Statistics Toolbox

Getting More Information About Cluster Links

One way to determine the natural cluster divisions in a data set is to compare the length of each link in a cluster tree with the lengths of neighboring links below it in the tree.

If a link is approximately the same length as neighboring links, it indicates that there are similarities between the objects joined at this level of the hierarchy. These links are said to exhibit a high level of consistency.

If the length of a link differs from neighboring links, it indicates that there are dissimilarities between the objects at this level in the cluster tree. This link is said to be inconsistent with the links around it. In cluster analysis, inconsistent links can indicate the border of a natural division in a data set. The cluster function uses a measure of inconsistency to determine where to divide a data set into clusters. (See Creating Clusters for more information.)

The next section provides an example.

Verifying the Cluster Tree

Example: Inconsistent Links