VIF value above 5

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
Catharsis
PLS Junior User
Posts: 7
Joined: Mon Mar 04, 2019 11:52 am
Real name and title: Human Kohzadi Student

VIF value above 5

Post by Catharsis »

Hello,

I am currently investigating some reflective research models with SmartPLS and have encountered a difficulty. In one model, the VIF value for a construct is 6.8 - greater than 5, but in my current work I assumed a VIF limit of 5. At this point, can you recommend what I should do at best? Just remove the construct? Or are there other solutions to be considered?

Hope you guys can help me!

Best regards!
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF value above 5

Post by jmbecker »

Are you talking about outer or inner model VIFs?

If you have reflective measures you do not care about outer model VIFs. In fact, VIFs for reflective outer models need to be high if you have reliable indicators with high loadings.
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
Catharsis
PLS Junior User
Posts: 7
Joined: Mon Mar 04, 2019 11:52 am
Real name and title: Human Kohzadi Student

Re: VIF value above 5

Post by Catharsis »

In this case it is the inner model...thats why I'm actually concerned.
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF value above 5

Post by jmbecker »

Do you achieve discriminant validity with these high VIFs? Usually the high correlations that come with high inner model VIFs are also a problem for discriminant validity.

What is your sample size?
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
Catharsis
PLS Junior User
Posts: 7
Joined: Mon Mar 04, 2019 11:52 am
Real name and title: Human Kohzadi Student

Re: VIF value above 5

Post by Catharsis »

In my work I use the Fornell-Larcker criterion to check the validity of discrimination. However, there are some inconsistencies when it comes to cross-loadings. I have, in fact, noted from some recent publications that cross-loadings are said to be inappropriate for testing discrimination via SmartPLS (Hair et al., 2017; Henseler, 2015). The sample size in this case is n=34.
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF value above 5

Post by jmbecker »

The sample size sounds quite small. Maybe the correlation is also just spurious because of not enough observations. Maybe there are even more problems.
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
Post Reply