Interpreting multiple correspondence analysis

Abstract
The practical aspects of interpreting multiple correspondence analysis are reviewed. The geometric concepts associated with simple correspondence analysis are shown to be inadequate for multiple correspondence analysis. An alternative approach, called joint correspondence analysis, is shown to be a more natural generalization of the simple case. The practical interpretation of homogeneity analysis is also discussed and a compromise is proposed between classical multiple correspondence analysis and joint correspondence analysis, preserving the optimal scaling properties of homogeneity analysis. A simple example is used throughout the discussion to illustrate the issues involved.