Brigitte Le Roux and Henry Rouanet (with a foreword by Patrick Suppes)

To many sociologists, a book with the title Geometric Data Analysis may at first seem of little relevance to their own work. The subtitle, however – From Correspondence Analysis to Structured Data Analysis – should indicate otherwise. As is well known, multiple correspondence analysis was the preferred statistical tool of the late Pierre Bourdieu, with whom the two authors collaborated from the mid-1990s and onwards. The book gives a state-of-the-art introduction not only to correspondence analysis, but also, through several examples, to dealing with problems met in practical multivariate analysis using methods for analysing latent structures in the data. The book is organised into 10 chapters, covering all the fundamental issues in geometric data analysis. Chapter 1, Overview of Geometric Data Analysis, gives a short review of the history of correspondence analysis, a general introduction to the chapters to come, and also some very clear messages about the book’s epistemological foundation. Citing Ludovic Lebart, the authors make it clear that ‘Statistics does not explain anything – but provides potential elements for explanation’ (pp. 19–20). The widely used and misleading opposition between so-called exploratory and explanatory techniques is thus strongly refuted. The same goes for the understanding that correspondence analysis and statistical inference exclude each other (pp. 18–19)

Correspondence analysis · Data science · Epistemology · Exploratory analysis · Exploratory data analysis · Inference · Linguistics · Mathematical economics · Sociology · Statistics · Subtitle · Computer Science · Mathematics · Sensory Analysis and Statistical Methods · Philosophy