Iterative Clustering Control Panel | JMP 13.2

JMP Support 800.450.0135 (US)

Documentation Feedback
Your feedback is important to us. us any comments about our documentation.

Multivariate Methods • K Means Cluster • Iterative Clustering Control Panel

•

Iterative Clustering Control Panel

The Control Panel for the Cytometry.jmp data table is shown in Iterative Clustering Control Panel. You can interactively fit different numbers of clusters or you can specify a range using the Range of Clusters option.

Iterative Clustering Control Panel

The Control Panel has the following options:

Locates multivariate outliers. Enter the maximum number of neighbors to examine. When you click OK, plots appear giving distances between each point and that point’s nearest neighbor, second nearest neighbor, up to the maximum nearest neighbor specified.

Caution: The total number of nearest neighbors specified is denoted using a small k. This number is not directly connected to the number of clusters, also sometimes denoted with a small k.

The following options appear beneath the nearest neighbor plots:

Specifies that excluded rows be treated as excluded in future cluster fits within the given report. For example, if you exclude rows based on nearest neighbor plots or the scatterplot matrix, you must select this option to exclude these rows from cluster fits performed within the report using the Control Panel.

Scatterplot Matrix

Creates a scatterplot matrix in a separate window that uses all of the Y variables.

Save NN Distances

Saves the distances from each row to its kth nearest neighbor as new columns in the data table.

Removes the Declutter report.

The following clustering methods are available:

KMeans Clustering

Described in this chapter.

Normal Mixtures

Described in the Iterative Clustering Control Panel in Normal Mixtures.

Robust Normal Mixtures

Described in the Normal Mixtures NCluster=<k> Report in Normal Mixtures.

Self Organizing Map

Described in Self Organizing Map Control Panel.

Number of Clusters

Designates the number of clusters to form.

Range of Clusters (Optional)

Provides an upper bound for the number of clusters to form. If a number is entered here, the platform creates separate analyses for every integer between Number of Clusters and the value entered as Range of Clusters (Optional).

Unless Single Step is selected, fits the clusters automatically.

Enables you to step through the clustering process one iteration at a time. When you select Single Step and click Go, a K Means Cluster report appears with no cluster assignments but containing a Go and a Step button.

–

Click the Step button to step through the iterations one at a time.

–

Click the Go button to fit the clusters automatically.

Use within-cluster std deviations

Scales distances using the estimated standard deviation of each variable for observations within each cluster. If you do not select this option, distances are scaled by an overall estimate of the standard deviation of each variable.

Shift distances using sampling rates

Adjusts distances based on the sizes of clusters. If you have unequally sized clusters, an observation should have a higher probability of being assigned to larger clusters because there is a higher prior probability that the observation comes from a larger cluster.

•

Help created on 9/19/2017