20th May, 2021

Question

Asked 28th Jan, 2018

# Multicollinearity test in PLS-SEM?

My study model has two Independent variables, one dependent and one mediating .

How to assess collinearity in PLS-SEM?

## Most recent answer

in a reflective model, can the multicollinearity issues can be established by only seeing inner model VIF values?? My inner Vif values are showing no multicollinearity issue, while outer model has few multicollinearity issues. Please guide.

2 Recommendations

## Popular answers (1)

After running the pls algorithm, just take a look at the results. SmartPLS 3.0 gives you this information automatically. VIF values are under "Collinearity Statistics (VIF)". To interpret this information, read this thread:

4 Recommendations

## All Answers (10)

After running the pls algorithm, just take a look at the results. SmartPLS 3.0 gives you this information automatically. VIF values are under "Collinearity Statistics (VIF)". To interpret this information, read this thread:

4 Recommendations

In addition @Alejandro Ros-Galvez , evaluate the collinearity in the inner and outer (if formative measure) models with the same evaluation measures: generally we consider tolerance values below 0.2 or VIF above 5 as levels critical of collinearity.

In SmartPLS gives VIF automatically. But in the package plssem for R you have to calculate. But it is an easy solver. You could run regression with a latent variable (score) and calculate the VIF (see VIF function).

2 Recommendations

As Adonai said, this regression to obtain the VIF values can be performed with statistical softare, such as SPSS or Stata.

Regards

1 Recommendation

Thank you so much, Dear Alejandro and Adonai for answer my Question,

I have been assessed the collinearity through SPSS,

The tolerance result of three variables was (.59, .79, and .70), while the VIF was ( 1.67, 1.27, 1.47).

So, Is that enough justification to avoid the Multi-collinearity problem?

2 Recommendations

I think so. But see too condition index (CI). Consider CI values above 30 critical levels of collinearity. The CI < 30 indicated that the variables would not present collinearity problems if they stayed together (Gujarati, 2003).

Note. You can use VIF or TOL. TOL is merely the inverse of VIF; that is TOL = (1/VIF).

Ref.

Gujarati, D. N. (2003).

*Basic econometrics*. New York: McGraw-Hill2 Recommendations

VIF values you reported are under the threshold of 3, 5 or 10. Then, it seems there are no collinearity issues.

1 Recommendation

I am trying to check for multicollinearity in R but am not winning. I have 4 times for the depend variable and 10 items for the in-dependend variable. I am fitting PLS-SEM and using plspm package in R

1 Recommendation

PLS SEM is a non-parametric data analysis method with a small sample size such that normality and classical assumptions are not an absolute requirement such that they can be avoided. If the sample is large, it is reasonable to use SEM with strict attention to normality and classical assumptions.

1 Recommendation

## Similar questions and discussions

Multicollinearity issues: is a value less than 10 acceptable for VIF?

- Alejandro Ros-Gálvez

## Related Publications

Die dritte Auflage des bewährten Lehrbuches, gegenüber der Vorauflage verbessert und inhaltlich erweitert. Zielgruppe sind Studierende und Praktiker, die zumeist keine Mathematiker sind, wohl aber statistische Verfahren verstehen und angemessen einsetzen wollen. Vorzug des Buches ist somit, dass die Verfahren ohne mathematische Vorkenntnisse voraus...

The following chapter provides the results of the empirical data analysis with PLS. For performing a PLS parameter estimation a few software packages, most notably PLS-Graph, SmartPLS, PLS-GUI, and Spad-PLS are available. This research study uses SmartPLS 2.0 M3 for conducting the SEM analysis. This software has a graphical user interface and sever...