Sir Ronald A. Fisher was a prominent mathematician who formalized and extended the field of statistics, and revolutionized the concept of experimental design. He worked for fourteen years as a research statistician and later held professorships in genetics, another field to which he made significant contributions. Fisher wrote some three hundred papers and seven books throughout his prodigious career.
The son of George Fisher, a partner in a fine arts auction firm, Ronald Aylmer Fisher was born in the north London suburb of East Finchley. The youngest of seven children, Fisher was a precocious child. In her biography, R. A. Fisher: The Life of a Scientist, Fisher's daughter Joan Fisher Box describes an incident that occurred when the scientist was about three years old. Fisher engaged his nurse in a breakfast-table conversation about the successive halving of the number two; after she answered the first three questions of his series, he concluded that "half of a sixteenth must be a thirty-toof."
During his school years at Stanmore Park and Harrow schools, Fisher developed a facility for visualizing complex geometrical relationships in his mind. Because of his poor eyesight, he was not allowed to read or write under artificial light, so he often listened to lectures without taking notes and solved problems mentally. This ability later proved fruitful, when his geometrical interpretation of statistics led him to new results.
In 1909, Fisher earned a scholarship to attend Gonville and Caius College in Cambridge, where he concentrated on mathematics and theoretical physics, while also pursuing interests in biometry and genetics. As an undergraduate, he published his first scholarly paper, discussing an absolute criterion for fitting frequency curves. Following his graduation in 1912, he continued his studies for another year, investigating statistical mechanics, quantum theory, and the theory of errors.
During his first six years after college, Fisher searched for an occupation that would suit him, even working briefly as a farm laborer in Canada. Primarily, however, he worked as a statistician for the Mercantile and General Investment Company in London (1913-15) and as a public school teacher (1915-19). Although he was unhappy and apparently ineffective as a teacher, Fisher was nonetheless recognized as a brilliant thinker who had some difficulty explaining his ideas to others. In 1917, Fisher married Ruth Eileen Guinness, the daughter of a doctor. They had eight children before eventually separating.
Even though his jobs did not support research opportunities, Fisher published several notable papers. One of his earliest accomplishments in statistics (published in 1915) was to establish, in mathematical terms, an exact method of sample measurement in statistics. A child of the upper class, he also wrote two papers on eugenics, the science of improving the human race through selective mating. His concern that the--as he thought--less talented lower classes produced offspring at a faster rate than the--in his mind--more capable upper classes influenced his personal choice to have a large family. This was, in addition to being jingoistic, a risk on his part considering his own genetic shortcomings regarding his poor vision, a trait he could have engendered to several large generations of Fishers. His 1918 paper on Gregor Mendel's theory of inherited characteristics laid the foundation for his later work on the statistical analysis of variance.
His growing reputation as a mathematician brought Fisher two promising job offers in 1919. One, from the noted statistician Karl Pearson (with whom he developed a lifelong feud), was to work at the Galton Laboratory in London's University College under Pearson's close supervision. Recognizing a better opportunity to conduct his own research, Fisher accepted a second offer from Sir John Russell at the Rothamsted Experimental Station, about twenty-five miles north of London. Established in 1843, this agricultural research laboratory had accumulated a sixty-six-year backlog of statistical data; it would be Fisher's job to analyze this material. For the next fourteen years, Fisher took advantage of the huge data resources at Rothamsted to derive new analysis techniques as well as agricultural results. On the theoretical side, he formulated the analysis of variance. Now a fundamental tool of statistical analysis, it isolates the effects of several variables in an experiment, showing what contribution each made to the results. Subsequently, Fisher advocated factorial experimentation, in which several factors are varied simultaneously, rather than varying one factor at a time. This approach not only speeds results by gathering information on the effects of several factors, but it also accounts for the possibility that the effect of a factor may be influenced by interaction with other factors.
