Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
GERTRUDE MARY COX January 13, l900~ctober 17, 1978 BY RICHARD L. ANDERSON THIS IS A FINAL TRIBUTE to a fellow statistician, fel- low graduate student, employer, and above allbest friend and well-wisher, the confidante and constant compan- ion of my wife and children. Gertrude Mary Cox had that rare combination of administrative strength and love for her fellow man we so desperately need at the present time. A gracious, patient, tenacious visionary, she brought out the best in people. As a pioneer in the development of statistics she was a servant to science who never lost her touch with people. ~ EARLY YEARS Gertrude Cox was born on a farm near Dayton, Iowa, where she spent several years "roaming in the woods by the river," as she put it, "and wandering over the hills." The fam- ily then moved to the small town of Perry, Iowa, where Gertrude attended public school. A lover of competitive sports, she played on the high school basketball team. (Iowa was the center of girls' basketball in those clays.) , . _ ~ ~ Much of what is printed here is excerpted from a 1979 obituary I prepared with Robert Monroe and Larry Nelson of North Carolina State University, "Gertrude Cox A Modern Pioneer in Statistics," Biometrics 35(1979):3-7. I have also included remarks from a letter Gertrude Cox wrote to me on October 10, 1975. 117
118 BIOGRAPHICAL MEMOIRS The Coxes were a close-knit, midwestern family with four chilclrentwo boys and two girls. Gertrude was especially close to her mother, Emma, and later wrote of her: "I learner! from my mother the value and joy of doing for other people. She nursed the sick for miles around and raised us to be active church workers." During those early years Gertrude also learned to like making bread perhaps because she was allowed to sell one pan of biscuits. Her excellent cinnamon rolls were famous. She always served them to us whenever we visited, and, when we left, provided one package for our son, Bill, and another for the rest of us. Gertrude lover! children and always joined us on Christmas morning to see our two youngsters open their gifts. Gertrude's early ambition was to help others. She took a two-year course in social science, then spent another two years as a housemother for sixteen small orphan boys in Montana. As preparation for becoming the superintendent of the orphanage, she decided to enroll at Iowa State College. Majoring in mathematics because it was easy for her, she elected courses in psychology, sociology, and craftscourses useful to her in her chosen career. In 1929 she received her B.S degree. To help pay her college expenses Gertrude did comput- ing, George Snedecorher calculus professor having asked her to work with the comptometers in his computing laboratory. Speculating (forty-six years later) as to why he had chosen her for this work, she told the Raleigh News and Ob- server in May 1975, that he hac! probably hoped that she, the only woman in the class, would have more patience for detail work than the men. Perhaps because of this computing experience, Gertrude became interested in statistics. But the Mathematics Depart- ment at that time would not aware! an assistantship to a
GERTRUDE MARY COX 119 woman, and she financed her graduate work with assistant- ships in psychology and art. In 1931, she received the first Master's degree ever given by Iowa State in statistics but was turned down for a job teaching high schc~o} mathematics be- cause she clid not have the requirec! courses in education. She decided to continue her graduate career. Because of her love of people and her desire to learn what "made them tick," Gertrude chose psychology as her research area. With a graduate assistantship at the University of Cali- fornia, Berkeley, she began work on a doctorate in psycho- logical statistics. Unfortunately for the fielc! of psychology, she stayed only two years. In 1933, Iowa State established its Statistical Laboratory under the direction of George Snedecor, Gertrucle's former mentor, ant! he persuaded her to return home to help him. Back in Iowa, she continued her interest in psychology and worked with several members of the Psychology Departmentincluding its chairman (later, dean of the School of Industrial Science), Harold Gaskill- on the evaluation of aptitude tests, test scoring procedures, and the analysis of psychological data. At the same time she was put in charge of establishing a Computing Laboratory and consulted! in ant! taught experi- mental designs. In 1934 she began to teach "Design of Ex- periments" a course that would become renownec! to fol- low Sneclecor's "Statistical Methods." Most graduate students in agriculture were required to take this sequence, a require- ment that was later extended to a number of other disciplines at Iowa State and was my own introduction to experimental statistics. Both the Snedecor and Cox courses were originally taught from mimeographed materials. In 1937, Snedecor's material came out in book form, but Gertrude only published her design material in 1950, when it came out as a colIabo- rative effort with W. G. Cochran (1950,7~. Gertrucle's course was built around a multitude of specific
