Maxwell J. Roberts and Riccardo Russo.
A Student’s Guide to Analysis of Variance.
London and New York: Routledge, 1999, 265 pp. $98.00 cloth (014-15165644), $31.99 paper (014-15165652).
For most students majoring in Sociology, the analysis of variance is not encountered in very much detail in their statistics courses. It may instead be a stepping stone, lightly traversed, toward regression analysis. Hubert Blalock’s classic undergraduate text Social Statistics, for example, treated analysis of variance that way. For instructors who wish instead to centre their course in statistics around analysis of variance, this is the book!
A Student’s Guide to Analysis of Variance is written by two lecturers in psychology at the University of Essex in England. The preface explains that they were seeking a middle ground between the very fleeting treatment of analysis of variance such as noted above and graduate student level coverage heavily leaden with formulae and proofs. “The main intended audience is undergraduate psychology students who have just completed a course in introductory statistics,” the authors advise.
It was interesting that in today’s software-conscious teaching environment, the authors opted not to fuse their book with a particular statistics package. In part they were protecting the shelf-life of their book and maximizing their market, but more pedagogically they note a concern that “exercises can substitute a knowledge of procedures for the understanding of fundamental principles, so that the outcome is highly proficient computer operators rather than competent statisticians.”
The presentation begins with two review chapters; material for “revision” after the “long vacation” as they say in the U.K. Thus Chapter 1 covers experimental design, random error, the concept of null hypothesis, Type I and II errors, significance levels and the notion of parametric tests. Chapter 2 proceeds to measures of central tendency and dispersion, variance, the t-test, and issues of statistical power. The pace is rapid here, it being presupposed that the student had a first exposure to such matters before that long vacation.
The analysis of variance proper begins with Chapter 3, presenting a good intuitive discussion on sources of variance in data. Chapter 4 covers F ratios for one-factor between-subjects designs. The concept of degrees of freedom, sometimes hard to make clear to students, is well-presented. The chapters move up through the more complex issues and designs: “rogue data” (I liked that term) in Chapter 5; more complex contrast comparisons in 6; within-subjects designs in 7; factorial design and interactions in 8; two-factor between-subjects designs in 9; and so on through to Chapter 12. Chapter 13 concludes with practical research advice to students using ANOVA in projects. The tone here is frank, supportive and feisty. E.g.: “If you do not like the look of three-way interactions, then do not use a three-factor design, and don’t be persuaded to use one by an enthusiastic project supervisor.”! (p. 228). Finally, an Appendix provides tips on writing up the results of analysis of variance.
This is an attractive and clearly-written book with a certain amount of “soul,” or conviction, even though it is a statistics textbook. Without becoming overly cute, modern-day graphics are well deployed to illustrate points and draw attention to salient features of data. Most likely, given the priorities for statistical analysis in our discipline, it will only find its way into the occasional sociology course, one probably at the graduate level by an instructor especially fond of ANOVA. As the authors stated at the outset, it’s mainly for psychology majors.
John Goyder
Department of Sociology
University of Waterloo
jgoyder@watarts.uwaterloo.ca
March-April 2000
© CJS Online
