Correspondence analysis as an explorative tool to analyse different data structures

Correspondence analysis (CA) is popular method for providing a graphical summary of the association between two or more categorical variables. It has gained a reputation for being a quick easily interpreted method of detecting relationships. Despite its popularity, and its acceptance amongst European researchers and those in the UK, the theoretical development of CA in the Australasian region has been relatively slow. Typically CA has been applied exclusively to two-way, or more generally, multi-way contingency tables. However recent, and not so recent, advances make CA a very useful tool for the analysis of other types of data structures. We will look at its application in situations where two categorical variables are nominal data, ordinal data, ranks, continuous and reflect geographical two-dimensional distances. We will also look at some modifications of the classical approach, focusing on identifying an asymetric relationship between the categories using non-symmetrical correspondence analysis (NSCA)