Book Review
The Really Understandable Stats Book
Neil Burdess
ISBN: 978-0-72481048X; 371 pages; Prentice-Hall;
Richard Jarrett
Department of Statistics, University of Adelaide, SA
Statistics is increasingly important in nursing as more nursing research is done. I have used two criteria to review this book-how well has the author communicated the message and is the material technically correct?
The author's stated aim here is to present statistical ideas in English, and he does well in the early chapters. He starts with three questions: What group of people are you studying? How and what will you measure on them? Why are you studying them? Early chapters cover different types of data, ratios, rates, index numbers, histograms, boxplots and scatterplots. Each chapter ends with some review questions. At the back of the book are answers, together with a glossary of formulae and technical terms.
The book is similar in style to Huff's How to Lie with Statistics (a great book if you don't take the title too seriously) but goes deeper into various statistical methods.
Unfortunately, I was disappointed by the later more technical chapters and I would not recommend it on its own for the following reasons. It presents statistics as being about the manipulation, assessment and presentation of data. This is not enough-a large part of statistics is about the front end-what is the purpose of our study and how can we plan our study to make sure we answer the major questions of interest? If we don't get this right, all else is a waste of time. Two examples illustrate this-in chapter 7 (p 217), Burdess tests whether a touch-typing course has done any good by comparing the grades of 200 students on a test done before the course and a (different) test done after the course. What about the fact that they are different tests? This is just bad design. Later (p 245), he takes 15 students, orders them alphabetically, gives the first five treatment A, the next five B and last five C. Why do this when it is so easy (and important) to do the allocation at random?
The author insists on calculating standard deviations using the sample size 'n' instead of '(n-1)' on the bottom line. This leads to all sorts of strange formulae. He is also incorrect in defining the five per cent for the usual significance test as 'the probability that the sample comes from the population specified by the null hypothesis' (p 185), rather than the correct 'given the null hypothesis is true, the probability of getting a value as large or larger than the sample value'. After all, any particular sample either does or does not come from the null hypothesis population!
Hence, after a promising start, the book ultimately disappoints due to flaws in the technical material. There are too many cases of bad design and incorrect or inconsistent explanations for me to be able to recommend this book.
eContent Home
