2

I'm using the clValid package (Brock et., 2008) to evaluate the stability of a cluster solution.

Interestingly, the package its you select Hierarchical Clustering and AGNES as if these were different algorithms.

It outputs a ranked list with the overall best combinations of distance metric, linkage method and clustering algorithm and here, hierarchical clustering and AGNES seem to achieve different results when using the same distance metric and the same linkage method.

So to gain some definite clarity: Do agnes() and hclust() indeed do the exact same thing? Is the clValid package just misleading in how the ranks are displayed or am I missing something?

John
  • 109
  • 1
  • 8

1 Answers1

3

Here is some relevant information detailing the differences in the use of the Ward method between the two functions.

It explains that hclust() uses two varieties of the Ward method - 'ward.D' and 'ward.D2'. 'ward.D2' is the genuine Ward method - while 'ward.D' is something else. Typing in just 'ward' in hclust() results in the use of 'ward.D'. In agnes() there is only one type of 'ward' option - which is the genuine Ward method.

For differences in other types of linkages read this post

deurali janta
  • 46
  • 2
  • Welcome to StackOverflow deurali! Please consider adding some relevant parts in the answer from the link you have pasted. So that, even if there is a change in the future in those websites, users would be able to understand the solution. – Shibaprasadb Oct 20 '21 at 08:52