Non-standard errors add up to lack-of-confidence intervals

Biomeasurement is suitable for undergraduate students of biological sciences and related disciplines taking an introductory statistics module. Dawn Hawkins covers the classical material one might expect. The book assumes a basic level of mathematics - consistent with the trend in UK universities to move away from more maths-based introductory modules.

The book is well structured, with aims at the beginning of each chapter and summaries at the end. Margin notes and boxes help to summarise important points. Worked examples are presented in a consistent and structured manner, using hand calculations and/or SPSS. Of particular use are screen shots of SPSS, allowing readers to analyse their own data easily. However, the choice of SPSS for the biosciences is questionable.

Hawkins uses real data and gives detailed descriptions. Statistical material from scientific articles is discussed. A companion website provides datasets, articles and other resources.

The book begins with a motivational chapter discussing scientific research.

The next introduces definitions, terminology and basic instructions in the use of mathematics and algebraic logic - useful for more mathematically challenged students.

Data analysis begins in chapter three, which examines a single sample of data, then moves on to standard errors and confidence intervals. This is followed by an overview of hypothesis testing and chapters devoted mainly to classical parametric and non-parametric hypothesis tests. This includes one-way Anova, simple linear regression and correlation. The book concludes with advice on selecting graphs and hypothesis tests.

The technical detail can be superficial - reasonable given the audience - but several points are worth noting. A few common graphs are missing, and boxplots could be more prominent. There is insufficient treatment of confidence intervals, and some worked examples end abruptly and could use more interpretation.

There are a number of errors and ambiguities. The statistical tables for the Mann-Whitney U test, as used, are incorrect, and critical values for the F-distribution are wrong. There are also potentially confusing typographical errors, so I would hesitate before recommending this book.

James Gallagher is senior statistician, Reading University.