Ratio maps and correspondence analysis

We compare two methods for visualising contingency tables and develop a method called the ratio map which combines the good properties of both. The first is a biplot based on the logratio approach to compositional data analysis. This approach is founded on the principle of subcompositional coherence, which assures that results are invariant to considering subsets of the composition. The second approach, correspondence analysis, is based on the chi-square approach to contingency table analysis. A cornerstone of correspondence analysis is the principle of distributional equivalence, which assures invariance in the results when rows or columns with identical conditional proportions are merged. Both methods may be described as singular value decompositions of appropriately transformed matrices. Correspondence analysis includes a weighting of the rows and columns proportional to the margins of the table. If this idea of row and column weights is introduced into the logratio biplot, we obtain a method which obeys both principles of subcompositional coherence and distributional equivalence.