job in obtaining NISS. Olkin: That’s absolutely correct. T here w ere ﬁv e prop osals of whic h t w o were ﬁnalists. One wa s fr om a consortium of the Universit y of Pittsbur gh and Carnegie Mellon Univ ersit y . Dic k Cye rt, then Pres- iden t of Carnegie Mellon, wa s instrumental in of- fering u s sp ace in th e future. Carn egie Mellon wa s building up. And the other p rop osal w as from a con- sortium in North Carolina, namely , of Duk e, North Carolina at Chap el Hill, North Carolina State and the Researc h T riangle. A t that time, the Researc h T riangle oﬀered land in their park, b ut the most critical p oin t w as that the North Carolina consortium was instrumental in getting their state legislature to oﬀer quite a bit of A CONV ERSA TION WITH INGR AM OLKIN 17 money . There was a lot of start-up w h ic h wa s dif- ﬁcult to come by , and so NISS ended up in North Carolina. But in contrast to some of th e mathemat- ics institutes, NISS w as not designed to b e an insti- tute f o cused on a sp eciﬁc area, but rather on cross disciplinary areas that w ould ha ve a p olicy impact. In fact, in the original call for p rop osals, we limited en tries to east of Ch icago b ecause we though t that if NISS’s w ork was going to inv olv e p olicy that it should b e with in access of W ashington, D.C. NISS w as started and no w is in its tent h y ear. And I think it h as b een very successful. Sampson: Also no w it’s inv olve d with SAMSI whic h is headed by Jim Berger, and they are hous ed in the same b uilding. Olkin: I think it’s v ery go o d for statistics. I w ould only argue that it’s a sh ame that we d on’t ha ve three such institutes b ecause I think th e profession is gro wing to suc h a degree that there should b e. A v ery natural marriage would b e b et w een statistics, genetics, and biology to ha ve a m ore fo cus ed role in thinking ab out ho w statistics can p articipate in what is clearly going to b e an imp ortan t and gro wing ﬁeld. If statistics isn’t inv olve d in the early stages, it m ay not b e inv olv ed later on. Sampson: Let’s mo v e from NIS S to NIST. Brieﬂy , what were yo ur in teractions with NIST? Olkin: W ell, I w as inv olv ed with Churc hill Eisen- hart, wh o was the ﬁrst leader of th e Statistical Engi- neering Group at the National Bur eau of S tandards. No w y ou ha v e to remem b er that the National Bu- reau of Standard s h ad a long history of statistics. In the 195 0s, it had on its staﬀ p eople such as Marvin Zelen, F rank Proschan, Richard S a v age and Joan Rosen blatt. After Ch urchill Eisenhart r etired, Joan Rosen blatt w as the head of this thriving group. I wa s in vited to participate in some of th eir activities. I actually ga ve a lecture on meta-analysis th ere, and o v er th e y ears I’ve b een inv olved in n umber of p ro jects. The most recen t one I’v e b een in vo lv ed with is an up date of what I’ll call the Abramo witz and S tegun ( 1972 ) b o ok, M athematic al F unctions , w hic h is one of the famous and b est selling b o oks eve r in mathematics. In it there’s a chapter on probabilit y and statistics b y Marvin Zelen and Norman Seve ro. T his chap- ter is b eing up d ated and I’ve b een inv olv ed with a n umb er of p eople on that pro ject. ST ANFORD Sampson: In gram, we hav en’t y et talk ed muc h ab out y our unive rsity careers. I know y ou started at Mic hi- gan State and then wen t to Minnesota b efore Stan- ford. Also m y rough calculatio n is that Stanford ’s Statistics Departmen t, is ab out 60 ye ars old. Y ou’ve b een there for approxima tely 45 y ears, whic h is 75% of the departmen tal life history and this migh t giv e y ou an inte resting p ersp ectiv e. Olkin: Let me talk a little bit ab out eac h piece, if I ma y . The ﬁrst p oin t is in 1951, w h en I graduated, there w ere v ery few statistics departments. Almost ev ery one wh o graduated, generally would go to a math department. What I didn’t recognize at that time was that man y of u s who w en t to math d epart- men ts ultimately tried to generate statistics dep art- men ts. S o th is started this general growth. Wh en I wen t to Mic higan State, there were n ot many j obs and I did n ot ha ve many oﬀers. Math departments w ould w ant one statistician, and so if they had one, Fig. 9. I ngr am Olkin in his Mi chigan State Uni versity Of- ﬁc e, 1950s. 18 A. R. SA MPSON they weren’t ready for a second one. And a lot of math departments didn ’t hav e any , and didn ’t wan t an y . In an y case, Leo Katz—once again random c h ance mec hanism plays in y our life—wa s a visi- tor at Chap el Hill. W e b ecame fr iend ly and