Subscribe Bookmark RSS Feed

DendoGram Interpretation


Occasional Contributor


Feb 25, 2017


Attached is the output for dendogram and two way clustering analysis.

How do I interpret this?

And how do i choose the best possible number of clusters from this data.CaptureDendoGram.PNG




Mar 21, 2013

You can use the black V-shaped handle at the top of your picture to select the number of clusters you want. I'm the wrong person to tell you how to pick the best number.

The dendrogram is a tree; the leaves are the rows. The leaves join together, one-by-one, with the next most similar leaf. The (tiny) vertical line that joins two leaves forms a new node that acts like a leaf and can join with another node or leaf. The vertical joiner lines move further to the right for subsequent joins. You might interpret (in your picture) the vertical black lines (further right) joining clusters that are more different than the vertical color lines (further left).

The ordering of the data from top-to-bottom is arbitrary; at the top of your picture the red/green/blue chunks could be rotated into blue/green/red by swapping the ends of the vertical black joining lines.

One of the cool things the dendrogram does is selection; you can click inside the dendrogram to select the corresponding rows in the table, then set the color/marker/hide/exclude state for the selected rows.