ABSTRACT
Cluster analysis is a statistical technique with a long history, and the types of clustering are many. The type of clustering used on the results from an Multiple correspondence analysis (MCA) is based on Ward's method and the minimum variance criterion. This criterion seeks to minimize the intraclass variance and to maximize the interclass variance, and the approach represents yet another way of partitioning the variance in the cloud of individuals. A partitioning based on Ward's criterion is good if the internal variation in the class is low, i.e. the within-variance is small and the between-variance between two or more classes is large. This is expressed in Ward's aggregation index, which is a measure on the quality of the classification or clustering. Using Ward's criterion, there will necessarily be an increase in the intraclass-variance for each cluster that is merged. The cluster analysis has also uncovered sub-groups that couldn't be identified by the MCA alone.
