H. O. Hartley Award

Herman Otto Hartley (1912-1980)

Herman Otto Hartley (HOH) was born Hermann Otto Hirschfeld in Berlin, Germany in 1912. While short in stature (5’ 1’’), Hartley stands tall professionally being widely recognized as one of the most prominent and influential statisticians of the twentieth century. He was affectionately known and referred to as HOH throughout his illustrious career. In 1934, shortly after completing his Ph.D. in mathematics from Berlin University, Hartley migrated to England to launch the start up of his highly productive career in many branches of statistics. For example, an early highly regarded and frequently cited Hartley paper in contingency table analysis provided the mathematical basis for the widely used multivariate technique called “correspondence analysis” (Proceedings of the Cambridge Philosophical Society, 1935). He did post-graduate work under John Wishart at Cambridge (1934-36) where he earned a Ph.D. in mathematical statistics in 1940. He also obtained a D.Sc. in mathematical statistics from the University of London. During the period 1936-38, he took a post as statistician with Harper Adams Agricultural College, Newport Shropshire, England, where he met and married Grace. At Adams he designed and analyzed poultry experiments at three Agricultural Experiment Stations. He lectured agriculture students and also published seven manuscripts that clearly exhibited his cleverness and innovative use of numerical techniques.

Prior to and during World War II (1938-1946), Hartley was heavily involved working as a Scientific Researcher with Scientific Computing Services, London, and with L.J. Comrie, a famous computer scientist. Much of his work during WWII involved wartime surveys and military applications.

Hartley’s first academic appointment after WWII began as lecturer in statistics at University College, London, where he became a close colleague of Egon Pearson. This collegial relationship led to the famous Pearson and Hartley two-volume series Biometrika Tables for Statisticians published in 1954 and 1972, respectively. He also developed Hartley’s F-max test for equality of variances (Biometrika, 1950). He was elected a Fellow of the Institute of Mathematical Statistics shortly after a brief association with S.S.Wilks at Princeton University in 1948-49.

In 1953, Hartley took a one year visiting Research Professorship position at Iowa State College. From 1954-63, Hartley remained at Iowa State as Professor of Statistics with a research emphasis in sample survey theory and methods as well as statistical computing. He was initially involved in scientific computing consultation for members of the entire Iowa State College and was available for statistical consultation as well. Hartley was instrumental in growing the stature and ranking of the Department of Statistics at Iowa State during this period. In fact, Hartley was the inventor of EM algorithm (Biometrics, 1958) and the EM algorithm is now one of the most widely used estimation methods. He was elected as Fellow of the American Statistical Association (ASA) in 1953 and was elected President of the Eastern North American Region of the Biometric Society in 1959. He also was an elected member of the International Statistics Institute.

Hartley left Iowa State in September 1963 to become distinguished professor and inaugural Director and founder of Texas A&M University’s Institute of Statistics. While at Texas A&M, he grew the Institute by hiring highly acclaimed statistical faculty members including two of his Iowa State doctoral students J.N.K. Rao and Ron Hocking. He published prolifically in top-tiered journals during his tenure at Texas A&M. Major research efforts included work in sample survey design for which he was highly regarded, in addition to mathematical optimization, estimation with incomplete data, variance component estimation and the development of designed experiments to estimate safe doses of carcinogenic agents.

While at Texas A&M, Hartley and J. N. K. Rao published a highly acclaimed manuscript using a maximum likelihood method for finding variance components in linear mixed models (Biometrika, 1967). They also developed a new estimation theory for sample surveys (Biometrika, 1968) that provided one of the earliest nonparametric arguments in what may be the very first maximum empirical likelihood estimator. Additionally, Hartley and R.R Hocking published a seminal manuscript on the analysis of incomplete data (Biometrics, 1971).  Hartley served as President of ASA in 1979 and was recognized and honored many times during his career that included winning the prestigious S.S. Wilks Medal in 1973.

After his mandatory retirement as Director of Texas A&M’s Institute of Statistics in 1977, Hartley remained active at A&M until 1979, when he took a full-time visiting professorship position at Duke University until his death in December of 1980 after undergoing heart surgery.

His legacy at Texas A&M is still unfolding with an annual Hartley Award given to a former Aggie Department of Statistics alumnus for distinguished service to the profession and a biannual lecture series named in his honor.

-Author: Michael Kutner, Ph.D (1971); Hartley Award Recipient (1984)

Primary documents utilized herein include:

Smith, William B. (1981). In Memoriam: Herman Otto Hartley (1912–1980). The American Statistician, 35(3): 142-143.

Smith, William B. (2005). Statistical Careers in History: Herman Otto Hartley.  Amstat News, Sept.:10-11.