120 BIOGRAPHICAL MEMOIRS examples (many of which ~ still keep in my files) in a variety of areas of experimentation. Members of her computing staff analyzed all of the data, which were then completely checked by Gertrude and the hundreds of graduate students in her course. Despite the fact that these experiments were con- ducted four or five decades ago, they could still furnish the basis for a sold course on the design of experiments, espe- cially in biology ant! agriculture. In her later "Advanced Experimental Design," Gertrude concentrated on her three basic principles for setting up an experiment: (1) Experiment objectives should be set forth clearly at the outset, the experimenter having answered the following questions regarding his or her experiment: Is it a preliminary experiment to determine the course of future research or is it intended to furnish answers to immediate ques- tions? Are the results to be put to immediate practical use or are they intended to help clarify theoretical questions? Does the researcher wish to obtain estimates or to test for significance? Over what range of experi- mental conditions do the results extend? (2) The experimenter should describe the experiment in detail, clearly defining proposed treatments, size, and materials: Is a control treatment necessary for comparison with past results? Will the funds available sup- port an experiment of sufficient size to yield useful results? Are the ma- terials necessary for the experiment available? (3) The experimenter should draw up an outline analyzing the data before starting the experiment. Both as a teacher and a consultant, Gertrude particularly emphasized randomization, replication, and experimental controls as procedures essential to experimental design: "Randomization is somewhat analogous to insurance in that it is a precau- tion against disturbances that may or may not occur and that may or may not be serious if they do occur. It is generally advisable to take the trouble to randomize even when it is not expected that there will be any serious bias from failure to randomize. The experimenter is thus protected against unusual events that upset his expectations. Of course in experiments
GERTRUDE MARY COX 121 where a great number of physical operations are involved, the application of randomization to every operation becomes time-consuming, and the experimenter may use his judgment in omitting randomization where there is real knowledge that the results will not be vitiated. It should be realized, however, that failure to randomize at any stage may introduce bias unless either the variation introduced in that stage is negligible or the experiment effectively randomizes itself." (1950,7, p.8) ~ . . . ~ . .. .. . ~ . units AS she pointed out, replication not only Increases the ac- curacy of treatment comparison, it also enables the experi- menter to obtain a valid estimate of the magnitude of exper- imental error. She also offered the following ways to increase accuracy by improving the control of experimental tech- niques: (1) Select the best experimental design for the proposed experiment; (2) Ascertain the optimal size and shape of the experimental unit; (3) Use uniform methods for applying treatments to experimental (4) In order that every treatment operate under conditions as nearly the same as possible, exercise control over external influences; (5) Devise unbiased methods for increasing treatment effects; (6) Take additional measurements (covariates) often to help explain final results; (7) Provide checks to avoid gross errors in recording and analyzing data. Though Gertrude was enrolled in a Ph.D. program in mathematics at Iowa State, her teaching and consulting du- ties cliff not leave her enough time to write a dissertation. An "assistant" from 1933, she was appointed research assistant professor in 1939, though her design course was listed under Professor Snedecor's name. In 1940 Snedecor was asked to recommend candidates to head the new Department of Experimental Statistics in the School of Agriculture at North Carolina State College. "Why didn't you put my name on the list?" Gertrude asked when
