STATISTICIANS IN HISTORY

EGON S. PEARSON

gon Sharpe Pearson was a long and leading figure on the world statistical scene and an honorary member of the EAmerican Statistical Association. He had been living in retirement in Sussex and, in spite of a heart condition, had been working on the history of statistics until a few months before his death in 1980. Pearson was born in London at a time when his father, Karl Pearson, was putting forward his famous system of frequency curves. Educated at Winchester and at Trinity College, Cambridge, Egon Pearson’s initial research was in solar physics. He joined Karl Pearson’s department of applied sta- tistics at University College London in 1921 and attended all of his father’s lecture courses. In 1922 there began the stream of important publications that soon established the younger Pearson’s reputation. When Karl Pearson retired in 1933 his department was split in two; Fisher was appointed Galton pro- By Herbert A. David fessor of eugenics and Pearson was appointed reader and then professor of statistics in 1935. With his father’s death in 1936 First published in The Pearson assumed the editorship of Biometrika, retaining it until American Statistician, his retirement in 1966; he continued until 1975 as editor of Biometrika auxiliary publications. May 1981, Vol. 35, No. 2 Karl Pearson deeply influenced his son by force of example and by providing him with many of the tools for future work. Egon Pearson’s statistical philosophy was shaped to a much greater extent by W.S. Gosset (“Student”), particularly by Gosset’s pioneering work in small-sample theory, which was later put on a rigorous footing and much extended by R.A. Fisher. Pearson was excited by these developments with which his father never felt wholly comfortable. In the 1920s both Gosset and Pearson became concerned with the too-ready assumption of normality that followed Fisher’s great success in development an elegant small-sample distribution theory under normality. As a result, Pearson developed a lifelong interest in what would now be termed robustness studies, with Tippett’s random numbers providing the means for simulation until the computer could take over. Also in the 1920s Pearson began the extraordinary 10-year collaboration with Jerzy Neyman for which both men are best known despite their numerous important individual contribu-

4 AMSTATNews tions. The Neyman-Pearson theory of contributed much to an understanding testing statistical hypotheses has become of issues in which he or his father were an integral part of every statistician’s principals. His biographical accounts in Egon Pearson education and vocabulary. In 1928 the the 1930s of Karl Pearson and Gosset likelihood ratio criterion was put for- are masterly. It appears however, to ward. No optimality was claimed, but have required the prodding of M.G. the criterion was shown able to repro- Kendall to persuade Pearson to turn Fisher’s work in high duce many existing tests and to provide more systematically in later life to such regard and was able appealing new ones. A major break- historical writing. Some of this work was to write in later years through occurred in 1933 with the opti- published in Biometrika in a still ongo- that he had long got- mality result that has become known ing series of articles (by various authors) ten used to Fisher’s as the Neyman-Pearson fundamental on the history of probability and statis- never having men- lemma. This settled the testing of a sim- tics, a series that with Kendall’s help was tioned his name in ple (i.e., fully specified) hypothesis instituted by Pearson in 1955. In 1978 print! With Fisher against a simple alternative; later papers Pearson completed the editing of lec- safely in Cambridge, dealt with the more complicated situa- tures, delivered by his father from 1921 the earlier postwar tions arising when null and alternative to 1933, on the history of statistics in the years were rather hypotheses are not both simple. Pearson 17th and 18th centuries. For about two pleasant. Students was the initiator of the partnership. years before his death in 1980, Pearson during this period Correspondence with Gosset, begun was engaged in editing a letter exchange found Pearson to be some six months before that with between Gosset and Karl Pearson (1908- an unassuming man Neyman, shows, according to Pearson, 1935) and between Gosset and himself and his lectures the that the concepts of the alternative (1926-1937). embodiment of common sense. He was hypotheses, two kinds of error, and University College always more concerned with the essen- use of the likelihood ratio principle London was closed tial ideas and their application than with to determine tests statistics were from 1939 to 1945 the mathematics, and would punctuate already formulated before being put and Pearson, togeth- his presentation with a large number of to Neyman. The latter’s contribution er with three mem- illustrative charts. Even though he did was absolutely bers of his staff, con- not regard himself as made to be a essential, how- stituted what was department head, he managed to attract ever, to give essentially an opera- to his staff, for significant periods, such precision to tions research group eminent people as Neyman, B.L. Welch, Pearson’s “loose in the Ordnance F.N. David, N.L. Johnson, and Hartley. ideas.” Neyman Board, Ministry of The result was a very stimulating envi- provided mathe- Defense. In spite of ronment. matical refine- bombings, V1 fly- Karl Pearson As editor of Biometrika Pearson ment and gener- ing bombs, and V2 established a reputation for fairness and ality, and took rockets, Pearson promptness. He was helpful and con- the lead in the found this assign- structive whenever possible and at times pursuit of opti- ment, in a way a would become deeply involved with a mal approaches. research job, “unde- W.S. Gosset submitted paper. One would never Pearson man- niably personally guess from his distinguished perform- aged to launch a rather enjoyable.” ance over a 30-year period that he second very successful collaboration, He had not been in regarded the editorship as a responsibil- this time with H.O. Hartley. Toward active service during ity thrust upon him. Apparently he the end of the 1930s he saw the need World War I because would have preferred to be left alone to for a complete revision of Karl of a heart murmur, get on with his own ideas. Pearson’s two-volume Tables for and the war work Pearson was inevitably the recipient Statisticians and Biometricians. Over provided an escape of many honors, including a C.B.E. for the years new tables were computed from the heavy dual his contribution to the war effort, a term by Pearson and Hartley, as well as oth- responsibility of Jerzy Neyman as President of the Royal Statistical ers, to be collected in Biometrika Tables running a depart- Society, and election as Fellow of the for Statisticians of which Volume 1 ment and editing Royal Society. Of most enduring signifi- appeared in 1954 and Volume 2 in 1972. Biometrika. cance, however, are the two volumes of With their attractive layout, easy means Pearson, who had none of his father’s Pearson’s selected papers and his joint for interpolation, and extensive, helpful aggressiveness, suffered during the papers with Neyman, issued by the introductory material, these tables have 1930s from the hostility of R.A. Fisher, Biometrika trustees in 1966 to celebrate been widely recognized as models of who was on the floor above. Naturally his 30 years as editor. He has put all stat- their kind. there were jurisdictional problems isticians in his debt. His modesty and Endowed with a sense of history and regarding the scope of their two depart- warmth will always be cherished by his an exceptionally lucid style, Pearson has ments. Pearson, however, held most of many friends. ■

SEPTEMBER2004 5