he said “W ell, wh en you graduate s en d me yo ur vita.” I d id, and they oﬀered m e a job. Leo Katz was the ﬁ rst mem b er of the statistics group at Michig an S tate. I cannot remem b er whether I wa s second or third, but the other one w as Kenneth Arnold wh o had b een in Wisconsin. After that was J im Hannan. L eo w as v ery goo d as a manager, and he arranged to get a lot of visitors, su c h as Herman Rubin and Esther Seiden (w h o late r wa s on the faculty). R. A. Fisher w as a visitor during my sta y th ere. I’m one of the few memb ers of th e profession n o w w ho can say th at R. A. Fisher h ad din ner at our hou s e. Sampson: That sounds lik e a story for another in- terview! Olkin: Y ou’re right. In an y case, once we had a core group, we w ere able to form the departmen t, and that was p robably by 1956 or 1957. And I really enjo y ed Mic h igan State very m uch. It was a great part of my life. Th e state of Mic higan had thr ee statistics groups. There was the Mic h igan State group. Don Darling, Paul Dwyer and C ecil Craig w ere at Michiga n. Milton Sob el an d Ben Epstein w ere at W ayne. W e h ad joint meetings and semin ars on o ccasion where we rotated the places. It wa s a v ery go o d time for statistic s. In 1960, I was oﬀered a v ery nice p osition at Minnesota, and I decided to accept that. As it turn ed out, the c hair at Min- nesota w as P almer John son wh o was an educational statisticia n—a wonderful p erson. He died shortly af- ter I arrived and I b ecame the c hair. Richard Sa v age w as there, as w ere Bernie L indgren and Leo Hurwitz, and we were starting to bu ild up . W e hired Milton Sob el and w e hired Mey er Dwa ss. (Mey er later d e- cided to go bac k to Northw estern.) So this started an increase in S tatistics. Later Kallianpur wa s h ired and a n umber of others. And, of cour se, Minnesota has n ow b ecome a large Sc ho ol of S tatistics. In an y case, I h ad this oﬀer f r om S tanford. I ha v e to say that at that time, I don’t kno w that I would ha v e accepted an ything other than Stanford . I had found S tanford to b e really almost an id eal p lace for someone like m yself b ecause I wa s inv olve d in m ultiv ariate, and the oﬀer w as joint b etw een Educa- tion and Statistics to build up the s tatistics program in Ed ucation. T h at was really app ealing to me. W e mo v ed to Stanford in 196 1. The d epartmen t at Stanford wa s found ed in 1948. Al Bo wk er wa s the ﬁrs t chair. There is an inte r- view in Statistic al Scienc e (Olkin, 1987 ) that d e- scrib es the early da ys and how h e b ecame c h air. He is a sup er b manager, an en trepreneu r for statistics. Shortly b efore I came, Herb Solomon b ecame chair, b ecause Al Bo wker b ecame Dean of the Gradu ate Sc ho ol, and Herb w as actually the one who hired me. Herb w as also a vigorous supp orter of s tatistics. The facult y consisted of Solomon, Herman Chernoﬀ, Charles S tein, Jerry Lieb erman, Manny P arzen, Ru- p ert Miller, Lin coln Moses, V ern on Johns and S am Karlin. I b eliev e I ha v en’t omitted any one. I was the next th ere and Kai Lai Chung came the same y ear I did. No w a critical p oint in Bo wker’s thin k- ing was that Statistics, b eing a small department, w ould never hav e to o large a facult y un less it had join t app oin tmen ts, wh ic h w ould mean t wo p eople in every billet. But furtherm ore, he ﬁrmly b eliev ed that statistics sh ould hav e all these tentac les and connections, and I think th at came out of his b eing at the statistica l researc h group at Columbia, wh ere once agai n Hotelling wa s one of the leading ligh ts. Before to o long, we had joint app oin tments with m yself in Education, T ed An derson in Economics, T om Co v er with Electrical En gineering, K arlin with Mathematics, and Moses with the medical s chool. The Stanford of the early 60s w as not only an ex- citing p lace, it was a p henomenally cohesiv e t yp e of place. At lunc h, there w as a game of hearts that wa s v ery cut-throat and had mathematicians and statis- ticians. There w as a bridge game, there was a go game, and some p eople wen t swimming. In 1961, Statistics d id not h a v e its o wn building, nor did Mathematics. Math was scattered o v er campus, so that some of the mathematicians we re in the same building as Statistics. In 1964, the mathematicians got their o w n buildin g, and Statistics had Sequoia Hall all to its o wn. But early on, it did not. Sampson: Ingram, it is my impression that S tan- ford has c hanged o ve r time in terms of the so cial relationships in the departmen t? Olkin: W ell, I think that’s deﬁnitely