Navigate Tree

Click Navigate on the Tree menu and ClustanGraphics displays a top down view of a hierarchical cluster analysis. It takes the following form:

Summary tree for 5 clusters, showing the cluster means for the variable Protein, cluster sizes and exemplars

This shows how the last 5 clusters in the Mammals case study combine into the whole data set over 4 final fusions. The boxes shaded yellow correspond to the 5 final clusters, and those shaded blue are the higher-order clusters into which they amalgamate - the colours can be changed. Each box is labelled by the cluster exemplar, and the value displayed is the mean percentage of protein in the milk for the cluster.

Right-click and drag your tree to re-position it; or double-click on a higher-order cluster to prune branches. In the following example, we pruned the branch below the cluster labelled "Sheep" by double-clicking on it:

Example of pruning a tree

Click Edit to edit the tree. This allows you to re-position or edit every node in your tree; prune or expand branches; re-organize the tree and its connections by point-and-drag; edit the text within the boxes or the border shapes and their colours and gradiants; change node sizes; add titles and legends; zoom in or out; combine two or more trees; and so on.

Navigate Tree's edit screen, in which you can re-position, move and generally control the look of your final tree

You can select any number of clusters and zoom in or out to display sections of special interest. For example, after some editing the full tree for the mammals case study looks like this:

Zoom out view for a full tree

In this tree, the yellow boxes correspond to cases while the blue boxes denote clusters of two or more cases. Point the cursor at any box and ClustanGraphics displays cluster-specific data, such as the cluster's position in the tree and its exemplar (if available). Recall that the tree can be optimally ordered, so there should be a meaningful horizontal arrangement of cases.

Navigate Tree can be used with incomplete data and mixed data types. When displaying a nominal variable the most common state within each cluster is shown together with its percentage of occurrence. For example, at the 3-cluster level in the following tree cluster -3 is characterized by Account Type 1, wihch occurs in 96% of the 132 cases, whereas 2 is the most common Account Type in the other two clusters (clusters +2 and +3).

Summary tree for a nominal variable, showing the most typical Account Type for each cluster and its percentage of occurrence within the cluster

Navigate Tree can be used on any size of cluster model. Check out how to use Navigate Tree with k-Means to cluster a survey of 16,000 cases in just 4 seconds on a PC, and then summarize the resulting 40-cluster model. Details here.

We're rather pleased with Navigate Tree! It's another example of why we can justly claim that Clustan is A Class Act.