Content area
Abstract
Today randomization is used routinely in many experimental situations; the randomized trial has become the gold standard in many disciplines. But until the 1920's the preferred method was one of systematic design: the researcher arranged the experimental treatments in a way that in her judgment would give the most reliable results. Sir Ronald Aylmer Fisher (1890–1962) changed the methods of experimentation and statistics and his principles of experimental design are now considered basic to much of scientific experiment. While randomization occurred in experimental work prior to Fisher's, most notably in psychology of the late nineteenth and early twentieth centuries, Fisher was responsible for a major change in the way randomization was, and is, used. He advocated that randomization was necessary even in experiments that could be done otherwise; he argued against using a non-random plan chosen by the experimenter. Fisher's reasons were that randomization would eliminate bias in the experiment and would also enable a valid test of significance. I argue that what led Fisher to the requirement of randomization in experimentation was his several years of working on the statistics of small samples and a realization of the role there that randomness plays. He extended the randomness that is inherent in sampling and made randomization a necessary part of experimentation. He developed randomized experimental design, and also analysis of variance, at Rothamsted Agricultural Station in Harpenden, England, between 1919 and 1933. Fisher was responsible for the wide acceptance of this new usage, of randomization, through his publication of many journal articles, his work with statisticians of many nationalities, and two of his most influential books, Statistical Methods for Research Workers (1925a) and The Design of Experiments (1935). Randomized experimental design is now used world wide, in many research disciplines, including agriculture, medicine and psychology, among others.
