Correspondence analysis is a technique for approximating a contingency table with lower rank tables to analyze the relationship between two categorical variables. It works by finding pairs of correspondence factors that have unit variance with respect to the marginal distributions and are maximally correlated. The correspondence factors and their correlations are obtained from the singular value decomposition of a normalized contingency table. Hypothesis tests can then be conducted to test the independence of the categorical variables and how well a lower rank approximation fits the data. The analysis also provides a spatial representation of the row and column categories in lower dimensions.