Correspondence Analysis

Abstract

Correspondence analysis is a variant of principal component analysis aimed primarily at categorical data, for example, aggregate count data in contingency tables or individual-level responses in questionnaire surveys. The method leads to visualization of the rows and columns of the data table in the form of a map, in which distances and relative positions of points have a specific interpretation. Simple correspondence analysis applies to the cross-tabulation of two categorical variables, while multiple correspondence analysis applies to more than two categorical variables. Both rely on the same computational algorithm, with the data coded in appropriate formats. Different ways of coding data extend the method to other types of data, such as rating scales, preferences, paired comparisons, and continuous data.