VIF for Type II second order constructs

BaerImPulli
PLS Junior User
Posts: 3
Joined: Fri Dec 18, 2015 9:39 am
Real name and title: O. Purner, B.A.

VIF for Type II second order constructs

Hello all,

I have a Problem to finding the VIF value for my second order factor type II.

I have a model in which there are a few second order constructs type II (reflective first order, formative second order). And I used the two stage approach in SmartPLS to calculate the model. At first I calculated the model with the reflective first order constructs and then I used the latent variable scores for the second calculation. (I did it just like James Gaskin showed in his tutorial https://youtu.be/kPeUTKjMF7o?list=PLnMJ ... wLKkAFBUCy).

The Problem is: I wanted to check the multicollinearity of the indicators (the refklective first order factors) for the second order formative constructs, but I don't know where I can find the correct VIF. Due to the two stage approach, I do have two different SmartPLS Reports: one with the reflective first order constructs integrated in the model and one without the indicators, but the latent variable scores as indicator for the formative second order factor.

In the first SmartPLS report, the outer VIF values just show the reflective indicators, but that isnt what I want, is it? I actually need the VIF for the whole reflective first order factors that act as the indicators for the second order formative construct. I do have values in the inner VIF values section in the report, but I am not sure if they are correct, since the VIFs of the particular first order factors are identical (for example: first order construct 1 = 1,002 and first order construct 2 = 1,002) and the inner VIF values are just for testing the inner model.

In the second SmartPLS report the outer VIF value is, due to the fact, that i used only one latent variable score, exactly 1.0. So this isn't correct for sure.

It would be so great, if anyone knew the answer to this problem and could help me solve it!

Thank you very much!

Kind regards,
Oliver

jmbecker
SmartPLS Developer
Posts: 952
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF for Type II second order constructs

Well the approach described in the video is not correct. The correct two-stage procedure uses the latent variables scores from the first-order constructs as formative indicators of the second-order construct. Not only a single-indicator. With that procedure, you should also obtain VIF values.
Another option is to simply use the repeated-indicator approach (but with formative / mode B operationalization at the second-order level) and model the influence of the antecedents through the first-order constructs assessing the total effect.

Those options are described in:
Becker, Jan-Michael, Kristina Klein and Martin Wetzels (2012): Hierarchical Latent Variable Models in PLS-SEM: Guidelines for Using Reflective-Formative Type Models, Long Range Planning, Vol. 45, Issue 5-6, 359-394. http://dx.doi.org/10.1016/j.lrp.2012.10.001
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf

BaerImPulli
PLS Junior User
Posts: 3
Joined: Fri Dec 18, 2015 9:39 am
Real name and title: O. Purner, B.A.

Re: VIF for Type II second order constructs

Dear god, thank you very much Mr. Becker!!! Good to know the approach was wrong.

So I follow your option in the paper and combine the repeated indicator approach and the sequential latent variable score method. I do have a formative second order construct with two reflective first order constructs, each with two indicators.

1) I model the two latent variables with their reflective indicators and then model a second order latent variable with a formative path from the two reflective first order constructs. Now i have to operationalize this second order construct with all indicators of the two first order constructs (4 indicators altogether). But with what operationalization? reflective like the first order constructs or formative?

2) I calculate the model and then use the latent variable scores of the first order constructs and operationalize them formative to the second order construct.

Would this be correct? Or did i misunderstand something in this procedure?

Thank you very very much for your help!

jmbecker
SmartPLS Developer
Posts: 952
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: VIF for Type II second order constructs

1) Formative.

2) Yes, you could do that if you want to do the two-stage approach.

OR, you could just use the repeated-indicator approach. We describe situations in the paper that are more suitable to the one or the other.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf

BaerImPulli
PLS Junior User
Posts: 3
Joined: Fri Dec 18, 2015 9:39 am
Real name and title: O. Purner, B.A.

Re: VIF for Type II second order constructs

Thank you very much! I'll do it with the two stages since I am now more familiar with that!

You helped me very much Mr. Becker!

shash
PLS Junior User
Posts: 5
Joined: Sat Nov 25, 2017 2:29 pm
Real name and title: Ms. Shasheela Devi

Re: VIF for Type II second order constructs

Hi,

I have the similar issue on the multicollinearity for my Formative Dependent Variable which is 2nd order construct. Just to double confirm, i need to use the original indicators for the other constructs/variables and just use the LV from executing the PLS algorithm for the 2nd order DV/construct (of course not the single indicator as Gaskin's video).
This is what i did, and my VIF output seems to be ok. I need to confirm if my method is correct. Please confirm. Thanks.