Some Aspects of Experimental Inference
Oscar Kempthorne · Journal of the American Statistical Association · 1966
L AM very happy that the Fisher Memorial Lecture* series was started by the joint statistical societies. This may, perhaps, prevent such happenings as the following. Last week I was saddened very much when I was trying to finid out what a student transferring into Ames knew about statistics. I asked him what name he associated with a lower bound on variance, what name with information aind what name with likelihood. He had had 3 or 4 years of graduate courses in probability and statistics, but had no answer. I then asked him if he had heard the name Fisher. The answer was Yes,-he was an applied statistician. Also it was clear that the fault did not lie with the student. The English poet Tennyson said Knowledge comes, but wisdom lingers. I think it is more appropriate in our society to say that knowledge is purchasable by dollars, but for wisdom we have to depend on the genetic process which Fisher understood so well. R. R. Race said, Wherever the sun of genius may have illuminated the British blood group work, it has been of Fisher's shining. How much more true is this of statistical work! I would also like to say that you do me great honor in asking me to give the lecture and I thank you. In my case you do honor to Fisher indirectly because almost all of the statistics I know derives from Fisher. I hope I do credit to both Fisher and you. In my lecture I make numerous references to Fisher's writings. The referencing code is: S, x means page x of Statistical Methods and Scientific Inference; D, x means page x of The Design of Experiments, Second Edition. The contributions of Fisher to the whole field of statistics have been extensively reviewed in many places, since his death. It is tribute to his genius that these review articles overlap very little. His contributions pervade the whole of statistical thinking with the exception of decision making, the only contribution to which in my opinion is his work on the discriminant function. I surmise that his writings will be subjected to re-examination and re-evaluation for many decades and centuries, just as is the case for the writings of Francis Bacon, David Hume and Thomas Bayes. Much of his work was purely mathematical and what was correct in this will be included in the corpus of basic mathematical statistics, no doubt with refinements and tightening of mathematical argument, but with little discussion. What will however lead to discussion is Fisher's writings on the philosophical nature of statistical procedures. Fisher's works are of so much interest because many of them were directed to the problem of inductive inference, a matter which has plagued philosophers for hundreds or even thousands of years. It seems to me from reading of the