122 BIOGRAPHICAL MEMOIRS he showed her his all-male list of candidates, and her name was aclded to the accompanying letter in the following post- script: "If you would consider a woman for this position, ~ wouIc! recommend Gertrude Cox of my staff." This terse note was to have far-reaching consequences for statistics, for not only was Gertrude considered, she was selected. Her resig- nation led to a heart-rending session with Dean Gaskill, she later told me, in which he tried to convince her that she was being disloyal to her native state anc! to Iowa State College, and that a woman would never be accepted as a department head in a southern state. SOUTHERN VENTURE Gertrude Cox became the head of North Carolina State's Department of Experimental Statistics on November I, 1940. The Board of Trustees of the Consolidated University of North Carolina authorizes! the establishment of the Depart- ment and confirmed! Professor Cox as its head on January 22, 1941. She hac! strong support from the U.S. Bureau of Agricultural Economics, which had been instrumental in es- tablishing the Department, and, in particular, from the Raleigh-based Division of Agricultural Statistics of its North Carolina Research Office. She encouraged researchers in the School of Agriculture to attend her experimental design course and recruited ca- pable applied statisticians to clevelop and teach basic statisti- cal methods. She made these statisticians available to consult with researchers on procedures for designing experiments and analyzing data. Most faculty had been trainee! in one of these disciplines and acquired some statistical training as a minor area. To secure at least one faculty member for every agricultural discipline, she had to start from scratch. "There weren't any statisticians to hire when ~ first started," she later wrote. "l haci to choose from other fields and train them."
GERTRUDE MARY COX 123 By the time Gertrude left Ames, ~ had hac! enough of Iowa's winters ant! told her ~ would like to join her group whenever a mathematical statistician position became avail- able. In 1942, ~ transferred from North Carolina State's Mathematics Department to handle statistical consulting with agricultural economists. Gertrude had decicled that it was necessary to bolster the methods courses with courses in sta- tistical theory; a graduate program was in the offing. Another innovative feature of the Cox statistics program was a series of one-week working conferences on specific top- ics, such as agricultural economics and rural sociology, bio- logical ant! nutritional problems, agronomic ant! horti- cultural problems, plant sciences, animal sciences, quality control, nutrition, industrial statistics, soil science, and plant breeding. Gertrude later obtained outside funds to hoist two summer conferences in the mountains of North Carolina, which were attended by statisticians from throughout the United States and abroad. In addition to experimental and mathematical statistics, these conferences covered many re- search areas involving statistics, including life testing, oper- ations research, clinical trials, surveys, pasture and rotation experiments, and genetics. Many were held cluring World War Il. Gertrude, realizing the importance of quality control methods to the war effort, inclucled engineering statisticians on the faculty. During this period Gertrude realized still another dream. She had become a close friend of Frank Graham, the Uni- versity's president, who had been instrumental in starting the statistics program in 1940. In 1944, Dr. Graham helped her get a grant from the General Education Board, founded by John D. Rockefeller, to establish and direct an Institute of Statistics to improve statistical competency in the South. This grant enabled her to add six faculty members to her depart- ment, including W. G. Cochran, who was to develop a grad-
124 BIOGRAPHICAL MEMOIRS Hate program. In 1945, the General Education Board made an additional grant to establish a Consolidated University of North Carolina Institute of Statistics, with a Department of Mathematical Statistics at Chapel Hill, to concentrate on graduate training and research in statistical theory. With complementary graduate programs, the two departments produced many outstanding applied and theoretical statisti- cians. Gertrude remained as heac! of the Experimental Statistics Department in Raleigh until 1949, when she decided to ad- minister the Institute almost full-time, with the exception of teaching her course in experimental design. In the School of Public Health at Chapel Hill, she helped establish the Biosta- tistics Department, the Social Science Statistical Laboratory in the Institute for Research in Social Science, and the Psy- chometric Laboratory in the Department of Psychology. These two laboratories were a culmination of Gertrude's life- long