true. Y ou ha v e to remem b er that in the 60s, most of the f acult y liv ed on campu s , and so we were not only close ge- ographically to the S tatistics Departmen t, we were also close to one another. The d emographics ha v e c h anged in the Departmen t and the h ousing has c h anged. I don’t kno w where every one live s no w, but s ome faculty migh t live in Redwoo d Cit y or San F r ancisco, and some m igh t liv e closer to San Jose. A CONV ERSA TION WITH INGR AM OLKIN 19 Fig. 10. Sam Gr e enhouse, R. A. Fisher, Unknown, Car ol Parzen, I ngr am Ol ki n, Manny Parzen (l-r), Paris, 1961. ISI Me et- ing. Also there are n o w a lot more pressures, a lot m ore t w o career f amilies than there w ere in the 1960s, and as a result of this, p eople are sp ending more time a wa y from the oﬃce, w h ereas in the early days the oﬃce w as s ort of “Grand Central S tation” in many w a ys. All this mak es for diﬀeren t interactio ns. Pe o- ple come in an d do their work, so to sp eak, and then ma yb e go h ome. Th e computer has certainly facil- itated all of this. There are more closed d o ors and there’s certainly less int eraction. My conv ersation with p eople in other unive rsities is that the same is true ev erywhere. P eople can d o an en tire job aw a y from the main oﬃce. It would b e in teresting to see a study across univ ersities’ d epartmen ts in whic h we coun ted the num b er of join t pub lications among fac- ult y wh o were in the same departmen t, and see ho w that h as c hanged o v er time. Sampson: Stanford has alwa ys b een a wo nder f ul b eacon of academic statistics. Do y ou think it’s main- tained that role and k ept its illustr iousness? Olkin: I’v e th ou ght a lot ab out ho w do great de- partmen ts mainta in thems elves, and wh y do some great departments go do wn and then other d epart- men ts, who are not that wel l-kno wn, sudden ly b e- come well -known. An imp ortant ingredien t is the t yp e of y oung p eople y ou b ring in. O f course you ha v e to b ring in very go o d p eople if the department is to m aintain its stature. But you also n eed p eople who in teract with the profession, who are not iso- lates, b ecause y ou wa nt the departmen t to in some w a y b ecome a “domain of attraction.” I thin k that Stanford has b een very , very f ortunate in b eing able to attract a n umber of y oung p eople who are clearly 20 A. R. SA MPSON Fig. 11. Jack Kiefer, I ngr am O lkin, Milton Sob el, Unknown, 1981. not only excellen t, but will k eep u p the stature of th e Departmen t. As I lo ok at other places, I think that w e hav e b een among the fortunate in the qualit y of our y oung p eople. There are sev eral other depart- men ts that I th in k hav e also b een successful in th is. F or instance, I think Carnegie–Mellon has done an excellen t job in b eing able to rep lace itself. Whether a d epartmen t will b e able to m ain tain its stature is a contin uing problem. Sampson: In gram, can you talk ab out wh at yo u’ve seen in terms of the change in the School of Educa- tion at Stanford d uring your time there? Olkin: No w, 1961 was a time wh en th e School of Education was in the th ro es of tr y in g to b ecome more researc h oriented, whereas previously they h ad b een oriented tow ards the practice of education. I wa s hired at the same time as Dic k A tkins on wh o was half in Psycholog y and h alf in Edu cation. Shortly af- Fig. 12. Ingr am Ol ki n in his Se quoia Hal l oﬃ c e, Stanfor d University, Spring, 1996. A CONV ERSA TION WITH INGR AM OLKIN 21 ter I came, Janet E lash oﬀ and Lee Cron bac h joined the faculty and we formed what yo u migh t call an educational s tatistics group similar to what a bio- statistics group wo uld b e. W e had semin ars and w e had a Ph.D. p rogram. O ne of the nicest times in my asso ciation with the Sc ho ol of Education w as w h en Rosedith Sitgrea ves w as a colleag ue. Sitgrea ves, Cron bac h and I f ormed a very nice tr io, and quite a n umb er of students graduated in educational statis- tics. What I wan t to note is th at aside from the p eople and wh at they were doing in the program, one of the adv an tages of b eing asso ciated with the Sc ho ol of Edu cation is that students had p roblems that w ere somewh at diﬀeren t, and it wa s an app lied area that p osed many new problems, m ostly in mul- tiv ariate analysis. It was also the genesis of meta- analysis, so I v alue that asso ciation tremendously . Sampson: Ingram, d o y ou wan t to men tion your closer fr iends on the facult y at Stanford, f or instance, Jerry Lieb erman and Herb Solomon? Olkin: Wh en I ﬁ rst we nt there