How to check multicollinearity

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
indrawati
PLS User
Posts: 20
Joined: Tue Nov 30, 2010 3:56 am
Real name and title:
Location: Indonesia

How to check multicollinearity

Post by indrawati »

Dear all,

Multicollinearity is a problem in any regression, including in PLS?

How to check the existence of multicollinearity? how to solve it.

Thank you

Regards
User avatar
alialashwal
PLS Junior User
Posts: 8
Joined: Wed Jul 20, 2011 9:58 am
Real name and title: Alashwal, A.M.

Post by alialashwal »

As SmartPLS is not provided with this function, you can simply use SPSS to get the VIF (not more than 10 according to some references). A new PLS software can indicate multicollinearity called WarpPLS..

Regards,
Ali M. Alashwal, PhD
Faculty of Built Environment,
University of Malaya,
50603, Kuala Lumpur, Malaysia.
indrawati
PLS User
Posts: 20
Joined: Tue Nov 30, 2010 3:56 am
Real name and title:
Location: Indonesia

WarpPLS

Post by indrawati »

Dear Ali,

Thanks for the info. How to get Warp PLS?

I have check multicall through SPSS but I am not so sure with the process I did.

Can you send me the step?
I have 8 constructs with 53 items, each construct has 4 to 8 items. How to correlate the construct. Should I add the items of each construct and find the average of each construct than correlate the average of each constucts by using SPSS?

Thank you

Regards

iin
User avatar
alialashwal
PLS Junior User
Posts: 8
Joined: Wed Jul 20, 2011 9:58 am
Real name and title: Alashwal, A.M.

Post by alialashwal »

Dear iin,

Well, to conduct the test you go to Analyze - Regression - Linear. Then you assign one of the variables in one construct as a dependent variable then you assign the rest of variables (of the same construct) as independents. In statistic box don't forget to tick on collinearity diagnostic, then hit OK.. You will get the output and you can notice the value of VIF as well as the Eigenvalue. You have to do this process again and this time choose different dependent variable instead of the first one you used first and put it with the independents..

I think choosing whether you want to chick muticollinarty for one or all of your constructs together will depend on your objectives..

Ali
Ali M. Alashwal, PhD
Faculty of Built Environment,
University of Malaya,
50603, Kuala Lumpur, Malaysia.
Post Reply