I am a biology student investigating a new method of creating a dichotomous identification key. I have created a dendrogram using data I have collected from a survey on how people rate how similar pictures of plant leaves are. I used ward's method to link the clusters. In the resulting dendrogram, I have a y-axis that ranges between 0 and about 50. I know that this axis represents at which the objects are joined in a cluster, thus how far they are from other objects, but I was wondering what exactly does the numeric value represent?
-
$\begingroup$ Do you have a reference for this graph? $\endgroup$– AndyJul 31, 2015 at 17:24
-
1$\begingroup$ Please read the documentation for the clustering program you used. $\endgroup$– ttnphnsAug 1, 2015 at 14:57
1 Answer
I'm going to, ahem, go out on a limb here, ahem, and guess that you built your tree via the hclust
function in base R with method = "ward.D2"
, which is Ward's original method. If you type ?hclust
and look for height
in the value
(output) section, it says "The clustering height: that is, the value of the criterion associated with the clustering method for the particular agglomeration." In this case, Ward's criterion is the total within-cluster error sum of squares, which increases as you go up the tree and make the clusters bigger.