An excerpt from my draft book;
Lavender Rising: An Intersectional History of the LGBT+ Struggle
Lives as dreadful as Ronald Aylmer Fisher’s can have profound generational effects. George Box reportedly married illegally in America (according to English law), in order to become Fisher’s son-in-law (Irwin Guttman, Dennis Lindley, personal communication), while leaving his actual wife with his children in England. Lindley, the 'Living God of the Bayesians' had described Box’s marriage to Joan Fisher in very negative terms, and this caused a lengthy feud between Box and Lindley (Box, personal communication).
Box depended heavily for his academic fame, credibility and rationality on his co-authors who included, among many others, the pre-eminent statisticians Gwilym Jenkins, Stuart Hunter, Norman Draper and George Tiao. My colleague Michael Meyer thought that Box ‘got away with it’ because of his occasional scientific ‘nuances’ and flashes of apparently deep insight.
When I joined George Box and his second (or third out of four if Irwin is to be believed) wife Joan Fisher-Box on Christmas Day 1979 in their plush mansion, Joan threw a fit while I was ahead in a second game of Scrabble, and she and her daughter fled to their rooms as if Armageddon had struck.
During early 1980, I effectively refuted George’s very strange heuristics in his interminably long, single-authored paper ( Box ,1980) by making his ‘prior predictive’ p-value for checking the sampling model look very silly when the sample size is large, after a neat piece of asymptotic mathematics. George promptly threw a fit and totally ignored my effective refutation of his much-hyped but utterly naive criterion.
Feeling as put down as Pynchon’s obsequious character, Teddy Bloat, I didn’t dare to publish my asymptotics in the discussion of Box’s much-acclaimed but deeply flawed 1980 invited paper to the Royal Statistical Society, particularly as he seemed to regard the paper as ‘another monument to his own greatness’. See Exercise 6.1.a on pp248-9 of Leonard and Hsu (1999) for the mathematical details which I published, with trepidation, 20 years after deriving them.
The question as to whether George E.P. Box (the vividly portrayed Croesus of sections 1.4, 1.5., and 1.9) was a closet-eugenicist will be addressed in detail elsewhere. His initial undergraduate training was in Chemistry, and his first post WW2 appointment was with the I.C.I. Dyestuffs division. He consulted with Ronald Fisher while working during WW2 on chemical warfare and the effects of poisonous gas on animals (Box, personal communication). I also recall Box’s post WW2 work on industrial chemicals and pharmaceuticals, chemical engineering, fractional factorial designs, Box-Wilson central composite designs, Box-Behnken designs, and Deming/ Taguchi-style total quality management ( See Moreno and Gitlow, 1999), and Box’s encouragement, on the quiet, at MRC of ‘Army Math’-style statistical research with military applications (‘lollipops’ for the generals).
Bisgaard (1992) extends an earlier paper on Quality Improvement by Box and Box’s much exploited and put upon graduate student from Gloucestershire, links several of the above-mentioned suspicious topics, and refers to much earlier papers out of Rothamsted Agricultural Research Station by the eugenicists Frank Yates and David Finney.
Box’s and Bisgaard’s research came out of the Center for Quality and Productivity Improvement on the notorious twelfth floor of the WARF building at UW Madison, where the Army Research Office and MRC no longer ruled. The research was instead financed by the right wing reactionary Alfred P. Sloan foundation. The material discussed by Bisgaard intersects with total quality management, and with the genetic justifications for experimental designs described by Mazumdar (1992, Ch3). This material has strong potential in terms of applications in eugenics.
As of 26 May 2023, this is all grist for the gander. While I strongly suspect that the much revered Box was a highly secretive eugenicist, the jury’s still out.
George Box liked Gilbert and Sullivan and the BBC Radio Goon Show, and wrote amusing scripts for the annual Department of Statistics Christmas Skits, some a bit biting in subtle sorts of ways. Maybe George is singing Gilbert and Sullivan’s ‘I’ve got a little list’ in his hereafter. For modern lyrics please listen to Suart (2015). The singer could have been hand-picked.
Richard Suart (2015) I’ve got a little list. English National Opera
https://www.youtube.com/watch?v=CWo_3CIcTBQ
George Box (1980) Sampling and Bayes' Inference in Scientific Modelling and Robustness (with Discussion) Journal of the Royal Statistical Society A. Series 143 (4), pp383-430
Thomas Leonard and John Hsu (1999) Bayesian Methods: An Analysis for Statisticians and Interdisciplinary Researchers. Cambridge: Cambridge University Press.
Henry Moreno and Howard Gitlow (1999) A comparison of Plato's Republic and Deming's System of Profound Knowledge International Journal of Applied Quality Management 2 (1) pp25-40
Søren Bisgaard (1992)The Design and Analysis of Inner and Outer Array Experiments
CQPI Technical Report No.90, Center for Quality and Productivity Improvement (UW Madison) https://minds.wisconsin.edu/bitstream/handle/1793/69151/r090.pdf?sequence=1
Richard Suart (2015) I’ve got a little list. English National Opera
https://www.youtube.com/watch?v=CWo_3CIcTBQ
No comments:
Post a Comment