(UC3M), (EIO) and (EIO). Received on November 28, 2023.
This academic year, the Facultat de Matemàtiques i Estadística (FME) is honoring the legacy of one of the most influential statisticians of all time, George E.P. Box (1919-2013). The keynote lecture of this dedication was delivered on September 14 by Professor G. Geoffrey Vining (see the Chronicle of that event authored by Professor Marta Pézez Casany in this issue).
On November 22, 2023, professor Daniel Peña, who was a friend and collaborator of Box, shared his reflections on the life and contributions of the eminent statistician in a lecture delivered at the FME. This piece is based on the slides he presented on that occasion.
Box was born into a working-class family in England. At age 16, he started working as an assistant in a chemical company. At age 20, he was studying chemistry at the University of London when he was called up for military service during World War II. In the British Army, he was involved in the design and analysis of experiments with poison gas, and turned to Ronald A. Fisher for advice. After the war, Box earned an undergraduate degree in mathematics and statistics and in 1953 he obtained a Ph.D. in statistics from the University of London under the mentorship of Egon Pearson.
After obtaining his Ph.D., Box worked at Imperial Chemical Industries as a statistician. During that time, he developed the foundations of evolutionary operation, which is based on sequential experimentation and continuous process improvement [1]. He took a one year leave to visit North Carolina State, where he met Professor Gertrude Cox and other prominent American statisticians. He and his family enjoyed his year in the United States and moved to Princeton in 1956. At Princeton, Box served as the Director of the Statistical Research Group. He further developed industrial statistics and worked on revolutionary work in the field of time-series analysis with Gwilym Jenkins (see [2, 3]).
In 1960, Box moved to the University of Wisconsin-Madison, where he wrote his most celebrated works in Bayesian data analysis, design of experiments, and quality control [4–6]. He built a strong Department of Statistics serving as its Chair, attracting some of the most talented statisticians at the time. Under his leadership, Wisconsin-Madison became the best applied statistics department in the US. Additionally, he was successful in conducting and promoting interdisciplinary research between statisticians, engineers, and scientists in a wide array of fields. Towards the end of his career, he founded the Center for Quality and Productivity Improvement with William G. Hunter.
When Box officially retired as a professor in 1992, he continued his involvement in the Center for Quality and Productivity Improvement. He kept on working on interdisciplinary problems and took a renewed interest in statistical process control and monitoring. He also wrote the memoir [7], a book that intertwines personal and professional experiences, providing insights into his life and influence as a statistician.
Professor Peña first saw Box at a conference on time-series analysis in Cambridge in 1976. At the time, Box’s work was controversial: theoreticians thought that the work lacked mathematical rigor, and applied statisticians thought it was too complicated. In 1979, Peña interacted with Box at the first Valencia Bayesian meetings (organized by J. M. Bernardo) and later invited him and George Tiao to give a course in Madrid. Peña and Box became friends and collaborators; they visited each other several times and they co-authored groundbreaking work on dynamic factor models [8]. In 1994, Box became doctor honoris causa by the Universidad Carlos III.
In the mid 1980s, Box visited the Statistics department at UPC invited by Albert Prat, where he taught a course on time series analysis together with Tiao, and he visited the department several times after that. In the late 1990s, Box spent a year at Universidad de Cantabria to work on his book on statistical process control [9] in collaboration with Professor Alberto Luceño.
Professor Peña sees George Box as a pioneering data scientist: his work on sequential experimentation, model combination, dimensionality reduction, and exploration of non-linear surfaces lies at the core of the field. Box’s work is alive in our everyday life, from the recommendations we get from streaming services (which are based on sequential design of experiments) to the answers we get from ChatGPT (which are based on neural networks that handle non-linearities in ways that are related to Box’s work on transformations).
References
[1] G. E. P. Box and K. B. Wilson, On the Experimental Attainment of Optimum Conditions, Journal of the Royal Statistical Society (Series B: Statistical Methodology) 13 (1951), 1-38.
[2] G. E. P. Box and N. Draper, Evolutionary Operation: A Statistical Method for Process Management, John Wiley & Sons, 1969.
[3] G. E. P. Box and G. M. Jenkins, Time Series Analysis: Forecasting and Control, Holden day, 1976.
[4] G. E. P. Box and G. Tiao, Bayesian inference and statistical analysis, Addison Wesley, 1973.
[5] G. E. P. Box and N. R. Draper, Empirical model-building and response surfaces, John Wiley & Sons, 1987.
[6] G. E. P. Box, J. S. Hunter, and W. G. Hunter, Statistics for experimenters: design innovation and discovery (2nd ed.), Wiley-Interscience, 2005.
[7] G. E. P. Box, An accidental statistician. The life and memories of George E.P. Box, John Wiley & Sons, 2013.
[8] D. Peña and G. E. P. Box, Identifying a simplifying structure in time series, Journal of the American statistical Association 82 (1987), no. 399, 836-843.
[9] G. E. P. Box and A. Luceño, Statistical Control: By Monitoring and Feedback Adjustment, John Wiley & Sons., 1997.