I have a question regarding cluster analysis.

I was wondering if I have only 9 data points, is it valid to use k-means methods in cluster analysis?

I have done a special molecular evolution analysis and it has generated 9 points, what I want to know is that whether there is any clustering among the groups. What I did was to just assign numbers to 3 groups and then do ANOVA and post-hoc t-test, but I was wondering if there is a more appropriate method for that?


  • 1
    $\begingroup$ What data have you got about these nine points? $\endgroup$ – Peter Flom - Reinstate Monica Dec 14 '13 at 16:26
  • $\begingroup$ It's unclear to me what you mean by "points," "groups," and "numbers." You 'have' 9 data points; the analysis 'generated 9 points'; you are interested in 'clustering among the groups'; and you 'assigned numbers to 3 groups.' Clarification would probably bring you more answers. $\endgroup$ – rolando2 Dec 15 '13 at 4:46

For small data sets, hierarchical linkage clustering is a good choice.

Even single-linkage may do the trick for you.

You can afford to use the most expensive clustering algorithm (most linkage variants are in $O(n^3)$) if $n=8$; it's only for really large data sets where k-means shines because of its linear runtime.

Plus, linkage clustering allows you to use a domain specific similarity, measure, whereas k-means assumes your data to be Euclidean.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.