Correspondence Analysis and Data Coding With Java and R by F. Murtagh
Stanley E. Lazic · Journal of the Royal Statistical Society Series A (Statistics in Society) · 2007
This book provides a readable introduction to the theory and application of correspondence analysis, and is suitable for senior undergraduate or post-graduate study. A good knowledge of R is assumed on the part of the reader (or the Java software can be used instead). The book briefly discusses the history of the field, highlighting the work of J. P. Benzécri and the ‘French school’, where much of the early work on this technique took place. There is a useful chapter on data coding, which demonstrates the flexibility of this method in handling various types of data in the same analysis. Practical examples are from a variety of fields, with many (and a chapter) devoted to the analysis of text. This book has recently been reviewed elsewhere ( de Leeuw, 2005; Holmes, 2006) and therefore the present review examines it purely from a user's perspective. A criticism of this book is the structure and organization of some of the chapters, particularly when illustrating examples. It breaks with the logical flow that many texts use: discussion of the method, technique, principle or concept to be demonstrated; introduction of the data set; use of software to conduct the analysis; display of numerical and graphical results; and, finally, discussion or interpretation of the output. The R algorithms that are used are at the end of each section or chapter and readers must flip back and forth, making the relationship between the original data, the coded or transformed data, the implementation in R and output somewhat unclear. The functions to implement the analyses are more an afterthought, rather than an integral part of the process. All of the information is present; it just requires some extra work to link it all together. Having the functions available as a package on the Comprehensive R Archive Network would have certainly been useful. In addition, reference could have been made to other packages on that network that contain correspondence analysis functions, if only to draw readers’ attention to these.