Nature : Review of the 1st edition of R. A. Fisher’s Statistical Methods for Research Workers

 

R. A. Fisher’s Statistical Methods for Research Workers (1925) was probably the most influential statistics book of the 20th century. It presented in book form for the first time Fisher’s work on maximum likelihood, t-tests (including applications to regression), the z-transformation of the correlation coefficient, the analysis of variance, randomisation and blocking in the design of experiments, etc. The first edition is available on Christopher Green’s Classics in the History of Psychology website

 

There were six major reviews of the first edition. One appears below and I the other five are available on the web:

 

All the reviews are worth consulting for each gives a different perspective on the book. See these other sites for further literature and links.

 

The opinion of Nature mattered more perhaps than that of the other journals because the “research worker” to whom Fisher’s book is addressed was most likely to read it. Like the other reviewers, this one perceived the book as a specialised work on “small samples.” Although the first small sample publication—Student’s 1908 paper—was nearly 20 years old, small sample work had as yet made little impact. All the reviewers, like this one, thought the book was too difficult, especially its first chapter.

 

This reviewer was the only one to voice the suspicion that Fisher’s arguments rested on “inverse probability,” i.e. that they were fundamentally Bayesian. For a good many years Fisher had to struggle to distinguish his “likelihood” from the Bayesian “posterior” as it is now called. There was a strong Bayesian vein in the textbooks of the time, Bowley’s Elements of Statistics and Yule’s Introduction to the Theory of Statistics.

 

The review was unsigned which was not unusual. My guess is that it was written by the psychologist Godfrey Thomson. He had the knowledge and range of interests required to write this review which few did, although Yule is another possibility. There is a recent account of Thomson’s statistical career by Bartholomew, Deary and Lawn (2009).

 

John Aldrich, University of Southampton, Southampton, UK. (home) August 2005/October 2010.

 

________________________________________________________________________

 

Unsigned review in Nature, 116, (1925) 815.

Joint Review of

·        The Fundamentals of Statistics. By Prof. L. L. Thurstone

·        Statistical Methods for Research Workers. By Mr. R. A. Fisher.

 

Modern statistical methods are now used in such widely different spheres of activity that it is natural that several books on the subject should be produced to meet the needs of the various persons concerned. It is of interest to notice that these books, being of the textbook variety, usually assume an air of certainty with regard to some things which are still almost within the regions of controversy. This becomes the more obvious as the subject matter becomes more advanced.

(1) Turning to the two books before us, we find that Prof. Thurstone has set himself the task of providing an elementary book on statistics for students of psychology who have little mathematical knowledge. The book is about as elementary as it can be, and it assumes that the reader is so poorly equipped as to need to have the graphical expression of a straight line and the most elementary aspect of the binomial series explained. It will, however enable these non-mathematical readers to follow, in a reasonable way, results obtained by others and expressed in terms that would be meaningless without some help such as this book gives. A good many elementary books of this kind have been published in recent years in the United States and this is one of the best of them.

(2) Mr. Fisher's book is written for a more advanced type of reader, and it has much to commend it. It treats of the interesting and important subject of small samples in statistical work; it has originality; its author is full of ideas; and its appearance is all that can be desired. But unfortunately the book suffers from an introductory chapter which seems unnecessarily hard to follow, and from the difficulty of the subject, which has, we fear often prevented Mr. Fisher from writing down to his reader. The book is intended for biological research workers, and it is apparently assumed that they already know sufficient of the theory to accept, without proof, the methods given, or that they will adopt these methods on Mr. Fisher's authority. A statistical “research worker” may be willing to dispense with rigid mathematical proofs when it can be seen from several arithmetical examples that a method carries its own justification, but in the present work the absence of proof goes rather far, and we fear that readers with little knowledge of the most recent statistical work will find the book as a whole difficult to follow, while those unfamiliar with the terms used in biological research work will have trouble with some of the examples.

In many places throughout the book a reader may hesitate, wish perhaps that he could share Mr. Fisher's confident assurance, and then find himself wondering whether deep down under much of the theory about which Mr. Fisher is so sure, there may not lurk the assumption that we can approximate to the whole population from a sample in a way that resembles the theory of “inverse probability” which he “wholly rejects.” [I. §2, p. 10].

It seems to us probable that the book will be read as much by statisticians who wish to study Mr. Fisher's methods and views as by those research workers who merely want to apply the methods he describes. Such readers will find so much that is interesting, suggestive, and useful that they will forgive the weaknesses we have tried to indicate.  

 

Note