Control variable and multi group analysis

Frequently asked questions about PLS path modeling.
Post Reply
luchow414
PLS Junior User
Posts: 4
Joined: Fri Jun 08, 2018 7:12 pm
Real name and title: Yuko Leo Assistant Professor

Control variable and multi group analysis

Post by luchow414 » Fri Jun 08, 2018 7:18 pm

Dear all,
I have a set of control variable (e.g. age, gender and education) in my PLS model. Can I use the same set of control variables (age, gender and education) to do multi group analysis (MGA) as a post-hoc analysis? If yes, what would be the major differences when it comes to result interpretation. Many thanks!

jmbecker
SmartPLS Developer
Posts: 821
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Control variable and multi group analysis

Post by jmbecker » Sun Jun 10, 2018 8:08 pm

First, you should not do both at the same time. If you have your grouping variables in the model the estimation will likely fail, because the variable may have only limited (zero) variance after splitting into groups (especially with variables such as gender).

The interpretation between control and grouping variable is also quite different.
If you have a variable as a control variable in your model, you are interested in the effect of your predictors beyond the effect of the control (i.e., for all observations equal on age and gender etc., what is the effect of X on Y). That is the very common interpretation of multiple regression coefficients.
If you do a multi group analysis it is more like a moderator analysis: If gender is male what is the effect of X on Y and if gender is female what is the effect of X on Y. Thus, you are interested in the change of X on Y due to gender, age, etc.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

luchow414
PLS Junior User
Posts: 4
Joined: Fri Jun 08, 2018 7:12 pm
Real name and title: Yuko Leo Assistant Professor

Re: Control variable and multi group analysis

Post by luchow414 » Mon Jun 11, 2018 3:54 am

Dr. Jan-Michael Becker,
Thank you very much for your quick reply!
A follow-up question: After I analyze the original PLS path model, I can still use the control variables (such as age and gender) to do a multi-group analysis, right? In Hair, Ringle and Sarstedt (2013), they stated that "When the effect of control variables are significant, the researcher should use this finding especially carefully when drawing conclusions or initiating ... PLS-SEM multigroup analyses." I am not quite sure what they meant here. What exactly should we be careful about when using the control variable in MGA?
Again, thanks you very much!

jmbecker
SmartPLS Developer
Posts: 821
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Control variable and multi group analysis

Post by jmbecker » Mon Jun 11, 2018 8:00 am

I don't know what they might have in mind when writing this.

I think there are two different aspects when thinking about control variables and MGA:
1) If your controls are not the grouping variables, you also want them in the MGA models. Because the control can also be important at the group level.
2) If your controls should serve as grouping variables you might have to exclude them from the model due to statistical reasons. Then you cannot control for them at the group level anymore. An alternative is moderation analysis where you introduce your control as a continous moderator for some relations.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

Post Reply