dendrograms in clustergram vs pdist->lin​kage->dend​rogram

Hello,

Can somebody explain why the dendrogram produced by clustergram is different than the one obtained by the traditional pdist, linkage and dendrogram process?

As I understand clustergram uses Euclidean distance metric and Average linkage. But when I run this functions with the aforementioned parameters the resultant groups are different than those displayed by clustergram.

How can I reproduce the same dendrogram produced by clustergram using the pdist->linkage->dendrogram approach?

Thanks in advance.

[Diego]

I thought this could be explained by the standardization performed by clustergram.

However, I use zscore for both, the matrix and each column in the matrix and the dendograms are still different than those of clustergram.

Any idea is much appreciated.

Answer 1

[Lucio Cetto]

Clustergram standardizes the data. I am not sure what release you are using but the way you control this option with the input arguments to clustergram has not been consistent. Also clustergram runs optimal-leaf-order.

HTH

Lucio

[Diego]

Thank you Lucio. Yes the optimalleaforder can be an explanation. I'm a little bit rusty in matlab and also it's my first time handling a Class. So it can take me sometime to perform the tests.

Thanks again.

Diego