interest in the use of statistics to study human relation- ships. During this time, North Carolina State statisticians began visiting a number of experimental stations to assist research programs in the use of statistical methods. Cox's Institute coordinated a number of short courses for researchers in industry and the physical sciences. One of her most impor- tant accomplishments was her successful effort, along with Boyd Harshbarger of Virginia Polytechnic Institute, to estab- lish the Southern Regional Education Board's Committee on Statistics to develop cooperative programs for statistics teach- ing, research, and consulting in the South. This Committee contributed tremendously to sound sta- tistical programs throughout the South and fostered the spirit of cooperation that Gertrude envisaged. From 1954 to 1973 it sponsored! a continuing series of six-week summer sessions and is now conducting an annual one-week Summer
GERTRUDE MARY COX 125 Research Conference modeled on the original Gordon con- ferences.2 Gertrucle's first contact with statistics came in the com- puting laboratory, and she remained a strong advocate of the integral connection between statistical analysis and an up-to- date computing facility. At Iowa State she hac! developer! an excellent computing laboratory. Early in 1941 she persuacled Robert Monroe, one of her chief associates there, to come to Raleigh to develop a similar facility. ~ remember those old Hollerith machines at Ames and Raleigh and the tremen- dous leap forward when IBM entered the electronic age. Gertrude Cox, naturally, had one of the first IBM 650s on a college campus, and North Carolina State subsequently de- signed for the 650 the best statistical software. From then on, Gertrude made certain that the Institute was in the forefront when it came to statistical software, and Raleigh statisticians designed the initial SAS programs. No account of Gertrude Cox's meteoric success at the Uni- versity of North Carolina would be complete without men- tioning her unique ability to secure outside financial support. Though her Institute was originally funded by General Edu- cation Boars! grants, Gertrude Cox persuaded the Rockefel- ler Foundation to support a substantial program in statistical genetics. She obtained funcis from the Ford Foundation for a joint program in dynamic economics with the London School of Economics. Finally, in 1952, she obtained a large grant from the General Education Board Matched by 1958) for a revolving research funs! enabling the Institute to fi- nance fundamental, nonsponsored statistical research for many years thereafter. 2 The Committee Cox and Harshbarger founded was still operating as of 1990 under the name of the Southern Regional Committee on Statistics. Though no longer affiliated with the Southern Regional Education Board, it continues to spon- sor summer research conferences.
126 BIOGRAPHICAL MEMOIRS Starting in 195S, Gertrude and seven other members of the North Carolina State statistics faculty worker! out proce- dures for establishing a Statistical Division in the proposed not-for-profit Research Triangle Institute (RTI) between Ra- leigh and Chapel Hill. RTI was established in ~ 959, and Ger- trucle retired from North Carolina State in 1960 to direct its Statistics Research Division, whose major component was the sample survey unit. Retiring from that post in 1965, she con- tinuect on as a consultant for many years, even occasionally teaching her design course at North Carolina State. During her five-year tenure, RTI- and especially the Statistics Divi- sion became an internationally recognizes! consulting and . ~ researc n organization. Gertrude Cox was a consultant to the Pineapple Research Institute of Hawaii, the WorIc! Health Organization in Gua- temala, the U. S. Public Health Service, the government of ThailancI, the Pan American Health Organization, and many other organizations overseas. She served on a number of gov- ernment committees including the U. S. Bureau of the BucI- get's Advisory Committee on Statistical Policy (1956-19581; the National Institutes of Health's Agricultural Marketing Service, Epidemiology, and Biometry Committees ~ ~ 959- 1964~; and the National Science Foundation's Office of Edu- cation (1963-1964) ant! Teacher Education Section (19661. Even after retirement she served on advisory committees to the Secretary of Health, Education and Welfare ~ ~ 970- 1973), the Bureau of the Census, and the Department of Agriculture (19741. PROFESSIONAL ACTIVITIES AND HONORS Gertrude Cox's major contribution to science was her abil- ity to organize and administer programs, but her early ac- complishments in psychological statistics and experimental design were wiclely recognized.
