Wednesday, December 30, 2009
Wednesday Math, Vol. 103: Sir Ronald Fisher
The vast majority of mathematicians who are renowned for their work in the sciences are physicists. Sir Ronald Fisher (1890-1962) was an important figure in the fields of statistics, evolutionary biology and genetics. On the Wikipedia page, Anders Hald proclaims Fisher "a genius who almost single-handedly created the foundations for modern statistical science". There was a time when people might have believed that, but the modern view usually gives a lot of credit to a very important predecessor, Karl Pearson (1857-1936).
Pearson came up with many of the basic ideas of modern statistics a generation before Fisher, and was a particular thorn in Fisher's side early in the younger man's career, disagreeing with him in print and disputing his results and methods. Pearson was co-founder of an important British publication Biometrika, and Fisher tried to have papers published in the journal, only to have them rejected or severely edited. Fisher did not take this well, and became Pearson's sworn enemy, his bitterness lasting for decades after the older man died. Later in his career, Fisher became the editor-in-chief of Biometrika, and did everything he could to repudiate Pearson's ideas, most notably the measurements of skew and kurtosis, once considered basic tools in statistics but now no longer covered in elementary classes. Many of Pearson's cast aside ideas are making a comeback, but even Fisher could not make people throw out the Pearson correlation coefficient, often written as rx,y or with the lowercase Greek letter rho, still a central concept in the field.
I do not want anyone reading this post to think that Fisher was just a minor Ronny-come-lately in the field. Many of his ideas are still vital in the field, as is his classic text The Design of Experiments.
In the 1950's, the British physiologist Richard Doll published several papers linking cigarette smoke to disease, and Fisher took great exception. His main disagreement was based on the completely valid assertion that correlation is not causation, but he also had bones to pick with Doll's methodology. Fisher would give Doll no help in correct his methods, though some other leading lights in the field, including the Polish-American mathematician Jerzy Neyman, were willing to help, and as the decade wore on, statistically valid studies in both the United Kingdom and the United States were published showing the linkage, even though the actual carcinogen in tobacco smoke was not isolated until the 1970's.
Fisher was correct that correlation is not causation, but his adamant position had much more to do with his general bloody-mindedness and his love for smoking. By the time of his death in 1962, no other important statistician was agreeing with him that the studies were all poppycock, and after his death, no serious voice in statistics took up the banner of his crusade.
As I listen to the arguments on both sides of the debate on global warming, I am reminded of Sir Ronald Fisher. Many skeptics and deniers of the man-made causes of global warming point to groupthink being the cause of the large consensus of people in the field who accept the basic premise. But when they speak, I hear the ghost of Ronald Fisher, brilliant in so many ways but afflicted with that all too common human defect, the deep desire to believe that nothing is really our fault.