How to calculate a weighted Hierarchical clustering in Orange

Question

I am doing my first cluster analysis with Orange (which I recently discovered and looks promising for this iterative and interactive process).

Apparently, there are several methods of creating clusters based on distance algorithm:

• Single linkage (which computes the distance between the closest elements of the two clusters)
• Average linkage (which computes the average distance between elements of the two clusters)
• Complete linkage (which computes the distance between the clusters’ most distant elements)
• Weighted linkage
• Ward

Since I have several columns, and some of them are more important than others in terms of defining clusters, it seems to me that using weighted linkage method may be what I am looking for. Unfortunately, I don't know how to do that, since I didn't find a way to assign a weight to each column.

To make things worse, I have only found an explanation about the first three on this Orange's blog post but nothing about Weighted linkage (nor Ward, which may be a recent addition, since it is not even mentioned on widget's help).

Am I on the right path to achieve what I am looking for? Is there any way to make some columns more or less important/definitory when calculating the distances?

Answer 1

Weighted linkage probably does not mean you get to specify weights of features (build the distance matrix yourself!)

Instead this most likely refers to the well-known weighted group average strategy you will find in most textbooks often called WPGMA. There are two different definitions of "average", so this is likely simply the "other" average linkage.