ABSTRACT
There are many reasons why cluster analysis may be worthwhile. It might be a question of nding the true groups that are assumed to really exist. For example, in psychiatry, there has been disagreement over the classication of depressed patients, and cluster analysis has been used to dene objective groups. Cluster analysis may also be useful for data reduction. For example, a large number of cities can potentially be used as test markets for a new product, but it is only feasible to use a few. If cities can be placed into a small number of groups of similar cities, then one member from each group can be used for the test market. Alternatively, if cluster analysis generates unexpected groupings, then this might in itself suggest relationships to be investigated.
