British Medical Journal : Review of the 1^st edition of R. A. Fisher’s Statistical Methods for Research Workers

John Aldrich, University of Southampton, UK. (home) 2006. Latest changes April 2017.

Introduction

R. A. Fisher’s Statistical Methods for Research Workers (1925) was probably the most influential book on statistics of the 20^th century. Yet “The book did not receive a single good review” writes Fisher’s biographer, Joan Fisher Box (1978, p. 130).

Box (p. 482) lists 5 reviews, signed ones by “Student” (W. S. Gosset), Leon Isserlis, Egon Pearson and unsigned ones, in the British Medical Journal and Nature. She missed the very enthusiastic review from the American statistician Harold Hotelling.

All the reviews are available on the web, the one reproduced here and

• Student’s review in the Eugenics Review.
• Leon Isserlis’s review in the Journal of the Royal Statistical Society.
• E. S. Pearson’s reviews in the journal Science Progress.
• Harold Hotelling’s review in JASA.
• Nature review

They are all worth reading for each gives a different perspective on the book. 

The BMJ review

Apart from Hotelling, all the reviewers had reservations about the book—it was too difficult and contained no proofs—but, as Frank Yates (1990, p. xii) says, only the BMJ review was “really hostile.” Yates quotes from the second paragraph of the review

The trained statistician interested in Mr. Fisher’s researches will miss a detailed justification of his conclusions … Even if the statement that Professor Pearson’s treatment of a fundamental problem contained a “serious error” had not been disputable, and therefore improper in a work addressed to elementary students, it would have reminded anyone of Macaulay’s remark on a similar situations—“just so we have heard a baby, mounted on the shoulders of its father, cry out, “how much taller I am than Papa!”

The reviewer was responding to Fisher’s statement that Karl Pearson’s paper of 1900 “contained a serious error, which vitiated most of the tests of goodness of fit made by this method” [I §4 p. 17] and to Fisher’s way of presenting his version of how the tests should be done. Yates remarks of Fisher’s presentation, “Actually the point was discussed by Fisher (Example 8) [IV. §20 p. 81], which for clarity of statement and convincingness of argument would be difficult to better.” Yates came on the scene when Fisher’s arguments had had time to sink in and I think he underestimated the difficulty for the first readers. It was not family partiality that led Karl’s son, Egon Pearson, to write in his review, “it is necessary to read the book in conjunction with the author's papers published elsewhere, and one must confess to some difficulty in following several of the proofs based on the idea of degrees of freedom.”

The author?

The review was unsigned. The quotation from Macaulay may be a clue to the author for this tart variation on the “shoulders of giants” theme was no commonplace and another put-down from the same source appeared in the statistical literature the following year. Greenwood and Isserlis borrowed Macaulay’s rebuke of the editor of Mackintosh's work: he “finds one mistake—such a mistake as the greatest scholar might commit when in haste, and as the veriest schoolboy would detect at leisure. He gloats over this precious discovery with all the exultation of a pedagogue.” I think it likely that the BMJ review was the work of one of the co-authors, the medical statistician Major Greenwood.

The reviewer’s assessment was pure Greenwood. In 1933 Fisher succeeded Pearson in the Galton chair. Greenwood wrote Fisher a friendly letter of congratulations yet still reminded him of his dependence on Pearson. (Letter of 10 June 1933 reproduced in Bennett (1990: 318)

There is, in my opinion, no other man alive worthy to sit in old K. P.’s chair; his mantle has descended on our shoulders and a double portion of his spirit is yours. Like all of us, you owe him much. If K. P. had never lived, you would have assuredly been one of our foremost men of science but very likely your field of work would not have been statistics, but, perhaps, pure mathematics. You will repay the debt not by silly adulation of the “illustrious predecessor”—as so often happens when a great personality leaves the stage—but by carrying the work further. This appointment really makes me happy for—although K. P. has often enough wounded my feelings and insulted my friends–I still love the man and venerate his genius, I should have been sad if his kingdom had fallen into the hands of a second rater.

Greenwood recalled how much Pearson meant to him—mistakes and all—in the opening paragraphs of his Presidential Address to the Statistical Society given a few months after Pearson’s death.

