How to do a three way interaction effect in Smartpls 2?

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
Jacks
PLS Junior User
Posts: 7
Joined: Thu Mar 15, 2012 4:17 am
Real name and title:

How to do a three way interaction effect in Smartpls 2?

Post by Jacks »

Dear Smartpls follower,

I was trying to do a three-way interaction moderation effect analysis, could anyone please advise how to do it in Smartpls 2? It seems there are different ways of doing it but no sure which approaches are more commonly used? I would be much appreciated if you could advise the detailed procedures of doing it!

Many thanks!
Jacks
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: How to do a three way interaction effect in Smartpls 2?

Post by jmbecker »

The only valid way is to do the two-stage approach by hand using a separate tool like SPSS.
Copy the latent variable scores, calculate the product terms from the latent variables scores and then calculate a regression with the product terms and the predictor and moderator variables.
You cannot reimport product terms into SmartPLS, because all variables will be standardized in PLS and you should keep the product terms unstandardized.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Jacks
PLS Junior User
Posts: 7
Joined: Thu Mar 15, 2012 4:17 am
Real name and title:

Re: How to do a three way interaction effect in Smartpls 2?

Post by Jacks »

Thank you very much for your reply Jmbecker, it is very helpful! A quick follow up question, I read the book "A Primer On Partial Least Squares Structural Equation Modeling (PLS-SEM)" By Hair et al, and the two-stage approach explained in the book is different from what you suggested. Basically, it just used the latent variable scores as a single indicator for all the constructs (some as in the first stage) and created interaction effect based on it in the second stage, all the analyses were still done by using SmartPLS. Would it be Ok for me to do the three-way interaction effect following the same approach (e.g. use latent variable scores as a single indicator for both single and the two-way interaction constructs and then create the three-way interaction effect in the second stage in SmartPLS)? If yes, would you please advise if there is any literature supporting this approach of doing three-way interaction effect?

Thank you very much for your help again!

Jacks
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: How to do a three way interaction effect in Smartpls 2?

Post by jmbecker »

I think the follow up version of the book corrected the recommendation. It is not correct to use the product terms in SmartPLS as single-indicator constructs because of the standardization problem. The correct way is to use a software that can do regressions and allows for bootstrapping. SPSS can do this with some plugins (e.g., the process macro by Hayes).
SmartPLS 3 can do correct two-stage approach for two-way interactions and we are working on multiple-way interactions as well, but currently you have to do it separately, especially if you are using SmartPLS2.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Jacks
PLS Junior User
Posts: 7
Joined: Thu Mar 15, 2012 4:17 am
Real name and title:

Re: How to do a three way interaction effect in Smartpls 2?

Post by Jacks »

Thank you so much for your great help and clarification! I will try to run it in SPSS as per your recommendation.
Happy New Year to you as well!
Jacks
PLS Junior User
Posts: 7
Joined: Thu Mar 15, 2012 4:17 am
Real name and title:

Re: How to do a three way interaction effect in Smartpls 2?

Post by Jacks »

Hi Jmbecker, could you please help me with this follow-up question with 3-way interaction effect. My model has three dependent variables, if I use the two-stage approach as you advised (i.e. calculate product terms use the latent variable scores and then run the regression to test for the two and three-way interaction effects), then I would not be able to run the model with three dependent variables simultaneously in the second stage, if using SPSS. Would this have any impact on the robustness of the final results if I run it separately with each dependent variable? Or is there a better way of doing this?

Would you also please recommend me any literature that supporting the two-stage approach using pls.

Thank you heaps!!
Post Reply