Clustering and scaling of transaction flow tables: A French interdepartmental migration example

Measures of functional linkages between analytical units are obtainable from transaction flow tables with the use of the iterative proportional fitting procedure (IPFP). Two quite different approaches-directed graph-based hierarchical clustering and robust metric multidimensional scaling-for the summarization of large tables of these measures can be taken. The results of the clustering are particularly responsive to strong linkages, while those from the scaling are more sensitive to weak ties. Residuals from both models indicate the adequacy with which the flow patterns are explained. The scaling method can be used to study the residuals from the clustering. These procedures are illustrated through the study of a 1954-1962 French interdepartmental migration table.