Agglomerative hierarchical clustering Using Ward Algorithm and Correlation Coefficient as a Distance Metric

Dear colleagues,

In the Matlab help page for agglomerative clustering (<http://www.mathworks.com/help/stats/linkage.html)>, it says that ward method appropriate (only) with euclidean distance.

I am wondering if other distance metrics (like correlation coefficient or cosine similarity measure) can be used. Correlation coefficient and cosine measures gave me better results when I used them instead of the euclidean distance.

So, I am wondering why it was written in the matlab documentation page that it works only with euclidean distance. I found some other research that use ward with correlation coefficient and some other people asking the same quesiton: http://r.789695.n4.nabble.com/hierarchical-clustering-with-pearson-s-coefficient-td4662788.html

Regards,

[Abdullah]

Here is some work that used Ward method with correlation coefficient http://openi.nlm.nih.gov/detailedresult.php?img=2891634_1471-2407-10-222-4&req=4