References

• The first edition of Statistical Methods for Research Workers is available on Christopher Green’s Classics in the History of Psychology website.
• Joan Fisher Box (1978) R. A. Fisher: The Life of a Scientist, New York: Wiley.

· Frank Yates’s foreword to Statistical Methods, Experimental Design and Scientific Inference, Oxford: Oxford University Press, 1990. This volume contains the 14^th edition of Statistical Methods for Research Workers and the final editions of Fisher’s The Design of Experiments and Statistical Methods and Scientific Inference.

· T. B. Macaulay’s review of Mackintosh’s History of the Revolution in England in 1688 is in the Edinburgh Review, July 1835, pp. 265-322. When Macaulay reprinted the essay he removed the ripest abuse though the shoulders of Papa stayed.

• M. Greenwood and L. Isserlis (1927) “A Historical Note on the Problem of Small Samples,” Journal of the Royal Statistical Society, 90, 347-352. (JSTOR)
• H. J. Bennett (1990) (ed.) Statistical Inference and Analysis: Selected Correspondence of R. A. Fisher, Oxford, University Press.
• M. Greenwood (1936) English Death-Rates, Past, Present and Future: A Valedictory Address, Journal of the Royal Statistical Society, 99, (4), 674-707.

Perspectives

· For a perspective on Fisher’s book see A. W. F. Edwards, (2005) “R. A. Fisher, Statistical Methods for Research Workers, 1925” in I. Grattan-Guinness (ed.) Landmark Writings in Western Mathematics : Case Studies, 1640-1940, Amsterdam: Elsevier.

• For Fisher generally, see A Guide to R. A. Fisher.
• For Karl Pearson, see Karl Pearson: A Reader's Guide.
• For many of the important figures see Figures from the History of Probability and Statistics (500 KB). This also has references on the History of Statistics.

________________________________________________________________________

Anon. (1926) Review of Statistical Methods for Research Workers (R. A. Fisher), British Medical Journal, 1, 578-9.

Mr. R. A. FISHER, the author of Statistical Methods for Research Workers, is known to the learned world as an able and original mathematical statistician, while as chief statistician to the Rothamsted Experimental Station, he has had considerable experience of the application of quantitative methods to a particular class of biological problems. The aim of his present volume is to supply a manual for biological research workers desiring to evaluate statistically their experiments. Since, in the kind of biological research with which Mr. Fisher has had to deal practically, small samples only are usually available, he has given considerably more attention to the particular methods applicable to small samples than authors of many textbooks have deemed necessary; he is well qualified to do so because of his previous mathematical investigations in that difficult branch of statistical theory. The chief merit, indeed, of his volume is that it provides numerous worked examples and tabular aids to calculation, which should be of service to those having to rely on small samples.

Mr. Fisher has designed his book for readers without special mathematical training, and is conscious that the inclusion of a good deal of matter which is “advanced,” and has, indeed, not been published before, needs some explanation. If he feared that he was likely to fall between two stools—to produce a book neither full enough to satisfy those interested in statistical algebra nor sufficiently simple to please those who dislike algebra—we think that Mr. Fisher’s fears are justified by the result. The laboratory worker will, as we have said, find the book useful when dealing with small samples, but will not find in it a sufficiently simple and comprehensive account of the general principles of statistical methodology. The illustrations are sometimes—for example, that dealt with on pages 94 et seq. [IV. §22]—only illuminating to students with special knowledge of modern genetics. The trained statistician interested in Mr. Fisher’s researches will miss a detailed justification of his conclusions, and may resent the somewhat arrogant way in which the law is laid down upon points respecting which there is difference of opinion among persons possibly as well informed as Mr. Fisher. A conspicuous example of this latter failing is the reference to Professor Karl Pearson on page 17 [I. §4]. Even if the statement that Professor Pearson’s treatment of a fundamental problem contained a “serious error” had not been disputable, and therefore improper in a work addressed to elementary students, it would have reminded anyone of Macaulay’s remark on a similar situations—“just so we have heard a baby, mounted on the shoulders of its father, cry out, “how much taller I am than Papa!”