I've been using the CH criterion to determine the optimal number of clusters for some data I'm working with. I know that in general you would evaluate the criterion for a number of different choices of clusters, and choose the number of clusters which gives the highest CH value, or one where there is a large increase in CH score.

For my data, the optimal number of clusters is 2. Plotting the CH score against number of clusters, I can see the score decreases very quickly if you increase the number of clusters.

My question is this: Since the CH criterion does not evaluate for 1 cluster, how can I be sure using 2 clusters is better than one cluster? I.e how can one determine if there is evidence for any clustering at all in the data?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.