A good case can be made for using statistics to help decision making and reporting in the ILS sector. Having firm evidence, regular reliable information, realistic comparisons, plausible projections are all important, and all these activities benefit from having enumerative data. Egghe (of Limburg University in Belgium and author of an Introduction to Informetrics, Elsevier, 1990) and Rousseau (of the Katholieke Hogeschool Brugge‐Oostende in Belgium) make such a case for statistics – for helping managers run their services better and explain them to others better. They have written a short readable book with a good subject index. It has three parts – information about the library (reports, data, sampling), descriptive statistics (graphical representation, regression, central tendency and variance), and inferential statistics (samples, confidence intervals, sampling again, and correlation).
Good tests of a useful statistics book for practitioners are: are the explanations clear and relevant; can I really use it at work (rather than just finding it, intellectually, an interesting read); and will people understand it if I do? Egghe and Rousseau clearly decided at the start to provide numerous examples, along with their solutions: most of these are credible (collections, overdues, users, suppliers and delivery, journal titles, sample sizes, budgets), and many bring together key factors (what we know and what we expect, how confident we can expect to be about the mean of a sample or the average delivery time in days). Some are artificial (like height of shelves/number of misplaced books, comparing two samples when you would not buy two simply to analyse the results, yellow books on mathematics which are also thin books) and we all recognise such examples in textbooks which know the statistics but struggle to find suitable factors to illustrate them. Some examples, too, were provided by H. Voorbij of the Royal Library in The Hague, courteously acknowledged each time they appear.
The credibility of the examples decides how useable they really are. The descriptive statistics here explain graphical representation (such as scatterplots) well enough, deal briskly with means and standard deviations and percentiles, and (from Rousseau’s research) provide an interesting discussion of the Lorenz curve based on CD loans in Belgian public libraries. They are good on simple regression too. The inferential statistics show a determination to avoid the usual approach which is usually significance‐test driven (like chi‐square, t and F). Unavoidably perhaps, they start with the normal distribution and move (using the central limit theorem) to Z tests and tails, using Z as a way of explaining and exploring significance, with one and more samples, after that. In a conversational and incremental way, they talk the reader through a growingly complex reasoning, providing examples (and solutions) along the way.
They apply it to reasonable situations – is there a significant difference between the visits of all users and just female users, are network loans significantly higher for corporate libraries than for other members of the network, is there a significant difference between two library suppliers, what size of sample satisfies our inquiry at the 95 per cent significance level? Correlation and its significance end the third section in a quaintly called “epilogue”. Throughout formulae are used (one of the appendices gives a notational guide, but no significance tables are provided). The decision, in the inferential section, to sidestep the grind of significance tests as such, and use the straightforward building‐block approach of Z has a lot to recommend it, although experienced readers and collection managers will probably compare earlier works like Peter Stephen and Susan Hornby’s Simple Statistics for Library and Information Professionals (LAPL, 1995), where the reasoning is less eccentric. For statistical software‐based readers, they will have to graft what this book says on to their knowledge.
Egghe and Rousseau provide interesting discussions of multinomial confidence intervals, and sample sizes for them, referring the reader to more complex works like R.A. Johnson and D.W. Wichern’s Applied Multivariate Statistical Analysis (Prentice‐Hall, 1988) and, in their conclusion, to D.H. Kraft and B.R. Boyce’s Operations Research for Libraries and Information Agencies (Academic Press, 1991), where they imply that statistical applications irresistibly lead. This background is there in their other work but, for a work like this, it would have been useful for the busy practitioner to have had a clearer indication of just what must be read and what might be if you have got the time. In consequence, some of it will be elusive to the layperson, and, if it does not go in, it will not come out as part of the persuasive machinery to convince and impress other people. So, for any good collection (for example where ILS management is taught) this is “should get”, for a keen student and teacher in the area, “could get”, and for the rest of us, get a library copy, use the index, and take what you need.
