March 5, 1962
BOOKREVIEWS
few academic scientists appear t o be seriously trying to teach their students how to use these methods.
883
This reviewer feels that authors of books on probability and statistics for scientists should not devote much time to the mathematical machinery of probability and statistics because there are already many books which do this exceedingly well. What the scientist needs is help in framing his problems and designing his experiments so t h a t the machinery will be able to help him interpret the results. KODAK PARK\VORKS EASTMAN KODAKCOMPASY GRANTWERNIMONT ROCHESTER 4, K E W YORK
Professor Parratt’s book was written, “to introduce the undergraduate student to the unifying concepts of probability a n d statistics as they apply to science” (pp. VU, viii). He devotes chapter one to a historical survey of ideas about probability and to a n explanation of classical ( a priori) a n d experimental (a posteriori) probability. I n the first part of chapter two, he points out the close relationship between scientific reasoning and the correct interpretation of measurements, and gives some basic definitions of errors, significant figures and the rounding of numbers. The rest of the chapter deals with frequency distributions, Principles and Applications of Paper Electrophoresis. By location indices and precision indices. CH ~J-UNDERLY, Member of Scientific Staff, Medical The statistics of measurements, in functional relationUniversity Clinic, Zurich (Switzerland). D. Van Sosships, is presented in chapter three. Equations that govern trand Co., Inc., 120 Alexander Street, Princeton, S e w the propagation of errors are given fOr sums and differences Jersey. 1961. xii 263 pp. 13 X 19 cm. Price, as well as for products and quotients. Use of the t-test $5.00. and the F-test t o examine the consistency of means and standard deviations is explained. The .method of least Since the publication on the quantitative analysis of squares for fitting curves to experimental data is developed. proteins by Consden, Gordon and Martin (Biochem. J . , The chapter closes with a short discussion of t h e design 38, 224 (1944)), paper chromatography has become a of experiments. powerful tool for the study of components in biological I n chapter four, the normal probability distribution is systems. It is interesting to note, however, t h a t the derived by way of the binomial distribution; and its mathedevelopment and use of paper electrophoresis dates from the matical properties are discussed in great detail. The x* isolation of a yellow protein from snake venom by Konig for goodness of fit is explained. and Klobusitskp (iVaunyn-Schmiedebergs Arch. exptl. T h e final chapter is devoted to the Poisson probability Pathol. Pharmakol., 192, 271 (1939)) which took place five distribution; and its application to measurements of radio- years earlier. active decay is explained in detail. This small book is a lucid and concise review of judiciously Many problems are given at the end of every chapter and selected methods and techniques in paper electrpphoresis. a glossary summarizes the equations developed throughout As might be expected, the predominant amount of space the entire book. is devoted to proteins and their derivatives, but a n excellent This book does not appear, to this reviewer, to be worth theoretical discussion and description of methods extends the price to the chemist, student or professional. Too much its usefulness to a much broader group of investigators. time is devoted t o mathematical probability and statistics; An extensive bibliography completes the necessarily scant and too little time is used to explain the logic of experimental details in a volume of this size. design which experimenters learn, with difficulty, by themPublications in rapidly developing areas suffer from unselves. avoidable delays between preparation and publication. The most disturbing thing about the book, to this re- This book is a completely revised edition in English of a viewer, is the way the author has jumbled inadequate tra- well-received three-year old German text by the same ditional machinery With the more useful small-sample author. Thus, one would have hoped to find in a work methodology. The equation for the normal probability of this high quality, a more complete discussion of recently distribution is derived with parameters h and z rather than developed low temperature techniques. This limitation u and p . T h e mean deviation is defended because i t is aside, the book is a valuable addition to the growing series simple to compute and puts less weight on the larger devia- of Elsevier Monographs. tions. Probable errors are used extensively tkroughout the OF CHEMISTRY book and the author justifies this because most experi- DEPARTMEST CHEMISTRY GROUP ELIE A. SHNEOUR menters have a sort of intuitive judgment about the reli- BIO-ORGAMC LAWRESCERADIATIOS LABORATORY ability of measurements as a consequence of their laboratory USIVERSITYOF CALIFORMA training” and because “the experimenter believes t h a t 1,CALIFORNIA systematic errors are never quite eliminated, a situation BERKELEY t h a t requires some guessing anyway” (see p . 148). The most inadequate machinery in the book is ChauNouveau Trait6 de Chimie MinBrale. Tome XIII. venet’s criterion for rejecting “bad data.” This procedure, Premier Fascicule. OxygBne, Ozone, Oxydes, Eau oxyfirst suggested by an astronomer in 1891, has been cited and recited in books on least squares and the theory of measure- ggnbe, La Combustion, Soufre(1). Edited by PAULPASCAL, ments. The rejection of data should not be justified on Membre de I’Institut, Professeur honoraire A la Sorbonne. Masson e t Cie, 120 Boulevard Saint-Germain, Paris 6, statistical arguments. The author sometimes presents misleading, if not erro- France. 1960. xxxix f 1124 pp, l i . 5 X 26 cm. iicous, explanations of statistical methodology. The F-test Price, brochi. 310 K F . ; cartonni. toile, 330 S F (two parts). comparing variances is set up on pp. 124-126 with the This is the first part of Volume XI11 which deals with larger variance always in the numerator (two-tail test); elements of Group V I , namely 0, S, S e , T e and Po. As in b u t the 5y0 limits of F in Table 3-2 are for a one-tail test most of the other volumes of this monumental treatise and would really be 10% for the situation described. published so far the introductory chapter (60 pages) is Again, in section 3-7, the concepts of correlation and written by the editor himself. It makes for interesting regression are badly mixed up and Figures 3-1 shows a scatter reading from the historical and systematic viewpoints, and diagram of physics and mathematics grades with “probablethe bibliography (some 300 entries) gives an idea of its scope error lines parallel to a regression line, rather than the ap- on both counts. Very few misprints and only one or two propriate 50% ellipse about the mean values of the two sets erroneous statements were found in this chapter. (On of grades.” p . 5 , the explanation for the different valence angles in the The book also fails to present some of the most useful series of compounds F20, ‘ 2 1 2 0 , etc., is not consistent with statistical techniques available to the experimenter. KO the accepted views; on p . 53, the configuration of the H& mention is made of control charts anywhere in the book; molecule is known to be neither cis nor trans). and the procedure of putting confidence limits about the The chapter on elementary oxygen by A . Pacault covers small-sample estimates of a mean, standard deviation, or the subject adequately in some 180 pages with 300 references. regression coefficient, and the interpretation of such con- There follows a 70-page chapter on ozone by Josien and fidence intervals is never discussed. Sourisseau, which may have been completed somewhat before the others in this volume since the bibliography ends Although the author shows how to use the F-distribution to test for the consistency of variances, he makes no men- earlier (1954 instead of 1958-1959). Thus the latest work tion of the fact that this distribution is much more powerful on the thermal decomposition of gaseous ozone was missed. At this point a very useful and well organized section of when used to make significance tests of results from properly designed factorial experiments. some 150 pages on the crystal structure of oxides and related
+
