The both are still orthogonal rotations, and since it has been stated clearly that not all the factors are correlated with each other, then there is no need to raise the power of correlation amongst factors which therefore means you can not use varimax any longer because it will reduce the number of factors you have (factors that show sign of correlation) , meaning items under these factors will come under one factor because they load high on each other.

You may want to go ahead with quartimax because these factors are not correlated as said by you . This will now hel you see clearly the items loading on each factor despite there might be some fairly correlated ones

Hello Anitha,

If you genuinely believe the factors you're trying to tap to be correlated, then, by all means, do opt for a rotation method that allows factors to be correlated (e.g., promax, oblimin).

Second, conducting EFAs on successive samples is often a frustrating experience, as they frequently yield apparently different (even if there is substantive overlap) structures. The literature is full of such examples, wherein researcher B calls into question the published structure of the "XYZ Scale" because their obtained factor solution isn't identical to that of researcher A.

Instead, I would urge you to try fitting the 8-factor EFA structure identified from the first data set to the data from the second sample. You might find that it fits sufficiently well to recommend the 8-factor solution (rather than the slightly more complex 9-factor solution which you apparently found in 'EFA-2').

Good luck with your work.

To know how to answer your question you should say why you are doing an EFA? What research questions do you have? If everything is correlated, would a bi-factor model be reasonable?

Thank you all for your valuable inputs. Daniel Wright , sir, even though I have developed based on the theoretical model, items are written by me. so I want to check whether the items are defining the factor that they are intended to measure. and the initial questionnaire consists of 79 items (10 factors) and I want to filter the items with low loading and commonalities. so I went for EFA on a sample of 329. but through the parallel analysis, I am getting 9 factors. I have finalized 51 questions for the second try-out. I have collected 1019 data and tried the second EFA on a sample of 400 which is giving 8 factors in the parallel analysis.

Unlike scales like the big five personality or attitude scales where all the factors directly measure a single construct. this scale consists of different factors which don't have a solid theory of influencing each other. but the literature review says they have a relationship.

The Bi-factor model might not fit the purpose. thank you sir..