Cluster analysis - using statistiXL package for excel - how to determine a cluster

#1
Hello,

I have a fairly basic question.

I've used the trial package of statisiXL for excel to run a cluster analysis on the following data:

9 elements (Na, Ca, etc..) were measured at 100 occasions, so 900 measurement points in total. All have a similar range.

Used settings: Quantitative data set; Euclidean distance; Nearest neighbour cluster method.

This gave a very nice hierarchical tree with two obvious clusters. It was obvious because you can 'see' two different clusters. But how do determine, statistically, the significance of a cluster/group? In other words; how do i determine a "critical distance"?

Could I use a simple statistical test? Including number of samples etc.?


My apologies if the question has already been answered on this forum, I couldn't find it.

Thank you in advance!
 
Last edited:

terzi

TS Contributor
#3
Re: Cluster analysis - using statistiXL package for excel - how to determine a cluste

The Duda-Hart index is a commonly used stopping rule for some iterative clustering techniques, it can also be useful for selecting the number of cluster with hierarchical approaches.