1. VIF > 10: Multicollinearity Problem

Interpretation:

Variance Inflation Factor (VIF) values above 10 suggest serious multicollinearity among your indicators or constructs. In SmartPLS, this typically applies to:

Indicator VIF (within constructs) → for reflective models

Lateral collinearity (between constructs) → especially important when assessing structural paths

Why It’s Problematic:

High VIF implies that the variable shares too much variance with others, meaning it's not providing unique information — this distorts path coefficients and weakens interpretability.

How to Justify or Fix:

Justify only if the constructs are theoretically expected to be closely related (e.g., in a higher-order model).

Otherwise, you must address it:

Remove or combine collinear indicators.

If it’s lateral collinearity, consider creating a second-order construct (Formative-Reflective or Reflective-Reflective) to account for shared variance.

Hi Amal! You should try to solve the collinearity and discriminant validity issues. In the case of reflective constructs you must assess the indicators loads and VIF values and see if there is one or more indicators that generate these problems and if there is the possibility to delete them. This can be the case with indicators that have very high loads (> 0.90). As Judit Albert said, higher HTMT values are accepted for constructs conceptually related, the maximum threshold in this case is of 0.90, so you should also check that. Using a formative measurement model is another possibility, but only if the constructs are defined as formative ones in the theory. The last solution and the hardest one is to increase sample size.

Thank you very much Vasilica Maria Margalina and Judit Albert

the constructs i'm using are reflective. Judit Albert can you please explain more "Justify only if the constructs are theoretically expected to be closely related". My focus is set on finding relevant references to argument and justify the high values of 2 constructs of my model. Thanks