If the usual correspondence analysis is viewed as a decomposition of departures from the model of independence, there are problems in applications to sociomatrices with regard to the treatment of the diagonal and cell estimates under independence. Other related techniques from the literature are more appropriate for analysis of sociomatrices. The different approaches are used to analyze a familiar sociomatrix, and the results of the techniques are compared.
Notes
The author is indebted to A. Kimball Romney for valuable help and advice. The Editor and anonymous reviewers made numerous suggestions throughout, particularly with regard to interpreting correspondence analysis without appealing to departures from a model and to issues of diagonal treatment.