Dendrogram

Purpose

Dendrogram Display a dendrogram based on cluster output.

Synopsis

dendrogram(fig,clusterdata,options)

dendrogram(clusterdata,options)

dendrogram(fig,action)

Description

A utility function called by the CLUSTER routine. This is generally not a user-accessible function. Input (clusterdata) is a structure with the following fields: order, options, classes, ins, labels, m, n, desc, dist, cls. All of which are defined within the code of CLUSTER.

Note: When running "Cluster" in Analysis window, the threshold value selected by the user in the dendrogram plot is stored in "x.userdata.threshold" after the user clicks on the "Keep..." button. When running "cluster" from the command line, for example: [results,fig,distances] = cluster(arch,options); the threshold selected by the user in the dendrogram plot is stored in appdata and can be retrieved as: cd = getappdata(fig,'clusterdata');cd.threshold. Important: The figure must remain open to obtain clusterdata.

Inputs

• fig: Figure object handle to display results.
• clusterdata: Results obtained from cluster analysis.
• options: Options structure.

Options

The following options are internal to the function but can be set using setplspref:

• linewidth: [2] defines the width of the lines in the dendrogram.
• maxlabels: [200] defines the maximum number of labels which will be shown on the dendrogram. If more than this number of samples exist, no labels will be shown.
• clickassign: [ 'off' |{'on'}] governs clicking to assign classes. If 'on', user can click on dendrogram to assign classes to samples based on splitting the dendrogram at the indicated distance.
• clickmarker: [ 'off' |{'on'}] governs showing of marker line where dendrogram would be split (used with clickassign).
• optthreshold: [ inf ] governs optimization of colors. Above this number of samples, the color pallet is optimized to improve contrast in the dendrogram. However, optimized colors will not match color order in other interfaces. Typical value is 10 to 20. inf = no optimization is ever done.

See Also

analysis, corrmap, dbscan, cluster, gcluster, knn, knnscoredistance, simca