Hello fellow researchers, I have issue regarding discriminant validity. I have 10 constructs in total, but i have 2 constructs with HTMT > 0,9. I want to do EFA for just these two constructs to know which items cross-loaded. I tried to do EFA for all 10 constructs and try to remove offending items but i got worse result. And then i tried to do EFA for just these two constructs and deleted the offending items and i got better result. Is it really appropiate to just do EFA for these two constructs? Thank you in advance.
Hello Gabriela,
Yes, you may explore the (joint) behavior of indicators for the two constructs / latent variables about which you are concerned.
Of course, if these represent existing scales (with extant prior research), then your obligation is to justify the jettisoning of items to achieve some specific goal (in this case, trying to tame what you are characterizing as unduly high HTMT values) and address the possible impact on the nature of the constructs under a new or abbreviated set of indicators.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Computer Science
- Research Topics