it was clear that Jerry and I w ould b ecome close friends. And we main tained that f r iendship unt il h is death a num b er of y ears ago. I was ve ry close with him. I was also v ery fr iendly with Herb S olomon from wa y bac k, and h e wa s also a close friend. These w ere friend- ships that were main tained o v er a v ery long p erio d of time. Sampson: I know that when Jerry was losing his strength and his abilit y to comm un icate b ecause of ha ving ALS , yo u we re one of his faithful visitors. Y ou were there sev eral d a ys a w eek to sp end time with him. That must h a ve b een b oth rewa rdin g and diﬃcult. Olkin: Jerry had the abilit y , ev en though h e could barely sp eak, to mak e p eople feel w elcome and wan t to visit. It was n ev er a d epressing time u nt il really close to the end . But it wa s a hard time b ecause it w as not clear h o w muc h h e u ndersto o d after a wh ile. Early on, he wa s quite lu cid and w as able to com- m unicate. But ALS is such a debilitating disease, so that after a certain p oin t one has to concen tr ate jus t on su rviving. It’s v ery diﬃcult to think ab out any other asp ects. Sampson: I remem b er visiting y ou during th at time and going with yo u to visit J er r y . He had lost totally his abilit y for sp eec h, bu t he had a syn thetic sp eec h b oard, and he “talk ed ” with a sense of humor. I r e- mem b er v ery vividly h e had programmed “tak e the rest of the da y oﬀ ” and h e would hit that as p eople w ere lea ving. Olkin: Y es, th at’s absolutely true. He had a n um- b er of his fa v orite commen ts that were programmed in to the synt hesizer, and he w as alwa ys u pb eat. He alw a ys had a sm iling hello, an d then we w ould talk. Basically to w ards the end, of course, it was a mono- logue. I wo uld try to b ring h im u p to date on wh at Fig. 13. I ngr am Olkin, Jerry Lieb erman and Ge or ge R esnikoﬀ, 1981. 22 A. R. SA MPSON w as going on, but ALS is a ve ry d iﬃcult disease b oth for the p er s on and for p eople around. ST A TISTIC AL S T A TESMAN Sampson: The term “statistical statesman” has b een applied to y ou, and I know the p eople wh o kno w you w ell would agree with this charact eriza- tion. I’m w ondering if yo u could say wh at y ou see y our role has b een and con tinues to b e as “statistical statesman?” Olkin: There’s no question I’v e b een an adv o cate of statistics. I think at times I’v e b een s omewh at forceful and my v oice may not hav e b een th at wel - comed in terms of b eing an adv o cate. But I w as ask ed to b e on a num b er of site visits, and that w as v ery go o d in that it ga v e me an opp ortunity to v oice to administrations the n eed f or a s tatistics p rogram or s tatistics dep artment. I w ould guess th at I was on ten or ﬁfteen suc h site visits durin g m y lifetime. A t a p lace lik e th e Un iv ersit y of Mic higan, w e w ere able to m ake some s uggestions whic h I think , su bse- quen tly , led to their making oﬀers and buildin g the departmen t. A t Mt. S inai Hospital, I think it’s al- most true to say that I help ed sav e one of the statis- tical programs. This also has o ccurred abr oad and, in fact, I just came bac k fr om Croatia where they’re talking ab out starting a do ctoral program in statis- tics. Th is is w ithin a mathematics department, and they do n ot ha v e an y sense of the role of statistics in applications, an d it wa s gratifying to b e able to illuminate them on a lot of these issues. So I think that I’ve had an eﬀect in that resp ect. Sampson: There are some m aterials I’v e seen from whic h I ha v e take n this direct quote. Someone de- scrib ed you as a “tireless campaigner for impro ving and increasing opp ortun ities in statistics for w omen.” Olkin: That’s a nice statemen t, and I really appr e- ciate it. I hop e I deserv e it. Let m e tell y ou a little bit of ho w it all s tarted. I think it’s apparent to most p eople that there are very f ew ten ur ed w omen in de- partmen ts of statistics. There are no w more ten ured w omen in biostatistics. But th roughout h istory th ere were not many ten ured w omen in academia. And what I n oticed w as that in many departmen ts, esp eciall y in small de- partmen ts, or in departments wh ere a woman would b e th e lone statistician in the math departmen t, when it came time for tenure, it w ould b e very dif- ﬁcult to mak e an assessmen t. The mathematicia ns w ould ev aluate app lied work v ery diﬀerently from theoretical wo rk. It wo uld b e ve ry hard f or this lone w oman to b e able to get letters of recommendation. I had the idea that if we could invite