GERTRUDE MARY COX 127 Gertrude was a founding member of the International Biometric Society in 1947, served as editor of its journal, Bio- metrics, from 1947 to 1955, and was a member of its Council three times and president from 1968 to 1969. She was proud that she had attended every international meeting of the so- ciety, and, in 1964, was awarded an honorary life member- ship. She was also active in the International Statistical Insti- tute ant! was a member of its Council in 1949, treasurer from 1955 to 1961, ant! chairman of the Education Committee from 1962 to 1968. She was president of the American Sta- tistical Association (ASA) in 1956. She was a fellow of the American Public Health Associa- tion, the American Association for the Advancement of Sci- ence, the Institute of Mathematical Statistics, ant! the ASA. She was also a member of the Psychometric Society, the Royal Statistical Society, ant! the Inter-American Statistical Insti- tute. In recognition of her international reputation she was named honorary vice-presiclent of the South African Statis- tical Association, honorary member of the Societe Aclolphe Quetelet of BeIgium, and the Thai Statistical Association, and an honorary fellow of the Royal Statistical Society. She was a member of the honor societies Delta Kappa Gamma (eclu- cation), Gamma Sigma Delta (agriculture), Pi Mu Epsilon (mathematics), Phi Kappa Phi (scholastic), and Sigma Xi (sci- ence). In 195S, Gertrude Cox's alma mater, Iowa State Univer- sity, conferred upon her an honorary Doctorate of Science as a "stimulating leacler in experimental statistics . . . outstand- ing teacher, researcher, leacler ant! administrator . . . Her in- fluence is worIclwicle, contributing to the development of na- tional and international organizations, publications, and councils of her fielcl." In 1959 she received the highest rec- ognition the Consolidated University of North Carolina can confer upon its faculty the 0. Max Gardner Award. The
128 BIOGRAPHICAL MEMOIRS citation named her a "statistical frontierswoman"a phrase suggested by the title of her ASA presidential adclress, "Sta- tistical Frontiers." In 1970, North Carolina State University honored her once again by designating the building in which the Statistics Department is located Cox Hall, and in 1977 a Gertrude M. Cox Fellowship Func} was established for outstanding gracI- uate students in statistics. Her most treasured honor came in 1975, when she was elected to the National Academy of Sci- ences. TRAVELS Gertrude Cox was a worIcl traveller who particularly en- joyed working in developing countries where she could offer advice and inspiration. All of Gertrude's trips were carefully planned, usually with reservations at excellent hotels. Fasci- natecl by Egypt, she helpecl establish a statistical program at Cairo University and, cluring the months she spent there, toured many historical sites. She was especially excited by her visits to the Sinai and to Abu Simbel. Thailanc! was another of her particular favorites, and ~ was touched, when ~ visited Bangkok in 1982, by how much the Thais lover! her. She loved wearing dresses she had had macle from colorful Thai silk, and a grower of orchids since her visits to Hawaii in the late 1940s she struck up a close friendship with Rapee Sagarik, Thailancl's principal orchid expert. (She grew these beautiful orchids for pleasure, not profit, and enjoyed giving them to her friends, as my own family can attest.) C LO S I N G REMARKS Gertrude Cox loved people, especially children. She al- ways brought back gifts from her travels and was especially generous at Christmas time. She considered the faculty mem-
GERTRUDE MARY COX 129 hers and their families to be her family and entertained them frequently. She was an excellent cook and had two hobbies that she inclulged during her travels: collecting dolIs and sil- ver spoons. She learned chip carving and block printing at an early age and spent many hours training others in these arts. She loved garclening, anal, when she had had a partic- ularly hard clay with administrators, would work off her ex- asperation in the garden. She had a fine appreciation for balance, clesign, anti symmetry. In 1976, Gertrude learned that she had leukemia but re- mained sure that she would conquer it up to the end. She even continued construction of a new house, unfortunately not completecl until a week after her death. While uncler treatment at Duke University Hospital she kept detailed re- corcis of her progress, and her doctor often referred to them. With characteristic testy humor she called herself"the ex- perimental unit," and diect as she had lived, fighting to the encI. To those of us who were fortunate to be with her through so many years, Raleigh will never be the same.
