MATLAB: Cut a dendrogramm at a specific height to generate clusters

Question

Hello, I am trying to group some columns of a complex-valued matrix with the hierarchical clustering method. I created the dissimilarity matrix "dissim" and extract the distance vector from this. I am then generating the linkage matrix and feeding it to the dendrogram function, where I used the example from the mathworks help center to find the (in this example very obvious) 4 groups of entries. The dendrogram looks fine, but as soon as I am applying the cluster() function to the linkage, I am getting rubbish output:

    distvec = distancevector(dissim);
    tree = linkage(distvec, 'average');
    
    cutoff = 0.5;
    CL = cluster(tree, 'Cutoff', cutoff, 'Criterion', 'inconsistent');
    dendrogram(tree,'ColorThreshold',cutoff);

The colored dendrogram looks like this, which shows the groups exactly as I want it:

So, the output I would be expecting from the cluster function would be something like this:CL =

CL =
    
    2
    2
    3
    3
    1
    4
    2
    1
    4
    1
    4
    4
    2
    1
    3
    3

i.e, 4 data entries are matched to each of the four clusters. However, CL is put out like that:

CL =
     5
     8
    12
     7
     4
     2
    11
     9
    10
     9
     3
    10
    11
     1
     6
    12

I think the dendrogram() function has a different understanding of the cutoff that the cluster() function. How can I fix it so that the clustering works as I want it to?

Answer 1

I solved it. It turned out that the cutoff used for the ColorThreshold option in the dendrogram-function is using the distance criterion cutoff definition of the cluster-function.

The issue was fixed when I replaced

 CL = cluster(tree, 'Cutoff', cutoff, 'Criterion', 'inconsistent');

with

  CL = cluster(tree, 'Cutoff', cutoff, 'Criterion', 'distance');