Which VIF values to report?

[Avneet]

Greetings! 

Please help me resolve the following queries:

1. In the case of formative constructs, which VIF values should be less than 3.3? Are the inner VIF values reported or outer VIF values? 

2. Can the threshold limit be kept as VIF  < 5 or it should only be kept as VIF < 3.3? 

Thank you! 

--

Regards

Avneet Kaur Puri

Junior Research Scholar 

University Business School

Panjab University

Chandigarh

[Neeraj Kaushik]

Dear Avneet

There are 2 terms in SEM:

Inner Model: The structural relationship of constructs

Outer Model: A construct with its statements

An Inner Model is also known as Path Analysis or the simple Regression analysis between constructs.

I made a demo model where Quality is a Formative construct. 

Quality and Satisfaction are IDVs for Loyalty.

Here, for the Formative construct, we shd report the VIF of the Outer Model

And for checking the Multicollinearity between the Quality and Satisfaction, we shd report the VIF values of Inner model.

Now feel plz to ask for any of your doubts.

Best wishes

Neeraj

--
The members of this group are expected to follow the following Protocols:
1. Please search previous posts in the group before posting the question.
2. Don't write the query in someone's post. Always use the option of New topic for the new question. You can do this by writing to dataanaly...@googlegroups.com
3. It’s better to give a proper subject to your post/query. It'll help others while searching.
4. Never write Open-ended queries. This group intends to help research scholars, NOT TO WORK FOR THEM.
5. Never write words like URGENT in your posts. People will help when they are free.
6. Never upload any information about National Seminars/Conferences. Send such information
in personal emails and feel free to share any RESEARCH-related information.
7. No Happy New Year, Happy Diwali, Happy Holi, Happy Birthday, Happy Anniversary, etc. allowed in this group.
8. Asking or sharing Research Papers is NOT ALLOWED.
9. You can share your questionnaire only once.
---
You received this message because you are subscribed to the Google Groups "DataAnalysis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataanalysistrai...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAN%2BNa9zsNq9d7uPn5GMEVt3DB6%3DaFJZNp8oY64hNW3ObmefByg%40mail.gmail.com.

[kritika gupta]

Sir In case of model with control variable how to report vif inner?

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAAd%3Dc8Oo%3DcZFfzO%3DyQ-stBcw4s-oh5neLeAz51EhkW2ynWMU7A%40mail.gmail.com.

[Avneet]

Respected Sir,

Thank you for the illustrative explanation. However, my query is if the VIF values for the formative constructs are supposed to be reported for the outer model i.e. statements to construct does that imply the responses collected from the respondents need to vary among themselves?  

For example:

A formative construct 'S' consists of 6 statements from S1 TO S6. 

The Issue of Multicollinearity is bound to arise since the responses received are similar and do not differ from each other. Wouldn't this affect the reliability/validity of the construct? Also, are VIF values above 3 also acceptable?

Thank you!

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAG3apP_8B7p%2BRUMH7ggptXKSSTUXVFqKei1Hizk_8xCmoZz%3DWg%40mail.gmail.com.

[Archana Mittal]

And sir, if the model is reflective-reflective in PLS-SEM, then which VIF values are to be reported, the Inner or outer model's? 

Thanks & Regards

Archana Mittal

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAAd%3Dc8Oo%3DcZFfzO%3DyQ-stBcw4s-oh5neLeAz51EhkW2ynWMU7A%40mail.gmail.com.

[Neeraj Kaushik]

Dear Avneet

Looking at the data, it does not seem to be a formative construct data.

If the respondents have replied so similarly to all statements, it appears to them all statements are the same.

Plz note high VIF value is again indicating the same that the responses are very similar, which rarely happens in formative constructs. Further, VIF<3 is good (denoted by green colour in SmartPLS), VIF between 3-5 is a concern  (denoted by black colour in SmartPLS)and VIF>5 is a problem (denoted by red colour in SmartPLS)

So, you are correct in your observations. Here we shd check whether we truly have a reflective or formative construct.

May refer https://www.youtube.com/watch?v=GXyZJCkHKQ4&list=PLzUJUtTJcj8SS71HovHrU2HIjoRW7L4sm&index=7&t=208s

Best wishes

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAN%2BNa9xchHiUCAvWOP_qKdO-MnAoeeekuW-edkovuwWiD7%3DLzA%40mail.gmail.com.

[Neeraj Kaushik]

Dear Archana

As I mentioned earlier, for the measurement models we refer to the outer model and for the structural model we refer to the inner model/

As regards the higher order construct, you shd use the two-stage approach (Stage-1 save the scores of LOCs and in Stage-2 use those LOC scores as items for making HOC) and here too you''ll refer to the outer model for HOC.

Best wishes 

 

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAAMvjqxbp6LhGr-z5EgXjHzeJxnEgnC0g8LtSh-FsLLYL%2ByybA%40mail.gmail.com.

[kritika gupta]

Sir In case of model with control variable how to report vif inner?

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAAd%3Dc8OL%2Bp6QeFgeHMS5-w63%2BKqa9LD1zvai%2BjS8OP0Uo6a1QQ%40mail.gmail.com.

[Neeraj Kaushik]

If the control variable is measured on the scale, then it should also be checked for validity and reliability

Then report the usual VIF values as given by model analysis.

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAG3apP9T%3DwT98QfEKxx5-c2YAxaD27721JZd1tRo2KVg74%3DkWA%40mail.gmail.com.

[Rinky Yadav]

Sir in my model vif value of 1 item is 5.276. should I remove it? And if yes then i need to report new values?

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CAAd%3Dc8Ny4e9peTY80TEpqazCgDGHnP8L%2BDVMg95%2BDw4%2BNmiy7A%40mail.gmail.com.

[Muhammad Rizqi Siregar]

Dear Rinky Yadav,

Are they key variables? Is the measurement of their effects part of your research goal? If the answers are resounding no, you can ignore large VIF.  If you work with large samples, VIF can be relaxed too.

I'd suggest that you focus on the effect estimates, the standard errors and confidence interval around those effects rather than taking "acceptable" VIF values (less than or equal 2.5, less than 3.3, or less than 5).

Best,
MRS

To view this discussion on the web visit https://groups.google.com/d/msgid/dataanalysistraining/CADdRJB-xfDKju3RTb4t7CiPZ082BMm0-15CFaoqzMEzdE0%3D7Sw%40mail.gmail.com.