some wo men in their fourth and ﬁfth yea r to wards tenure to Stan- ford for tw o su mmers, that it would giv e them a c h ance to broaden their scop e, write-up their r esults and, hop efully , get in vo lv ed with some of the facult y . The optimal outcome of these visits wo uld b e that one of the Stanford facult y wo uld know this p erson suﬃcien tly w ell to b e able to write a letter of rec- ommendation. And certainly this op ened do ors for some of these women. Sampson: When did y ou start this? Olkin: It’s no w at least 15 y ears. Sampson: Y ou started it on your own NSF grant? Olkin: The wa y it started is that I p rop osed this to the National Science F oundation. They said, “w ell, w e’ll try this” and th ey ga ve m e a sup plemen t to m y m ultiv ariate gran t. I was able then to go to the Asso ciate Prov ost and mak e th e case that this wa s a unique pr ogram in the United States, that it would help women, and that Stanford could b e part of this b y w aiving o verhead w hic h w ould p ermit us to in- vite more women with the s ame amount of money . Stanford did wa iv e the ov erhead. Sampson: How man y women ha v e come in the y ears that yo u’ve b een doing this? Olkin: W ell, we used to ha ve somewhere b et wee n 1 and 4 w omen eac h s ummer. Ov er a 15 y ear p e- rio d, there w ere probably 15 to 20 w omen w h o came through the p rogram, and I will sa y most of them did get tenured. I thin k it just w as an inspiration to them. It also lo ok ed go o d on their vita to b e able to sa y th at they w ere invit ed under an NSF program to b e at Stanford. Sampson: W er e yo u in v olv ed in other wa ys w ith men toring w omen? Olkin: I had a n umber of female Ph.D. studen ts and, in general, I was able to advise them on what w ould b e in store for them after graduation, and to try to help them in deciding on j ob oﬀers. F rom man y women, n ot only from Stanford , I got phone calls trying to discuss d iﬀeren t oﬀers, what was p osi- tiv e and negativ e, and I wo uld try to giv e th em some unbiase d advice. Not from my p oin t of view, but more from th eir p oint of view and w hat their n eeds w ere. It turn ed out that this is a con tagious pro cess. If yo u h elp one p ers on, the w ord gets around, and then you help another one, and another one, and b efore y ou kn o w it, you’re sort of a cen tral agency for giving advice. A CONV ERSA TION WITH INGR AM OLKIN 23 Sampson: It is to y our tribu te that you were the ﬁrst male to win the Elizabeth Scott Awa rd. That m ust ha ve b een gratifying to you. Olkin: It w as really gratifying. Of course, y ou h a ve to rememb er I’m the father of three daught ers. I b e- came a ware of gender bias in s cience v ery early on when w e w ere in England. Our oldest daugh ter was in school, and she wan ted to tak e some science courses and they w ant ed to counsel h er out of it. W ell, I foun d m yself b eing an adv o cate in that dir ection and in- sisting on this not happ ening. An d this happ ened o ver and o v er again in diﬀerent w a ys. So I wa s re- ally aw are of the problem. I’ve also talk ed to a lot of w omen who ha v e Ph .D.s and I’v e disco ve red that of- ten there wa s a paren t in v olv ed who stated that their daugh ter w as not to b e discrim in ated against an d fough t for them. W e no w know that man y w omen are counseled out of the sciences, and it do es tak e an advocate. Sampson: I know that y ou ha ve h ad a n umb er of other honors. Y ou receiv ed an honorary degree from DeMon tfort Univ ersit y in England; also a CCNY Distinguished Alum n us Awa rd, an AS A F ounders Aw ard, a Wilks Medal, a L if etime Contribution Aw ard from the American Psyc hological Ass o cia- tion, and a Fisher L ecturesh ip, among others. I’m w ondering, are there an y fav orites among these for y ou? Olkin: Certainly the Elizab eth Scott Aw ard, but the other one that I really v alued a lot was the Wilks Medal. Wilks was one of my hero es, in part, b ecause he w as really a statistical statesman. He fought in man y wa ys for the fur therance of statistics, and eve n though I was not at Princeton, I kn ew ab out this and on o ccasion discussed this with him. Wilks w as a theoretician, h e had stu den ts at Princeton, he w as an editor, and he was inv olv ed in helping the pro- fession. In many wa ys I though t of myself tryin g to follo w in the v arious directions that he had. So that a ward w as really a very pleasing one. Sampson: W ould y ou call him a role mo del? Olkin: In man y w a ys he w as a role mo d el. Hotelling w as a d iﬀeren t kin d of p erson and fough t diﬀerent kinds of battles, but