American Statistical Association History of Statistics

2/ He developed the sequential probability ratio test SPRT for quality control of munitions production between March & April. 1943 His paper was so important to the military it was classified until the end of WWII. Similar tests were developed independently by Barnard (1946) and Turing .

#OTD Abraham Wald b (d 13 Dec 1950) ASA Fellow 1945 Best known for his WWII work on aircraft survivability & survival bias, estimation and hypothesis testing. & sequential analysis. He was a member of the Statistical Research Group at Columbia formed in 1942 to support the war effort. 1/2

#OTD 1933 Jerzy Neyman (ASA Fellow 1942) and Egon Pearson (Honorary ASA Fellow) read a paper to the Royal Society where they first use the terms “Type I and Type II error”. In 1928, they had called them “first and second source of error” & in Feb 1933 “errors of first and second kind”.

5/ Probably a bit biased by having developed the concept of the Average Man, Quetelet thought Verhulst’s early death was because he was too tall (he was 1.89 m, 6ft). It was more likely due to tuberculosis.

4/ Quetelet was not convinced about the model's utility because there was no counterpart in physics & he had his own competing model.

3/ He then said he didn’t have time to update his curves & anyway he only published them b/c his professor Adolf Quetelet made him.

2/ Verhulst published 3 papers on the logistic equation between 1838 & 1847. He didn’t say how curves were fitted, and nobody knows why he chose to call it 'logistic'. He checked the model empirically against population data for 4 countries.

Happy Belated 🎂to Pierre François Verhulst b 28 Oct 1804 (d 15 Feb 1849), who developed the logistic function as a model for limited population growth. Forgotten for almost 80 years, it was rediscovered by Pearl & Reed (1920) & Yule (1925) who revived the name ‘logistic’. 1/5

#OTD 1908 Margaret Gurney b (d 19 Mar 2002) ASA Fellow 1968. At the US Census Bureau she pioneered methods of variance estimation, sample surveys & non-sampling error. Awarded Department of Commerce Silver Medal for her work as an international statistical consultant & trainer.

#OTD 1893 Karl Pearson publishes a letter in Nature suggesting the method of moments for estimating parameters of asymmetrical frequency distributions. He trained as a physicist & borrowed the concept from physics, but it was first used by Chebyshev in his 1887 proof of the central limit theorem

Russell's diagram first appeared in print in Nature May 7 1914

#OTD Henry Norris Russell b (d 18 Feb 1957). An astronomer, he codeveloped the Hertzsprung-Russell diagram showing relationships of star luminosity & temperature. It is now a classic data set demonstrating the effect of extreme data outliers on regression model fits

2/His student Betty Allan did all the hard tedious calculations. Fiducial inference was never widely accepted: Neyman called it a “conceptual fallacy”, Savage said it was an attempt to "make the Bayesian omelette without breaking the Bayesian egg"

#OTD 1930 Ronald Fisher writes his first article on fiducial inference. He uses the observed sample distribution of r to generate a probability distribution (“fiducial density”) for the unknown population correlation coefficient ρ without priors. 1/2

8/ It was verified in 2005 by Georges Gonthier using general-purpose interactive theorem-prover software. So much for Lewis Carroll's idea of turning it into a party game.

7/ The proof stirs up a storm of controversy because it cannot be conventionally verified by manual calculations. The New York Times even refuses to report on it for a while because so many previous proofs & disproofs had been wrong.

6/ In 1976 by Kenneth Appel and Wolfgang Haken publish a proof more than 100 pages long & requiring hundreds of hours of computer time. This may be the first ‘computer proof’.

5/ In 1886 Lewis Carroll makes it into a 2-person party game. It is set as a challenge to schoolboys with the proviso that “no solution may exceed one page, 30 lines of manuscript, and one page of diagrams."

4/ In 1879 Alfred Kempe publishes in Nature his claim that he had solved it (without proof) & is made FRS as a result. His result is later shown to be incorrect.

3/ In 1878, Arthur Cayley writes the first paper on the conjecture. It is not a proof but explains why the problem is so hard. Francis Galton suggests Cayley publish it in the Proceedings of the Royal Geographical Society.

2/ De Morgan’s student Francis Guthrie had noticed that only 4 colors were needed to color a map of England such that no adjacent areas had the same color. The question was published in The Athenaeum in 1854 under the initials F.G.

Augustus de Morgan

#OTD 1852 Augustus De Morgan first reports on the four-color conjecture (the problem of ‘unavoidable sets’) in a letter to William Rowan Hamilton. The problem is finally solved in 1976. 1/8🧵

#OTD 1903 Malcolm C. Rorty (ASA Fellow 1921, ASA President 1932), then an engineer at Bell Telephone, writes a memo that turns into a pioneering paper on the application of probability theory to estimate future service demand.

#OTD 1854 Florence Nightingale leaves for the Crimea with her aunt & 38 nurses. They arrive 13 days later. After 2 strenuous years, she returns to England to work with William Farr, reforming the collation & communication of medical statistics.

#OTD 1850 Gustav Fechner publishes Fechner’s Law describing the relationship between the magnitude of a physical stimulus & its perceived intensity. It was a precursor of probit regression later developed by Chester Bliss who rediscovered it & coined the term.

The Bisset Hawkins Medal is a triennial award made by the Royal College of Physicians of London for advances in public health and sanitary science

#OTD 1796 F Bisset Hawkins b (d 7 Dec 1894) an English physician, he promoted statistical analyses over clinical opinion for assessing treatment efficacy. The Bisset Hawkins Medal is awarded for work in sanitary science and public health.

Charles Sanders Peirce

Joseph Jastrow

#OTD 1884 Charles Sanders Peirce & Joseph Jastrow publish the first randomized and blinded trial in experimental psychology. Random sequence allocation was performed by shuffling ordinary playing cards, blinding by conducting experiments in pairs with allocation concealed from one partner.

Karl Pearson in 1890

Walter Frank Raphael Weldon

#OTD 1900 WFR Weldon writes to Karl Pearson about Gregor Mendel's 1866 paper on the inheritance of traits in peas. They conclude Mendel’s ratios are too good to be true, triggering a bitter decades-long feud between them (the ‘Biometricians’) & the ‘Mendelians’ (Bateson, Saunders, and Punnett).

3/ Karl Pearson said “While the existence of ten fingers in man is a distinct advantage in the matter of personal identification - or if we like a distinct misfortune to the criminal - it is also something of a misfortune to the geneticist”

2/ Although confident the data showed general evidence of heritability she was perturbed by the low correlation coefficients, later admitting that 2 prints per person were not enough to determine the “intensity of heredity”

In Oct 1920 Ethel Edgerton (1874-1954) publishes an ambitious study attempting to measure the heritability of fingerprints. Francis Galton had pioneered their forensic use. She analysed Galton’s collection of several hundred familial forefinger prints. 1/3