I need help with my data analysis. I am using a non-validated translated questionnaire in my survey. The purpose of it is not to validate the measure in a different language and its construct validity, I only translated it for the purpose of my study. Still, I am obligated to test the construct validity of the questionnaire and the problem is the following:
When I run the principle axis extraction with oblique rotation, I initially get 2 factors-two of the items load very poorly on the 1st factor (zero corrleations), and some have small values of 0.3, but also correlate with the second one. Majority loads on the first factor and the MAP test also gives me a reason to extract only one factor. If I run the principle axis extraction without any rotation-it only suggests me to remove one item, based on the correlations with the factor, not two.
My question is-can I decide to extract a "fixed number of factors" and extract only one after concluding that MAP test gives me a better solution? The second question is-if I do so, am I allowed to only remove that 1 item that it offers me to remove after extracting 1 factors and ignore the almost zero loadings in the initial EFA with oblique rotation of the second item? The other problem with removing too many items is that it messes my variance too much. It is in my interest to keep as many items as possible, but only the mentioned one I think I must remove (the second one I am not sure whether I must).
Thank you so much for your response, I am really stuck because of this problem.