1

My question is similar to How do I interpret high residuals for the reproduced correlations in factor analysis?.

I have 5-item Likert-scale questionnaire. I carried out a PCA with N=1253 and I extracted just one factor, and this is consistent with what my items are about. KMO, Cronbach's Alpha, Bartlett's test are all good, but I have 70% of residuals with absolute value higher than .05.

  1. I'm unsure if I have to use a technique other than PCA (frankly, I'm not familiar with any other).
  2. Moreover, I was wondering if there is a way to keep my factor like it is while citing authors who did the same and motivated it in a convincing way. I did not find anything... Does anyone know an exemple?

Thanks!

Community
  • 1
Luke
  • 11
  • 3
  • I think you should read some book(s) about PCA and FA first. The question you link to directly addresses - partly at least - your concern. If you are going to reproduce pairwise correlations (for, you speak of their "residuals") you should better do FA proper, not PCA. – ttnphns Feb 19 '16 at 17:08

0 Answers0