I have mixed type variables (3 quantitative and 3 qualitative) and I calculated Gower's dissimilarity distance between my objects. I wanted to do a hierarchical clustering with hclust, but I am not sure of which method to use as I found in some sites that Ward.D2 must only be used with Euclidean distances, but I dont really understand why... Can someone please help me on this ? Thank you !

  • $\begingroup$ stats.stackexchange.com/a/15313/3277 answers your question. Towards the end, it discusses "Gower distance" and whether it can be used with methods like Ward geometrically or conceptually. (It is irrelant which implementation, Ward or "Ward2", to use ever, if the method is used properly.) $\endgroup$ – ttnphns Jun 25 at 18:19