0
$\begingroup$

I computed a PCA and am interestet in the explained variance of the first unrotated component. The same procedure was used in a previous study.

Question: How do I test whether the two explaned variance ("mine" and the one from the previous study) differ significantly?

Note: Procedure and input variables are the same, samples are independent and differ in size.

$\endgroup$
  • $\begingroup$ - Some samples could don't come from the same distribution - In some sample could exist lacks of data - Did you remember about scale variables before start of PCA? $\endgroup$ – fuwiak Jul 19 at 16:32
  • 1
    $\begingroup$ Suppose yours explains more variance than the previous one, then what does it mean? I do not think it mean that your PCA is better than previous PCA. $\endgroup$ – user158565 Jul 19 at 18:22
0
$\begingroup$

Since the "explained variance" is a proportion, you could do a test of the differences in the proportions. For instance, R has the prop.test function.

Why you are interested in this is a good question; what do you hope to do if they are significantly different? What about if they are not?

$\endgroup$
  • $\begingroup$ This is being automatically flagged as low quality, probably because it is so short. At present it is more of a comment than an answer by our standards. Can you expand on it? We can also turn it into a comment. $\endgroup$ – gung Jul 20 at 11:53
  • $\begingroup$ I added some more. $\endgroup$ – Peter Flom Jul 21 at 12:48

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.