His employment as a statistician at the Rothamsted Agricultural Research Station led him to work on challenges in analysing and interpreting data from crop trials. He proposed and named a statistic, the variance, as a way of studying variation due to various influences or factors, devised the theory for partitioning such variation into identifiable components, and devised a powerful and insightful approach, the analysis of variance, to performing the calculations.
He realised that how experiments were designed, in particular the way that known potentially biasing factors were controlled and the degree of replication, would have a great effect on the precision with which statistical estimates could be produced, and that efficient experiments might involve clever simultaneous study of many factors. He founded experimental design as a subject in its own right. He emphasized that the way in which the statistician intended to analyse the experiment should guide its design. To deal with the possible influence of hidden factors in a way that would avoid unconscious bias, promote trust, and permit valid calculation of estimates and associated variances, he championed randomisation His general theory of experimental design proved itself applicable to other disciplines apart from agriculture, and is now applied with great success to subjects as different as manufacturing and clinical medicine.
Many of the statistical tests now in common use were developed by Fisher, and the associated probability distributions were derived by him. He produced a general theory of estimation based on the concept of likelihood, now central to many data-analytic disciplines, not just statistics. He developed efficiency and sufficiency as a way of judging the extent to which a statistical estimate was satisfactory.
He had exceptional geometrical insight and ability to visualise multidimensional spaces. An early paper of his used this to derive the sampling distribution of the correlation coefficient, a matter that had defeated Karl Pearson and his colleagues. He worked on problems of estimation associated with simultaneously varying observations, thus founding the field of multivariate analysis, and provided the necessary statistical techniques for studying palaeomagnetism. His pupil, C. Radhakrishna Rao, described him as 'The Founder of Modern Statistics'.
Fisher also pioneered mathematical models of how both natural selection and random fluctuations (due to the sampling effects of finite population size) affect the frequencies of genetic variants in populations, introducing mathematical methods that are still in use today. This work was summed up in his 1930 book The Genetical Theory of Natural Selection , which is full of brilliant insights, such as why the sex ratio tends to be one to one, into the evolutionary process, and is widely regarded as the most original book on evolution since Darwin's Origin of Species.
Fisher later turned his attention to human genetics, developing statistical methods for analysing linkage between human genes, which have been central to the identification of genetic variants underlying disease traits. He led the development of blood group research at University College London, which pioneered the population genetics of human blood groups. An outcome of this work was Fisher's masterly analysis of the genetics of the Rhesus blood group system. Incompatibility between the Rhesus make-up of the mother and foetus is the cause of haemolytic disease of the new-born, and understanding the genetics of Rhesus has enabled this disease to be eliminated.
Smoking and Eugenics are dealt with in greater detail in further statements.
This page last updated 3 March 2021