Unrealistic high VIF values in formative SEM

[burton9887]

Deal all,

as part of my PhD project I am creating a new formative scale.
Right now I am testing a pure formative single-item vesion (see exhibit SEM) vs. a reflective-formative multi-item version.

SEM

Messmodell.jpg (104.78 KiB) Viewed 3328 times

The outer VIF values in the single-item version (see ehibit VIF) seem to be unrealistic high.

VIF values - realistic!?

VIF.jpg (128.79 KiB) Viewed 3328 times

Could they be truth? What might be the reason for it? Or coud it be a software bug? I do not know!

Looking for your comments, ideas...
Thank you.

Best wishes
Felix

[jmbecker]

Hi Felix,
could you also provide the correlation matrix of your indicators? That would give clarity if VIFs are realistic or subject to a calculation problem.
It seems that there are two pairs of indicators which are almost identical in terms of covariation.

In addition: Where is the reflective part of your "reflective-formative multi-item version"??

[burton9887]

I solved the problem.
It was a problem regarding missing data replaced by mean in European number format - but i did not change the SmartPLS settings from US to European number format. :-)

Thank you!

Felix