the battles that Wilks fough t w ere v ery similar to the battles that I as p erson could see myself doing. And so, yes, Wilks w as def- initely a mo del for me. F AMIL Y LIFE Sampson: W e’v e b een talking a lot ab out y our pro- fessional life and the fullness of y our professional life, but I kno w yo u’ve got a full family life, to o. Y ou’ve b een equally in vo lv ed with Anita and your three daugh ters. I’m curious ho w y ou met An ita. I don ’t think I ev er heard that story . Olkin: What seems to come out in our con versa- tion is that c hance pla y ed a big role in many as- p ects of my life, and this wa s another suc h o cca- sion. I wa s in th e orc hestra at Cit y College. I pla y ed trom b one. Not very w ell, I migh t say , b ut I pla y ed trom b one. I p la y ed well enough to b e in th e orc hes- tra, and the orchestra used to pla y at the b ask etball games in Madison Square Garden wh en the Cit y College basketbal l team w as one of the great teams. (They we re one of the few teams who w on b oth the NIT and NCAA championships.) My close f riend, Andy Gregg, wh o also p la yed tromb one, had a friend b y th e n ame of Anita, and And y in tro d uced me to Anita. I met her shortly after I joined the service. I was stationed in the east and I was able to meet with her f airly often. Because I w as a meteorologist, I w as stationed at airp orts. A t that time ev ery plane th at to ok oﬀ from an airp ort had to ha ve the signature of an op erations oﬃcer and a meteorolo gist. So I knew ab out ev ery plane that was taking oﬀ. When I had free time and someb o dy was ﬂying to New Y ork, I would ask the pilot if he could take m e, and the answ er w as often get a parac hute and come along. I would all of a sudden come into New Y ork at nine o’clock at n ight and call Anita an d we ’d go out. W e sa w eac h other for sev eral y ears this wa y and also corresp onded . Then I was stationed at LaGuard ia Airp ort. I w as in New Y ork, actually living at home, and w as able to see Anita quite a bit. Next San F ran- cisco con tacted LaGuardia and told them that they needed an extra p erson b ecause th ey w ere short- handed, and would LaGuardia transf er one of its p eople. But they also had the stipulation that b e- cause h ousing was very diﬃcult in San F rancisco, they would prefer if they transferred a single p er- son, and I was chosen to b e that p erson. I n an y case, I had to go to San F rancisco. W ell, Anita and I d iscussed this and w e decided to get married. In spite of S an F ran cisco’s demand f or a sin gle p erson, w e did get marr ied after ve ry sh ort n otice. W e b oth to ok a train to S an F rancisco, and that wa s a v ery happy part of our life. Sampson: Y ou w ere married r igh t at wa r’s end? Olkin: ` W e w ere married in Ma y of 1945 and the war wa s coming to an end. W e were able to ﬁnd h ousing in San F ran cisco. A t the time a lot of 24 A. R. SA MPSON Fig. 14. Anita and Ingr am Olki n, New Y ork, 1945. p eople r en ted ro oms. Anita w as w orking an d I w as w orking and we w ould meet afterw ards an d go to the Oﬃcer’s Clu b for din ner, and then just go out. It w as a v ery n ice time. I was disc harged a ye ar later in 1946, and at that time w e returned to New Y ork where I ﬁnished Cit y College. Sampson: Vivian, y our oldest, was b orn in North Carolina? Olkin: After Columbia, w e wen t to Ch ap el Hill, where Vivian w as b orn in 1950. Vivian return ed to Chap el Hill ab out 15 y ears ago where she no w liv es with h er family . Then in 1951 , we w ent to Mic higan State and Rho da, my middle d au ghter was b orn in 1953 in Lansing. And my youngest Ju lia is a S tanf ord c h ild, b orn in 1959 when I was on sabbatical th ere. Sampson: Y our daughters are very close to you and Anita. I think yo u sp ent a lot of time with them when they we re gro w ing up. Olkin: Oh y es. I was really very fortunate in ha v- ing suﬃcient energy to b e able to wo rk d uring the da y and come home in the late afterno ons and ev enings to sp end time with the c h ildren. F or most y oung children ﬁve o’clock is the “witc hin g” hour and th ey’re tired after the day . A t Mic h igan State, I r emem b er I w ould come h ome a b it early and sev- eral of us, my colleagues and I, w e w ould tak e our c h ildren and go oﬀ to the Cow Barn. Mic higan State had an agricultural school, so th at it h ad a lot of ar- eas whic h were attractiv e to c hildr en. It had a duck p ond and co w barn s, for example. The children lik ed the change of scenery . But I had a lot of energy an d that’s very fortunate when y ou hav e children and a