Sunday, April 24, 2016

Harold Jeffreys

Sir Harold Jeffreys (1891–1989) was an English mathematician, statistician, geophysicist, and astronomer. As an noted physical scientist, he was the first one to demonstrate that the core of the Earth was liquid; but more along our line of interest, he contributed to the revival of Bayesian statistics with a prior probability distribution, known as the Jeffreys prior, in the 1930s. His astronomer-mathematician life allowed him to develop this highly mathematical based method, aiming for applications in the physical sciences. Meanwhile, his expertise in multiple fields shaped his philosophy of science--as a Bayesian statistician Jeffreys regarded probability as a degree of reasonable belief. Prior probability distribution, as its name indicated, is one's belief assigned to the distribution before a statistical test, and Jeffreys prior is specifically a non-informative prior for any parametric model.  
Jeffreys was born in Fatfield, England, and his father was the headmaster of a secondary school. After studying mathematics, physics, chemistry, and geology at Armstrong College in Newcastle, he went to St John's College in Cambridge and became a fellow in 1914. He worked in applied math and physics related job positions before the 1930s, and then returned to Cambridge. There, Jeffreys taught mathematics until 1932. From 1932 until 1946, Jeffreys taught geophysics. His students did not regard his lecturing skills highly; his teaching style was apparently quite unclear, which led many students to drop out of his courses. Nevertheless, Jeffreys became Plumian Professor of Astronomy and Experimental Philosophy in 1946.
Jeffreys was further recognized for his scientific and mathematical contributions in 1953, when he was knighted. In 1960, The Royal Society awarded Jeffreys their Copley Medal in recognition for his work in geophysics, astronomy, and probability. Regarding his geophysical work, Jeffreys coauthored The Earth: Its Origin, History and Physical Constitution (1924) and Earthquakes and Mountains (1935), in which he explained the origin of monsoons and showed how cyclones contribute to the circulation of the earth’s atmosphere. Interestingly, Jeffreys was a critic of the theory of continental drift and later of plate tectonics. Despite the mounting evidence for these theories, Jeffreys remained skeptical against them, believing that there’s no single force great enough to move pieces of earth.
Although his major focus was the physical sciences, Jeffreys’ interactions with one of the most influential statisticians in the twentieth century, Ronald Fisher, lead him to publish his major statistical work: a book called the Theory of Probability (published in 1939). The two first began to interact in 1933, and discovered that they had opposing viewpoints on Bayesian statistics. Jeffreys embraced Bayesian statistics, while Fisher rejected Bayesian statistics with his notion of probability based on relative frequency. Their continued criticisms and responses finally ended in reconciliation. In the Theory of Probability, Jeffreys says about Fisher: “I have in fact been struck repeatedly in my own work, after being led on general principles to a solution of a problem, to find that Fisher has already grasped the essentials by some brilliant piece of common sense [...]”
Jeffreys was described as a shy man and a heavy smoker, who rode his bike well into his 90s. He was greatly respected for his warmth and intelligence. Jeffreys passed away at the age of 97 in Cambridge.

References:, “Bayes, Jeffreys, Prior Distributions and the Philosophy of Statistics” by Andrew Gelman, “Harold Jeffreys and R.A. Fisher” by John Aldrich

No comments:

Post a Comment