Bayes, Thomas
BAYES, THOMAS
(b. Bovingdon, Hertfordshire[?], England, 1702; d. Tunbridge Wells, Kent, England, 7 April 1761)
probability. For the original article on Bayes see DSB, vol. 1.
Bayes is rightly remembered chiefly for his work on the determination of the support for a statistical hypothesis provided by experimental data (or, on the determination of the probability of a cause from knowledge of an observed occurrence). However, he also provided results concerning the approximation of the skew beta distribution, the series expansion of log z!, and (with Richard Price) a series expansion of the Normal probability integral.
Genealogy. The eldest of the seven children of Joshua (who was not a Fellow of the Royal Society) and Anne (née Carpenter) Bayes, Thomas might well have been born in Bovingdon, near Hemel Hempstead in Hertfordshire. William Urwick (1884, p. 390) states that Joshua Bayes ministered at the Box Lane Chapel, in Bovingdon, after his ordination (in 1694), and remained there for about eleven years before returning to London. There he became an assistant to John Sheffield at St. Thomas's, Southwark, and he subsequently officiated at the dissenting chapel in Leather Lane in Hatton Gardens. Here Thomas later assisted him, before moving to Tunbridge Wells in the early 1730s.
Education. While it is uncertain where Thomas received his early education, there is no doubt that he later studied at Edinburgh University. The records of that institution for 1719 (Scottish style) show that Thomas was admitted on the nineteenth of February, and his name is also recorded in the List of Theologues in the College of Edinburgh since October 1711. Like many intending to become Nonconformist ministers, he did not take the MA degree, and returned to London licensed but not ordained.
The certificate detailing Bayes’s admission to the university library was signed by James Gregory, brother of the Scottish mathematician and astronomer David Gregory, his immediate predecessor in the chair of mathematics, and nephew of the illustrious James Gregory (1638–1675). David Gregory has been credited with the introduction of Isaac Newton’s Principia to Edinburgh students, and it is possible that his brother James continued with this practice. Thus Bayes might have been exposed to the fluxionary calculus, and this could well have accounted for the precision and understanding evinced in his Introduction to the Doctrine of Fluxions.
Works. In his Divine Benevolence Bayes argued that the works of God are inspired and motivated by benevolence, contra the view expressed in John Balguy’s earlier tract that rectitude was the guiding principle. Subsequently Henry Grove saw wisdom as God’s most fundamental attribute. All three of these tracts were studied at dissenting academies of that period.
In addition to An Essay towards Solving a Problem in the Doctrine of Chances, volume 53 of the Philosophical Transactions of the Royal Society contains a Letter from Bayes on a semiconvergent series. Here he showed that the well-known James Stirling–Abraham de Moivre series “for” log z! in fact diverges for any z (the divergence for z = 1 had been shown by the Swiss mathematician Leon-hard Euler some six years before Bayes’s death).
On the matter of the scholium in the Essay, Stephen M. Stigler interprets Bayes as showing that, for the balls-and-table example, Prob(M=p)=1/(n+1) for p=0, 1, …, n, a characterization of the uniform distribution on [0,1] due to Bayes. Suppose next that we have an event (say E) of whose probability (say 0) we are ignorant. Because the probability that E occurs p times in n trials is also 1/(n+1), 0 must again be uniformly distributed. That is, that 0 has a uniform distribution is implied by the constancy of Prob(E=p). When the scholium is viewed in this way, the problems sometimes thought to arise from the apparently arbitrary choice of a uniform prior disappear.
The Essay was followed by a Supplement, partly by Richard Price, published in volume 54 of the Philosophical Transactions. In addition to Bayes’s famous result, these two papers contain, in the rules, an approximation to the two-sided beta probability integral that is considerably better than the Normal approximation (in essence, the skew beta probability density function, obtained by Bayes as a posterior density, is approximated by a symmetric beta probability function multiplied by a factor tending in the limit to 1, the two betas having the same maximum and points of inflexion). Also to be found in the Supplement is a series expansion, by Price, of the Normal approximation to the posterior distribution derived in the Essay.
In the library of the Royal Society of London is a letter from Bayes to John Canton commenting on work by Thomas Simpson on errors in observations. Bayes’s view was that an increase in the number of observations will not cause a decrease in the probability of error if the measuring device is inaccurate. With this is connected Simpson’s assumption that errors in excess and in defect of the true value are equally probable.
A serendipitous find by David R. Bellhouse in the Kent County Archives in Maidstone, U.K., suggests that Bayes might well have been used by Philip Stanhope, 2nd Earl Stanhope, as a mathematical critic and commentator. Bellhouse discovered several items in the Stanhope Collection: (1) preliminary drafts of Bayes’s paper on series; (2) a derivation, in Bayes’s hand, of the series for log z! that does not use the semiconvergent series; and (3) letters between Bayes and Stanhope, supportive of the above suggestion, referring to some observations by Patrick Murdoch. There is also a note, in Stanhope’s writing, headed “The Reverend Mr Bayes’s Paper concerning Trinomial divisors.”
The Equitable Life Assurance Society in London holds a notebook by Bayes. With passages in French, Latin, English, and an obscure seventeenth-century shorthand, the notebook contains notes on matters as diverse as mathematics, electricity, optics, and music.
Death. Although giving up his ministry in 1752, Bayes probably remained in Tunbridge Wells, where he died in 1761, being interred in the family vault in the Bunhill Fields Burial Ground, near Moorgate, London. Despite frequent restorations of the vault, the inscriptions have at times been difficult to decipher. This is perhaps the reason for the Canadian philosopher Ian Hacking’s recording of Thomas’s death as the seventeenth, rather than the seventh, of April.
SUPPLEMENTARY BIBLIOGRAPHY
Bellhouse, David R. “On Some Recently Discovered Manuscripts of Thomas Bayes.” Historia Mathematica 29, no. 4 (2002): 383–394.
———. “The Reverend Thomas Bayes, F.R.S.: A Biography to Celebrate the Tercentenary of His Birth.” Statistical Science19, no. 1 (2004): 3–43.
Dale, Andrew I. Most Honourable Remembrance: The Life and Work of Thomas Bayes. New York: Springer-Verlag, 2003. Reprints of and commentaries on Bayes’s works.
Grant, Alexander. The Story of the University of Edinburgh during Its First Three Hundred Years. 2 vols. London: Longmans, Green, 1884.
Hald, Anders. A History of Mathematical Statistics from 1750 to 1930. New York: Wiley, 1998.
Stigler, Stephen M. “Thomas Bayes’s Bayesian Inference.’’ Journal of the Royal Statistical Society, ser. A, 145, pt. 2 (1982): 250–258.
Urwick, William. Nonconformity in Herts: Being Lectures upon the Nonconforming Worthies of St. Albans, and Memorials of Puritanism and Nonconformity in all the Parishes of the County of Hertford. London: Hazell, Watson, and Viney, 1884.
Andrew I. Dale