work da y . After the children wen t to sleep, I was able to ev en con tinue w ork . Sampson: Y ou u se the past-tense “had a lot of energy .” I’ve b een at conferences with y ou recen tly and at the end of the conference we are all tired. But y ou are f u ll of energy and tell us “I’m going to ha v e dinner with a grandc hild tonigh t,” and you set out for an h our on a bu s line somewhere. Olkin: I think those are singular ev en ts now-a- da ys. Th e abilit y to wo rk late at n igh t is no longer with me. Sampson: Y ou are a vicious tennis p la y er, and I witnessed that p ers on ally . I once h eard Harry Jo e, who w as just half your age when he play ed you in tennis, sa y “the only w a y I kno w how to b eat Ingram”—he felt a bit guilt y ab out this—“was to hit the ball at wide ends of the court and mak e y ou just run so hard, th at even tually yo u w ould wear do wn.” Harry said he could n ot ﬁ gure out an y other w a y to b eat you. Olkin: W ell, I lov e tennis and I pla y ed quite a b it. I w as v ery fortunate alw a y s in b eing able to ﬁnd a v ery go o d doubles partner who could run and h elp out. Sampson: Harry w as describin g this in a singles game, Ingram. Olkin: On e of m y tric ks wa s to we ar the wrong so c ks and what one w ould call “schlump y” clothing, and try to psyc hologically h a v e my opp onent n ot ex- p ect v ery m uch. Th at w ork ed up to a certain p oint, unt il th ey disco ve red w hat was going on. [Laugh ter] A CONV ERSA TION WITH INGR AM OLKIN 25 Fig. 15. Ingr am and Anita Ol kin and their gr andchildr en, L e ah, No ah, Jer e d, Sophia, Jer emy and R achel (l-r), 2005. But ye s, I did lik e tennis a lot. I ha ve n’t pla y ed recen tly , b ut I’m going to get bac k to it. Sampson: Y ou’re still swimming. Olkin: Until the recent energy crisis, our swim- ming p o ol at the house was heated suﬃ cien tly all y ear r ound. I really enjo y ed getting up in the morn- ing and going swimm in g for t w ent y minutes or half an hour ev ery d a y . No w I just do it half a year and I hav e to ﬁgure out some wa y to sw im more often. Sampson: Also, m y recollect ion is that y ou enjo y hiking. Olkin: I ha ve gone on hiking trips—I really lo ve Y osemite. It’s b een just b eautiful up there, and I’v e gone bac kpac king, and often with one or tw o p eople whom I had k n o wn from the hiking groups. I w ould go, not alwa ys with the same grou p s, and sometimes part of the trip with one, and part with another. There’s some b eautiful h ikes in Y osemite. I w ould go up to the high country and then hike for ab ou t a week, and that was ju s t absolutely magniﬁcent . Y our could hik e, and then all of a sudd en there’s an op ening and the vista is just br eath taking. I re- ally enjo y ed that, and I did th at r egularly for man y y ears. PERSPECTIVES Sampson: W e are coming to the end of our con- v ersation. And there a couple of things y et I’d like to yo u touch up on. The ﬁrst is to ask if there is any advice y ou’d like to oﬀer y oung statisticians wh o are just s tarting their careers. Olkin: I think th at there are a num b er of profes- sions wh ere the notion of pr o b ono service is intrin- sic in the profession. The medical profession is one, as is the legal profession. And I think the statistical profession should b e, in many w a ys, one suc h. There are a lot of go vernmen t agencies and panels that d is- cuss imp ortan t p roblems relating to so ciet y . Most of these inv olv e the analysis of d ata, and if th e statisti- cians do n ot participate in this, I thin k , it would b e a serious mistak e. S tatisticians bring to the und er- standing of these problems a diﬀerent orient ation, in my opinion, that is not generally th e pur view of most of the p eople who are on these p anels. 26 A. R. SA MPSON The term pr o b ono is in teresting b ecause it r eally refers to lac k of ﬁn an cial comp en s ation. Ho w ev er, it’s not pr o b ono int ellectually . That is, there’s a v ery big return b y b eing on a lot of these panels. Sampson: Finally I’d lik e to ask y ou if y ou could, and I realize that this is a hard th in g to do, lo ok bac k at y our long and fru itful career in statistics, and fr om th is v ant age p oin t say wh at has giv en y ou the most satisfaction? Olkin: There are v arious things th at I’v e done that, I th ink, in retrosp ect ha v e b een v ery fulﬁ lling and satisfying. Of course, all of us hav e a research career, and that’s p art of it. The multiple collab o- rations that I ha v e h ad with y ou, with Al Marshall, Leon Gleser, Milton Sob el, Larr y Hedges, Michael P erlman, Jim Press, and