Statistics Toolbox

Specifying Arbitrary Clusters

Instead of letting the cluster function create clusters determined by the natural divisions in the data set, you can specify the number of clusters you want created. In this case, the value of the cutoff argument specifies the point in the cluster hierarchy at which to create the clusters.

For example, you can specify that you want the cluster function to divide the sample data set into two clusters. In this case, the cluster function creates one cluster containing objects 1, 3, 4, and 5 and another cluster containing object 2.

T = cluster(Z,2)
T =
    1
    2
    1
    1
    1

To help you visualize how the cluster function determines how to create these clusters, the following figure shows the dendrogram of the hierarchical cluster tree. When you specify a value of 2, the cluster function draws an imaginary horizontal line across the dendrogram that bisects two vertical lines. All the objects below the line belong to one of these two clusters.

If you specify a cutoff value of 3, the cluster function cuts off the hierarchy at a lower point, bisecting three lines.

T = cluster(Z,3)
T =
    1
    3
    1
    2
    2

This time, objects 1 and 3 are grouped in a cluster, objects 4 and 5 are grouped in a cluster, and object 2 is placed into a cluster, as seen in the following figure.

Finding the Natural Divisions in the Data Set

Linear Models