Fornell-Larcker Criterion

Frequently asked questions about PLS path modeling.
Post Reply
Catharsis
PLS Junior User
Posts: 7
Joined: Mon Mar 04, 2019 11:52 am
Real name and title: Human Kohzadi Student

Fornell-Larcker Criterion

Post by Catharsis » Tue Mar 12, 2019 2:34 pm

Hello,

I am currently investigating some reflective research models with SmartPLS and had a few questions about checking the measurement model, specifically discriminant validity.

I would like to know if my Fornell-Larcker criterion is met. I'm not sure about the FC construct because the value on the diagonal is not greater than the values in the corresponding row. According to my assumption this should be so. Would be very grateful about clarification. Maybe I just misunderstood something.


Best regards!
Attachments
Bildschirmfoto 2019-03-12 um 15.29.04.png
Bildschirmfoto 2019-03-12 um 15.29.04.png (125.08 KiB) Viewed 107 times

jmbecker
SmartPLS Developer
Posts: 982
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Fornell-Larcker Criterion

Post by jmbecker » Tue Mar 12, 2019 8:59 pm

Fornell-Larcker criterion says that you diagonal value should be larger than all values in the same row and column. Thus, your EE and FC constructs are not distrinct based on Fornell-Larcker.
In addition, you should also consider HTMT criterion (smaller than 0.85) which is more sensitive to discriminant validity problems than Fornell-Larcker and the current state-of-the-art in PLS-SEM research.
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

Catharsis
PLS Junior User
Posts: 7
Joined: Mon Mar 04, 2019 11:52 am
Real name and title: Human Kohzadi Student

Re: Fornell-Larcker Criterion

Post by Catharsis » Wed Mar 13, 2019 6:35 am

Hello Dr. Becker,

thanks for the answer. I followed your instructions, and it actually looks like there's a discriminant validity problem.

My question now would be how to solve the problem. Should I discard the weakest item from the construct (FC4) and recalculate the PLS algorithm? Or do you remove the whole construct directly in such cases? In this case FC? I have discarded the FC4 item as a test. The Fornell-Larcker criterion is now fulfilled. But the HTMT value does not improve.
Attachments
Bildschirmfoto 2019-03-13 um 07.25.40.png
Bildschirmfoto 2019-03-13 um 07.25.40.png (110.93 KiB) Viewed 98 times

Catharsis
PLS Junior User
Posts: 7
Joined: Mon Mar 04, 2019 11:52 am
Real name and title: Human Kohzadi Student

Re: Fornell-Larcker Criterion

Post by Catharsis » Wed Mar 13, 2019 6:53 am

...
Attachments
Bildschirmfoto 2019-03-13 um 07.26.03.png
Bildschirmfoto 2019-03-13 um 07.26.03.png (240.93 KiB) Viewed 95 times

jmbecker
SmartPLS Developer
Posts: 982
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Fornell-Larcker Criterion

Post by jmbecker » Wed Mar 13, 2019 8:28 am

There are multiple ways that could solve a discriminant validity problem. Pleas search the forum and the literature for more information on that issue. It would need quite a lengthy answer to provide all possible options.
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

Post Reply