130 BIOGRAPHICAL MEMOIRS SELECTED BIBLIOGRAPHY 1930 A statistical study of industrial science students of the class of 1926. Iowa Acad. Sci. Proc., 37:337-41. 1931 The use of the individual parts of the aptitude test for predicting success of students. Iowa Acad. Sci. Proc., 38:225-27. 1933 With C. W. Brown and P. Bartelme. The scoring of individual per- formance on tests scaled according to the theory of absolute scaling. l. Educ. Psychol., 24:654-62. 1935 With G. W. Snedecor. Disproportionate subclass numbers in tables of multiple classification. Iowa Agric. Exp. Stn. Res. Bull., 180:233-72. Index number of Iowa farm products prices. Iowa Agric. Exp. Stn. Bull., 336:297-328. 1936 With G. W. Snedecor. Covariance used to analyze the relation be- tween corn yield and acreage. J. Farm Econ., 18:597-607. 1937 With H. Gaskill. Patterns in emotional reactions: I. Respiration. The use of analysis of variance and covariance in psychological data. l. Gen. Psychol., 16:21-38. With G. W. Snedecor. Analysis of covariance of yield and time to first silks in maize. J. Agric. Res., 54:449-59. With W. P. Martin. Use of discriminant function for differentiating soils with different asotabacter populations. Iowa State Coll. l. Sci., 11:323-32. 1939 The multiple factor theory in terms of common elements. Psy- chometrika, 4:59~8.
GERTRUDE MARY COX 131 With M. G. Weiss. Balanced incomplete block and lattice square designs for testing yield differences among large numbers of soybean varieties. Iowa Agric. Exp. Stn. Res. Bull., 257:290- 316. 1940 Enumeration and construction of balanced incomplete block con- figurations. Ann. Math. Stat., 11:72-85. With R. C. Eckhardt and W. G. Cochran. The analysis of lattice and triple lattice experiments in corn varietal tests. Iowa Agric. Exp. Stn. Res. Bull., 281: 1-66. 1941 With H. V. Gaskill. Patterns in emotional reactions. II. Heart rate and blood pressure. l. Gen. Psychol., 23:409-21. 1942 With H. McKay, et al. Length of the observation period as a factor in variability in calcium retentions. I. Home Econ., 34:679-81. With H. McKay, et al. Calcium, phosphorus, and nitrogen metab- olism of young college women. I. Nutr., 24:367-84. 1944 Modernized field designs at Rothamsted. Soil Sci. Soc. Am. Proc., 8:20-22. Statistics as a tool for research. l. Home Econ., 36:575-80. 1945 Opportunities for teaching and research. I. Am. Stat. Assoc., 229:71-74. 1946 With W. G. Cochran. Designs of greenhouse experiments for sta- tistical analysis. Soil Sci., 62:87-98. 1950 The function of designs of experiments. Ann. N.Y. Acad. Sci., 52(Art. 6~:800-7.
132 BIOGRAPHICAL MEMOIRS With W. G. Cochran. Experimental Designs. New York: John Wiley & Sons, Inc. 1953 Elements of an effective inter-American training program in agri- cultural statistics. Estadist., 11: 120-28. 1957 Statistical frontiers. l. Am. Stat. Assoc., 52:1-12. (Institute of Sta- tistics Reprint Series, no. 99.) With W. G. Cochran. Experimental Designs. Ed ed. New York: John Wiley & Sons. 1964 With W. S. Connor. Methodology for estimating reliability. Ann. Inst. Stat. Math. The Twentieth Anniversary, 16:55-67. 1972 The Biometric Society: The first twenty-five years (1947-19721. Biometrics, 28~2) :285-311. 1975 With Paul G. Homeyer. Professional and personal glimpses of George W. Snedecor. Biometrics, 31~2~:265-301.