(hierarchical techniques, k-means methods and probabilistic clustering) may be used in
principle.
Agglomerative hierarchical techniques can be - dependent on your hardware - applied for small (e.g.
n=50) and moderate sample sizes (e.g. n=500). K-means methods require at least a moderate sample
size (e.g. n=300). K-means can be used for large sample sizes, too. Probabilistic techniques require
large sample size (e.g. n=3000). The size of sample depends on the structure of the data. If well
separated clusters exist the sample size can be smaller. Therefore, general threshold values cannot be
given.
Explorative Cluster Analysis and Confirmatory Cluster Analysis
Clustering techniques are regarded as explorative methods in many text books. This is
correct only to some extent. They only require a specificatoin of the variables and cases that
should be used. A specification of the number of clusters is not necessary in advance. The
number of clusters can be determined in principle. It is also not necessary to specify certain
characteristics of the cluster (e.g. clustering variables: country A and B are in the same
clusters; or clustering cases: cluster 1 prefers countries A, B and C, cluster 2 countries A and
D, and so on). Therefore, clustering techniques may be used in an explorative way. But they
can also be applied as confirmatory techniques. In this case the number of clusters and certain
characteristics of the clusters are fixed. Figure 1-2 shows the differences between exploratory
and confirmatory cluster analysis.