A two-day short course on Correspondence Analysis was presented by Michael Greenacre on 18-19 December 2006. It gave a comprehensive introduction to the technique as well as recent developments in its application.

Correspondence analysis is now found in many computer software packages.  In this course two alternative ways of computing correspondence analysis were demonstrated: the R computing language freely available on the internet (www.r-project.org), using R functions for simple, multiple and joint correspondence developed in collaboration with Oleg Nenadić of the University of Göttingen; and the commercial package XLSTAT (www.xlstat.com), an add-on package to Excel.  

About the Presenter

Michael Greenacre is Professor of Statistics at the Universitat Pompeu Fabra in Barcelona. He obtained his doctorate in 1978 at the Université Pierre et Marie Curie, Paris, under the supervision of Jean-Paul Benzécri, the originator of correspondence analysis. Since then he has authored and coedited five books and numerous journal articles on correspondence analysis and visualization of categorical data. The most recent book Multiple Correspondence Analysis and Related Methods has just been published by Chapman & Hall (June 2006). Michael has given short courses on correspondence analysis in Germany, USA, Canada, Italy, Switzerland, UK, South Africa and Norway.

About the Topic

Correspondence analysis has found applications in every area of research that involves the collection and interpretation of large sets of categorical data; for example, questionnaire data or tables of frequencies and percentages. Popular areas of application are the social sciences (e.g., comparing demographic groups on their responses to a questionnaire), marketing research (e.g., analyzing brand–attribute relationships) and environmental research (e.g., analyzing species–environment relationships), and also many applications in linguistics, archaeology and geology. It is a method that leads to a visualization of the data in the form of a spatial map, where points in the map represent individuals (or other sampling units), groups of individuals, response categories and/or variables. This type of graphical representation facilitates the understanding and interpretation of large complex data sets and is an essential first step in data analysis before proceeding to more formal statistical modelling.  Correspondence analysis has proved to be the most versatile of the family of multidimensional scaling techniques because of its flexibility in handling different data types and the natural way it incorporates various weighting systems in the analysis.