1
$\begingroup$

My psychology dissertation will be a cluster analysis with one grouping variable. How do I know how many participants I need? I imagine there should not be more than 5 clusters. 79 items make up 9 attributes that the participants will be clustered on. I realize that power analysis is moot since cluster analysis is not inferential. I have seen that for factor analysis it is 10 per item but cannot seem to find a guideline for cluster analysis.

Thanks for any help or insight you can offer.

$\endgroup$

1 Answer 1

1
$\begingroup$

You never know, before you have the data.

In particular with small sample sizes, it's easy to tweak parameters, preprocessing, algorithms etc. to get out whichever answer you like.

Do not assume that the output of a cluster analysis is "correct". There is no such thing as correct. Either the results lead to insights (which you should then verify in a second experiment) or not. Consider cluster analysis a tool to find out which questions to ask; but not as a means to prove that you have found anything.

$\endgroup$
2
  • $\begingroup$ Unfortunately this won't fly for my proposal. Can I use G-Power and simply calculate for MANOVA? Or use factor analysis recs of 10 per item? And what would be the precedent for that? I am having NO LUCK finding anyone with recommendations. $\endgroup$
    – Hether
    Jul 10, 2014 at 12:01
  • $\begingroup$ You've got a recommendation: Don't do it! $\endgroup$ Mar 10, 2019 at 11:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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