VIF value above 5

Frequently asked questions about PLS path modeling.
Post Reply
neethu856
PLS Junior User
Posts: 5
Joined: Wed Jan 16, 2019 12:27 pm
Real name and title: Nethu M

VIF value above 5

Post by neethu856 » Tue Jan 29, 2019 2:34 am

Hi

When I am running the model using SmartPLS 3.0, I am getting 5 out of 10 items VIF value above 5. I am not quite clear how to deal with this like grouping it into composite indicators (where can I get the correlations?, whether PCA can be done with SmartPLS 3.0).

Content validity proves that all items are distinct. Higher ordering as suggested by Hair et.aI tried was not working. Whether I have to remove correlated indicators iteratively?

Could someone make it clear how to deal with this high VIF value?

jmbecker
SmartPLS Developer
Posts: 971
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF value above 5

Post by jmbecker » Thu Jan 31, 2019 12:25 pm

From what you say I assume you are talking about outer model VIF.
Are your items formative or reflective? If they are reflective you do not need to worry about VIF. In fact, you want high VIFs there as this indicates high correlations and therefore supposedly high reliability.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

neethu856
PLS Junior User
Posts: 5
Joined: Wed Jan 16, 2019 12:27 pm
Real name and title: Nethu M

Re: VIF value above 5

Post by neethu856 » Thu Jan 31, 2019 10:07 pm

Yes, outer model VIF. Formative constructs I got high VIF above 5.

jmbecker
SmartPLS Developer
Posts: 971
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF value above 5

Post by jmbecker » Mon Feb 11, 2019 1:53 pm

Why do these formative indicators correlate so heavily? Is it a conceptual issue or could it be a data quality issue (e.g., people not paying attention to the questions and just marking all items in a block consistently)?
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

neethu856
PLS Junior User
Posts: 5
Joined: Wed Jan 16, 2019 12:27 pm
Real name and title: Nethu M

Re: VIF value above 5

Post by neethu856 » Mon Feb 11, 2019 10:49 pm

Yes, I think its a data quality issue. I have actually bought the participants from a third party recruitment company.

jmbecker
SmartPLS Developer
Posts: 971
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF value above 5

Post by jmbecker » Wed Feb 13, 2019 2:53 pm

You might need to exclude some of your respondents (and/or renegotiate with your data provider). I usually screen respondents of answers that are less than half of the median (speeders) and those answering more than 90% of the time in the same response category (straighliners).

You may also include trap questions (where respondents should give a specific answer) even though they are heavily debated.

Yet, that all will not help you with your current problem. You either have to life with the low data quality of get new data.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

neethu856
PLS Junior User
Posts: 5
Joined: Wed Jan 16, 2019 12:27 pm
Real name and title: Nethu M

Re: VIF value above 5

Post by neethu856 » Wed Feb 13, 2019 10:20 pm

Thank you very much for your reply. I will do that.

Post Reply