Page 1 of 1

Collinearity Statistic

Posted: Fri Dec 19, 2014 3:46 am
by alirezarousta
HI every body
As you now first step in structural model is testing for Collinearity issues. I am using PLS 3 and i saw that PLS 3 can calculate it directly and no need to use spss as we did before for PLS 2. it is true ? and how can interpret the result?

Re: Collinearity Statistic

Posted: Fri Dec 19, 2014 6:51 am
by jmbecker
Yes it is true. In SmartPLS 3 you get the collinearity statistics for the inner and outer model. No need for SPSS anymore. You interpret the VIF like you did in SPSS.

Re: Collinearity Statistic

Posted: Tue Dec 30, 2014 10:33 am
by j_terstriep
Dear All,

I was just wondering whether you are using R2 or the corrected R2 for the calculation of VIF values shown in the output of smartpls 3.0.

Many thanks in advance for your reply.

Best regards,
Judith

Re: Collinearity Statistic

Posted: Tue Jan 27, 2015 9:19 am
by saudah
Hello,

I am new to SmartPLS 3. As both inner and outer model's collinearity readings are given by the algorithm, I am not sure which one should I report. Based on my reading of the PLS-SEM book, I need to assess the collinearity among the predictor constructs, for every set of the subpart of the structural model.

Accordingly, I thought I should report the outer model's collinearity readings. But, when I looked at the output of the SmartPLS 3 , I am confused as the output provides individual VIF readings for each of the indicators. Whereas, if I calculated manually as required when using the SmartPLS 2, the readings are for the constructs rather than individual indicators.

Does this, means that I should just report the inner model's VIF readings for each of the latent construct included in the structural model?


Thanks and regards,
Saudah

Re: Collinearity Statistic

Posted: Tue Jan 27, 2015 10:55 am
by jmbecker
VIF statistics are important for the inner model and formative outer (measurement) models.

If you have only reflective measures, then using only the inner model VIFs is ok.
If you also have formative measures, then you should assess the VIFs of the indicators of the formative measures as well.

Generally, you get one VIF for each predictor in a regression. Hence, you have VIFs for all indicators in the outer model and for all constructs in the inner model. Sometimes people report only the highes VIF per regression (per dependent), yet the VIFs of the other predictors might be informative as well. I would always report all VIFs and not only the highest.

With regard to PLS you will find good information of VIF assessment in Hair, J. F., Hult, G. T. M., Ringle, C. M., and Sarstedt, M. 2014. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM), Sage: Thousand Oaks.

Re: Collinearity Statistic

Posted: Tue Feb 10, 2015 10:13 am
by saudah
Thank you very much Dr. Becker for the clear and helpful answer.

Now I realised that I got confused because I forgot about the need to assess VIF of formative constructs. My model contains only reflective constructs.
So, only the inner VIFs are relevant.
I will report all VIFs given in the output.

Thanks again.
Saudah

Re: Collinearity Statistic

Posted: Tue Jan 25, 2022 10:36 pm
by LFaridoon
Can we still retain the construct if VIF is 6 or 7?
And do we only check the multicollinearity between the predictors, what if there is a collinearity between one of the predictors and the dependent constructs? in my model, I have three predictors and one dependent construct, there is only collinearity between a predictor and the dependent construct, is that accepted?

Re: Collinearity Statistic

Posted: Wed Dec 21, 2022 2:23 pm
by jimciv
I have the same questions, in my case VIF 7.