Which data for VIF calculation

[Markus R.]

Hi all together,

what data should I choose to use as input to SPSS for the VIF calculation? The Original Sample (O) or Sample Mean (M) calculated by SmartPLS for path coefficients and outer weights?

Thany you in advance,
Markus

[Hengkov]

Hi,

I create new formula for compute VIF manually in PLS. This is very easy.
VIF = 1/1-f^2, where f^2 is effect size.
For using another program like SPSS, I think if you model complicated (include moderating and intervening), this is not easy, because SPSS limited.

Best regards,

[Diogenes]

Hi,

If you want to compute VIF for formative indicators:
- use these indicators as independent variables in a multiple regression in SPSS and compute VIF there.

If you want to compute VIF for the structural model:
- run the model in SmartPLS (PLS algorithm)
- report
- copy “PLS / Calculation results / Latent Variable Scores”
- paste them to SPSS
- use these scores as independent variables in a multiple regression in SPSS and compute VIF.

The dependent variable could be any variable in the both cases.

Best regards,

Bido

[Hengkov]

Hi,

SPSS only for simple model research for compute VIF. But, if you cases include many variable (moderating and intervening), SPSS cannot used. I recommend for used WarpPLS software or compute by hand with my formula above.

Regards,

[Markus R.]

Hello all together,

first of all, I want to thank you for your solutions. But both of your solutions aim for the VIF of variables, but I want to get the VIF of indicators to evalute my formative measure modell. I am sorry, I should have made a more focused question.
So please, can you tell me how to reach these VIFs?

[Hengkov]

Hi,

Problem collinearity not only in measurement variables (formative indicators), but also in structural models. So, you must check all together (reference you see Kock and Lynn article 2012 in JAIS).

Regards,

[Diogenes]

Hi,

As I said, if you want to compute VIF for formative indicators:
- use these indicators as independent variables in a multiple regression in SPSS and compute VIF there.

More detailed:
1) Open your dataset in SPSS
2) Analyze / Regression / Linear /
3) Use as dependent variable any variable (could be the case number)
Use as independent variables all formative indicators of the same LV.
4) Statistics / Collinearity diagnostics / Continue / Ok
5) In the output, see the Coefficients table --> Tolerance and VIF
Of all indicators.

Best regards,

Bido

[FHH]

In the PLS report I get negative LV scores!

Can I still use these scores to test for collinearity among the LVs using VIF?
Or are all conclusions drawn from negative LVs scores wrong?
Is there a way to tranform/standardize the negative LV scores to make them interpretable?

[cringle]

Hi,

The LV scores are standardized and thus there are postive and negative values. Yes, use these scores to test collinearity in the structural model. We also explain things in the PLS-SEM book: http://www.sagepub.com/books/Book237345

Best regards,
Christian Ringle