Doubts: VIF values and high correlations

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
agalvez
PLS Expert User
Posts: 39
Joined: Mon Jul 04, 2016 10:17 pm
Real name and title: Alex

Doubts: VIF values and high correlations

Post by agalvez »

Hi mates

I have the following two doubts with SmartPLS. I would greatly appreciate your help.

On the one hand... I checked that some items of my model have VIF values higher than 5 and than 10 (some have VIF values of "20" or "30"). I have read that in this case one has to delete the problematic items.

But, what should be process to follow? I think I should delete them step by step, deleting first the item with highest VIF, given that VIF values may change when one of them is deleted.

On the other hand... Item loadings of my model have values < 0.90 and correlations among latent variables are < 0.80. However, discriminant validity is confirmed.

Constructs are conceptually very similar. In fact, some researchers argue that they could be identical. This happen, i.e., with service quality and satisfaction. Both are closely related and I think in this case correlations can be very high.

THANK YOU IN ADVANCE
User avatar
cringle
SmartPLS Developer
Posts: 818
Joined: Tue Sep 20, 2005 9:13 am
Real name and title: Prof. Dr. Christian M. Ringle
Location: Hamburg (Germany)
Contact:

Re: Doubts: VIF values and high correlations

Post by cringle »

Do you try to assess a reflective or formative measurement model? Sounds like a reflective one since you care about the loadings being higher than .7. For reflective measurement models the VIF value is not relevant.

Best
CR
agalvez
PLS Expert User
Posts: 39
Joined: Mon Jul 04, 2016 10:17 pm
Real name and title: Alex

Re: Doubts: VIF values and high correlations

Post by agalvez »

Dear Ringle

I'm sorry. Reflective model.

So, VIF values are not relevant? I though I shows care of LV VIF (inner model)?

What about correlations? Higher correlations among constructs (< 0.80) make difficult to guarantee discriminant validity, even being them conceptually different.

Thank you in advance
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Doubts: VIF values and high correlations

Post by jmbecker »

For reflective models you care only about VIFs in the inner model. VIFs in the outer model are allowed to be high.
Are the high VIFs that you report in the inner or outer model?

Your first messages sounds like they appeared in the outer model, because you were talking about deleting "items". You should not delete items because of high VIFs in the outer model for reflective constructs.
If you have high VIFs in the inner model you might check cross-correlations of items and delete those who are problematic. Please also read viewtopic.php?f=5&t=16006
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
agalvez
PLS Expert User
Posts: 39
Joined: Mon Jul 04, 2016 10:17 pm
Real name and title: Alex

Re: Doubts: VIF values and high correlations

Post by agalvez »

jmbecker wrote:For reflective models you care only about VIFs in the inner model. VIFs in the outer model are allowed to be high.
Are the high VIFs that you report in the inner or outer model?

Your first messages sounds like they appeared in the outer model, because you were talking about deleting "items". You should not delete items because of high VIFs in the outer model for reflective constructs.
If you have high VIFs in the inner model you might check cross-correlations of items and delete those who are problematic. Please also read viewtopic.php?f=5&t=16006
Now, I understand everything better. I also checked the other thread. Thank you again. It is helpful to have great professionals in this forum.
Post Reply