others hav e b een b oth en- jo y able and satisfying. Bu t I thin k the friend s and student s whom I ha ve inﬂu en ced in a p ositiv e w a y ha v e b een ve ry satisfying. I’v e really enjoy ed the pro cess of helping a stud en t from b eginning to fruition. It is v ery , ve ry satisfying when students complete a d issertation and go on to live a fruit- ful life. If they b ecome totally and indep endent ly pro du ctiv e, I ﬁn d th at the p ersonal satisfaction is immeasurable. Also I think m y accomplishmen ts in terms of the profession h av e also b een very gratifying and, of course, m y role in helping build Stanford in to a great statistic s dep artmen t complemen ts all of these. I think this comp osite w as v ery fulﬁlling and really mak es me feel satisﬁed when I lo ok bac k at m y ca- reer. Sampson: T hank you Ingram. That’s an insp ira- tion f or all of u s, and thank yo u for this interview. A CKNO WLEDGMENTS W e w ish to thank Diane Hall who did a sup erb job, as alw a ys, of transcribin g the original tap es. Also we wan t to thank Leon Gleser for his advice prior to the con versatio n and his assistance in edit- ing the print v ersion of this con v ersation. REFERENCES Abramow itz, M . and Stegun, I. , eds. (1972). Handb o ok of Mathematic al F unctions with F ormulas , Gr aphs , and Math- ematic al T ables. National Bureau of S tandards A pplied Mathematics Series 55 . US Go vernment Printing Oﬃce, W ashington, DC. MR0167642 Acz ´ el, J. (1966). L e ctur es on F unctional Equations and Their Appli c ations. Academic Press, New Y ork. MR0208210 Cram ´ er, H. (1946). Mathematic al Metho ds of Statistics. Princeton U niv. Press, Princeton. MR0016588 Das Gupt a, S., Ea ton, M. L., Olkin, I., Perlman, M . , Sa v age, L. J. and Sobel, M. (1972). In equalities on the probability conten t of conv ex regions for elliptically conto ured distribution. Pr o c. Sixth Berkeley Symp. Math. Statist. Pr ob ab. 2 241–265. Univ. California Press, Berke- ley . MR0413364 Deemer, W . and Olkin, I. (1951). The Jacobians of cer- tain matrix transformations useful in multiv ariate an alysis. Biometrika 38 345–367. MR0047300 Feller, W. (1950). Intr o duction to Pr ob ability The ory and Its Appli c ations 1 . Wiley , New Y ork. MR0038583 Glass, G. V., McGra w, B. and S mith, M. L. (1981). M eta- analysis in So cial R ese ar ch . S age, Pub lications, Beverly Hills, CA. Gleser, L. J., Perlman, M. D., Press, S. J. and Sampson, A. R. (1989). Contributions to Pr ob ability and Statistics : Essays in Honor of Ingr am Olkin . Springer, New Y ork. [Bi- ograph y reprinted in Li ne ar Algebr a and Its Appl i c ations 1–15 (1994). MR1274406 ]. MR1024318 Hedges, L. V. and Olkin, I. (1985). Statistic al Metho ds f or Meta-A nalysis . Academic Press, New Y ork. MR0798597 Hoeffding, W. (1940). Masstabinv ariante korrela tionstheo- rie. Schriften des Mathematischen Instituts und des Insti- tuts fur Angewandte Mathematik der Universitat Berlin 5 179–233 . Jones, L. V. and Olkin, I. (2004). The Nations R ep ort Car d : Evolution and Persp e ctives. Phi Delta Kappa, Blooming- ton, IN. Kendall, M. G. ( 1944). The A dvanc e d The ory of Statistics . 1 . J. B. Lippincott Co., Philadelphia. MR0010934 Kendall, M. G. ( 1946). The A dvanc e d The ory of Statistics . 2 . Charles Griﬃn and Company , Ltd., London. MR0019869 Marshall, A. and Olkin, I. (1967). A multiv ariate exp o- nential distribution. J. Amer. Statist. Asso c. 62 30–44. MR0215400 Marshall, A . and Olkin, I. (1979). Ine qualities: The ory of Majorization and Its Applic ations. Academic Press, New Y ork. MR0552278 Olkin, I. (1987). A conv ersation with Alb ert H. Bowk er. Statist. Sci. 2 472–483 . MR0933739 Olkin, I. (1997). The History and Development of M eta- Ana lysis (DS041) and A Conversation with Ingr am Olkin (DS042). Di stinguishe d L e ctur e Vide otap es . American Sta- tistical A ssociation, W ashington, DC. Olkin, I., Ghur ye, S. G., Hoeffd ing, W., Madow , W. G. and Mann, H. B. , eds. (1960). Contributions to Pr ob a- bility and Statistics : Essays in Honor of Har old Hotel l ing. Stanford Un iv. Press, Stanford, CA. MR0120692 Olkin, I. and S ack s, J. (1988). Cr oss-Disciplinary R ese ar ch in the Statistic al Scienc es . (R ep ort of a Panel of th e Insti- tute of Mathematical St atistics.) IMS , Hayw ard, CA. Pra tt, J. W. and Olkin, I. (1958). On a multiv ariate Tc hebyc heﬀ inequality . Ann. Math. Statist. 29 226–234. MR0093865

A Conversation with Ingram Olkin

Original Paper

Comments & Academic Discussion

Leave a Comment

Original Paper

Related Papers

Comments & Academic Discussion

Leave a Comment