0
$\begingroup$

I have a short question. I've performed a principal component analysis and obtained two components. Are two components enough two perform a cluster analysis (number of participants > 400)? Thanks for your help!

$\endgroup$
  • $\begingroup$ As many as neccessary. Visualize your data. If you can't see clearly cut clusters in 2d, then algorithms won't return meaningful results either. $\endgroup$ – Has QUIT--Anony-Mousse Aug 12 '14 at 7:46
0
$\begingroup$

If the goal is to identify optimal number of clusters from the data, one way is to use Consensus Clustering which performs repeated random sampling on the data, and identifies the number of clusters that has the most robust cluster membership across variables.

To answer your more question more directly, N top components should correspond to 2*N clusters in the linear space as each component represents variables correlated and anti-correlated with it.

$\endgroup$

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.