Browse Books

Reviewer: William A Fahle

The sheer power of today's hardware allows one to work with enormously large amounts of data. If the data are of a numerical nature, then, typically, one will not be able to draw any useful conclusions from just looking at millions of numbers. A technique to visualize the numbers in a way that clearly shows the important features (and, if possible, suppresses any irrelevant information) is required. Normally, graphical methods are the methods of choice in such a situation. The data may come from very different sources. Sometimes we talk about huge amounts of statistical data (in particular, in finance and medicine), and sometimes we deal with the numerical simulation of physical processes by, say, a finite-element scheme. This book only covers the former class of problems, not the latter. Regrettably, the title of the book does not indicate this at all, and the cover text is not very explicit in this respect either, so the only hint that the reader gets is the fact that the book has been published as part of a book series called "Springer Handbooks of Computational Statistics." Thus, there is a certain danger that the buyer of this very expensive book will not find what he or she is looking for. Within the restricted range of topics covered in the almost 1,000-page book, it is probably fair to say that no important aspects are missing. Section 1 gives an introduction to and a brief survey of the contents of the other parts. Section 2 consists of ten papers discussing the basic principles of visualization: an overview of the history of the area; guidelines for the preparation of graphics to make them reader friendly; requirements to impose on corresponding software; and various visualization techniques (graph-theoretic methods, techniques for tree models in statistics, methods for dealing with multidimensional data, and linked views). Section 3, more than 500 pages, is the longest part of the book. This section contains 18 papers, each addressing one special aspect of graphical visualization in a very detailed way, including: smoothing and other techniques to handle noise in the data; the visualization of geography-related data; and efficient ways to use Java and Extensible Markup Language (XML) to develop software for visualization applications. Section 4 consists of seven papers that discuss concrete applications and the methods required to visualize the corresponding data. The applications arise in medicine (genetic networks and medical imaging) and finance (analysis of bankruptcy data, statistical aspects of eBay auctions, and the modeling of certain aspects of insurances). More than 60 authors have contributed to this book. Therefore, many questions have been addressed from different points of view, which is certainly an attractive feature. On the other hand, it also means that a reader looking for very specific information may have difficulty finding it, as various topics tend to be scattered over two or three papers. The scientific level of the book is very high. Thus, my impression is that it is written for people with a certain degree of familiarity with the problems, and not for beginners or practitioners from other areas who just need to visualize a set of data that they have encountered. This is already evident from the introduction, where readers find statements such as, "Everyone knows about Chernoff faces." Maybe everyone in the statistics-visualization community knows about Chernoff faces, but most mathematicians, engineers, and computer scientists probably do not. Finally, the presentation of the graphics is worth a few words. In a book with such a topic, and in particular in a book that is so expensive, I would expect to see graphics of the highest possible level. Whereas most of the figures in the book are good, some of them are really disappointing. Two issues in particular stand out. The first may be found in the text on page 6, related to Figure 1.2: "The scales have been chosen to maximize the data-ink ratio. An alternative version ... makes [it] clearer." This is a strange attitude toward the reader. Surely, "Choose the representation that saves ink even if it means the message is not clear" cannot be serious advice in the context of visualization. My second issue is with the way the publisher handled the color graphics that some authors presented in their papers. There is a small section in the middle of the book where all of these color figures are collected, whereas the papers themselves only contain grayscale versions of these figures. To get the additional information provided by the colors, readers need to browse forward and backward; this is a true nuisance. In spite of these weaknesses, the book is useful for members of the target group: namely, people who deal with visualization problems every day. For people who only deal with a visualization problem in some other field, I have doubts about its suitability. Online Computing Reviews Service