Acritical issue in cluster analysis is the determination of the appropriate number of clusters Unfortunately, no generally accepted criterion exists. Researchers are therefore reduced to using existing theory to identify a natural number of
clusters that are interpret able in terms of the research question. However, as a complement, more formal ‘‘rules of thumb’’ can be used. One such method is to examine how the distance between objects within clusters changes (the vertical axes in Fig. 2) as the number of clusters decreases. The idea is to identify the points where within-cluster
distance makes a sudden jump. In Fig. 2a two ‘‘jumps’’ were identified between two and three clusters, and between three and four clusters. Both solutions were examined. The three-cluster solution was chosen because it provided clusters that
were consistent with previous research in the area
(see also section Definition of constructs above).
On the same grounds, the three-cluster solution in
Fig. 2b was considered the most interesting (see
also section Theory development).