830 



SCIENCE 



[N. S. Vol. XXXIX. No. 1014 



is better as an introductory work for most 

 American students, while Brown's more tecli- 

 nical and more comprehensive book is indis- 

 pensable to those who care to do really serious 

 research. In this connection the work of 

 Urban'^'' may also be mentioned. 



The most comprehensive text on modern 

 statistics — an introduction to statistical theory 

 rather than a guide to the application of 

 statistical analysis in any particular field — is 

 that of Tule,^^ who has brought to his task 

 the training of an assistant in Karl Pearson's 

 laboratory. To all those who must use statis- 

 tical formulae without being able to read with 

 ease the fundamental papers — and this com- 

 prises all but a handful of the workers — ^Yule's 

 book is indispensable. The simpler statistical 

 processes are all treated with care and clear- 

 ness, and with a terseness refreshing after the 

 verbiage of works written from the biological 

 vievsrpoint. One must commend the careful 

 system of cross reference, which should greatly 

 facilitate the use of the book, and the very com- 

 plete and annotated bibliography which should 

 inspire the student to stUdy original sources. 

 The student, moreover, should remember that 

 even Mr. Yule's book does not render this un- 

 necessary, for in the text some subjects of 

 great importance are not treated at all. For 

 example, one misses a discussion of Pearson's 

 various tsTJes of frequency curves which have 

 been used with such success for the past sev- 

 eral years in describing a great variety of phe- 

 nomena. The classical fourfold correlation 

 method is also omitted. The most flagrant 

 fault of the book is the putting forward of 

 certain measures of association and correla- 

 tion, which are worthless,^^ but which, because 

 of their simplicity, are likely to be widely used 



urements, " Cambridge University Press, 1911. 



10 Urban, F. M., "The Application of Statis- 

 tical Methods to the Problems of Psycho-Physics, ' ' 

 Philadelphia, 1908. 



11 Yule, G. U., "An Introduction to the Theory 

 of Statistics," London, Chas. Griffen and Co., 

 1911, 2d edition later. 



12 See Heron, BiometriTca, Vol. 8, pp. 109-122, 

 1911; K. Pearson and D. Heron, Biometrika, 

 Vol. 9, pp. 159-315, 1913. 



by those who prefer ease of calculation to cor- 

 rectness of result. 



The guides which are available the beginner 

 must use : but he should know that there is not 

 only no royal road, but as yet no Baedeker, to 

 statistical analysis. Here, as is wont to be the 

 case with text-books, the guides have for the 

 most part been written by men who have not 

 distinguished themselves by exploration into 

 the territory through which they now propose 

 to lead others. The real student will, there- 

 fore, be vigilantly skeptical, and wiU go as far 

 as he is able to the original sources to read 

 and weigh for himself. 



Taking the books mentioned in this review 

 as a class, the greatest criticism that can be 

 made is not that there are blunders in state- 

 ment and misprints in formulae, but that 

 throughout there is an incautious attitude 

 towards the real difficulties which the student 

 is to encounter, or even a positive assurance 

 that dangers are only apparent. Nothing could 

 be farther from the truth, for in higher sta- 

 tistics innumerable pitfalls surround the in- 

 vestigator. Serious harm has been done by 

 telling the beginner that he need not under- 

 stand the formulae to use them. Neither is it 

 necessary for a chemist to understand his 

 reagents ! To be sure, most of us have of 

 necessity to work with but little knowledge of 

 the mathematics which lie back of the for- 

 mulse, but it is the obvious duty of the student 

 who proposes to use these newer tools of re- 

 search to learn all that he can concerning 

 the assumptions upon which the formulae 

 rest in order that he may apply them with 

 intelligence. 



Finally, the beginner must realize that it 

 is as impossible to gain a working knowledge 

 of statistical methods from a text-book with- 

 out experience in the routine of measurement, 

 computation and interpretation as to acquire 

 proficiency as an organic chemist outside the 

 laboratory. It takes years of hard work to 

 make a proficient biometrician — although 

 some have to their own satisfaction qualified 

 for the writing of texts much more easily. 



J. Arthur Harris 