In another innovation of experimental design, Fisher advocated the random arrangement of samples receiving different treatments. Traditional agricultural experiments arranged samples according to elaborate placement schemes on checkerboard plots to avoid bias from extraneous factors such as variations in soil and exposure to weather. Fisher showed that assigning these positions randomly, rather than according to a systematic pattern, facilitated statistical analysis of the results. His 1925 textbook Statistical Methods for Research Workers is considered a landmark work in this field, although it is so difficult to read that, as Fisher's friend and colleague M. G. Kendall wrote in Studies in the History of Statistics and Probability, "Somebody once said that no student should attempt to read it unless he had read it before."
During the course of his career, Fisher's theoretical work also included improvements to different tests of statistical significance. He refined the Helmert-Pearson chi-square test (including the addition of degrees of freedom) and the t-distribution test, also developing what would eventually be called the F-distribution test after Fisher himself. Fisher introduced the concept of the null hypothesis to designate random processes. Deviations from the null hypothesis indicate significant correlations in statistical samples. Fisher developed procedures for determining when results deviate from the null hypothesis sufficiently to justify an assumption that correlations are significant. He derived the distributions of numerous statistical functions, including partial and multiple correlation coefficients and the regression coefficient in analyses of covariance. Covariance is a term used to describe samples in which statistical results are influenced by different factors. Regression analysis allows the statistician to screen out the effect of all factors other than the one whose significance is being tested. In his 1922 paper "On the Mathematical Foundations of Theoretical Statistics," Fisher analyzed and formalized existing knowledge in the field. Fisher became a Fellow of the Royal Society in 1929. That same year he published a paper on sampling moments that would provide the foundation for future development of that topic. During the 1930s, he wrote several substantial papers on the logic of inductive inference, building on earlier work on the maximum likelihood estimate.
In the agrarian setting of Rothamsted, Fisher also pursued his interest in genetics by breeding various animals such as mice, snails, and poultry, even in his own home. He applied his mathematical prowess to Mendel's work on inheritance, resulting in the 1930 publication of The Genetical Theory of Natural Selection. In it, he showed that Mendelian selection always favors the dominance of beneficial genes and concluded that Mendel's results were mathematically compatible with Charles Darwin's theory of natural selection. His work solidified the growing consensus among theorists of evolution that the Darwinian model, favoring selection over genetic mutation as the explanation for evolutionary change, best fit the available data. Fisher left Rothamsted in 1933 to occupy the Galton Chair of Eugenics at University College, a position he held until 1943. In 1935, he established a blood-typing department in the Galton Laboratory, which developed important information on the inheritance of rhesus blood groups. That same year, he published Design of Experiments, another landmark text in statistical science. The following year, he published his first presentation on discriminate analysis, an approach to statistical samples in which several factors influence outcomes that is now used in such areas as weather forecasting, medical research, and educational testing. During a 1936 summer lectureship at Iowa State College's agricultural research center at Ames (where he had also taught during the summer of 1931), Fisher established contacts that helped popularize his techniques among American educators and psychologists, as well as agriculturists.
In 1943, Fisher joined Cambridge University as Balfour Professor of Genetics. He was knighted in 1952 and served as president of the Royal Society from 1952 until 1954. Both the Royal Society and the Royal Statistical Society awarded him several prestigious medals during his tenure at the University of Cambridge. He formally retired in 1957, but continued working until a successor was found in 1959. In 1950, Fisher published Contributions to Mathematical Statistics, an annotated collection of forty-three of his most significant papers, many of which had originally appeared in rather obscure journals. During the late 1950s, he wrote several articles criticizing the presumption of a cause-and-effect relationship between smoking and cancer based only on the establishment of a correlation between them. When he left Cambridge in 1959, he moved to Adelaide, Australia, to join several of his former students as a statistical researcher for the Commonwealth Scientific and Industrial Research Organization. Fisher died in 1962, at the age of 72.
This is the complete article, containing 1,497 words
(approx. 5 pages at 300 words per page).
