Modeling moderating effects with formative indicators

Frequently asked questions about PLS path modeling.
Post Reply
u-ing
PLS Junior User
Posts: 7
Joined: Mon Nov 28, 2005 10:09 am
Real name and title:

Modeling moderating effects with formative indicators

Post by u-ing »

There are a lot of posts in this forum about moderating effects with reflective indicators or with mixed measurement models, but what about formative indicators?

I've read the paper of Chin/Marcolin/Newstead about modelling moderating effects with reflective indicators and there is a hint on building moderator variables with formative indicators in the last chapter, but I'm not sure if I understand that correctly.

It says, that you should use PLS to create underlying construct scores for the predictor and moderator variables and then take those single composite scores to create a single interaction term. Does that mean you just export the latent variable scores of predictor LV and moderator LV to Excel, multiply the scores, get the data back to smartPLS and implement the Moderator as LV with one Indicator?

Would be a great thing if someone could help me here. Maybe anyone even knows some literature about Moderators with formative indicators, all I found was about reflective models.

Thanks in advance and have a happy new year!

Sebastian
User avatar
cringle
SmartPLS Developer
Posts: 821
Joined: Tue Sep 20, 2005 9:13 am
Real name and title: Prof. Dr. Christian M. Ringle
Location: Hamburg (Germany)
Contact:

Post by cringle »

Hi,

I do usually do not include moderating effect in PLS path models. However, Chin provides at hint in the following document (slide 52):

http://disc-nt.cba.uh.edu/chin/icis2000plstalk.pdf

He recommends to test this approach in Monte Carlo simulation. In addition to that, I suggest to analyse the correctness of such an analysis from a (formal) statistical point of view.

Best
Christian
User avatar
joerghenseler
PLS Expert User
Posts: 40
Joined: Fri Oct 14, 2005 9:59 am
Real name and title:

Re: Modeling moderating effects with formative indicators

Post by joerghenseler »

u-ing wrote:It says, that you should use PLS to create underlying construct scores for the predictor and moderator variables and then take those single composite scores to create a single interaction term. Does that mean you just export the latent variable scores of predictor LV and moderator LV to Excel, multiply the scores, ...
Up to here it's definitely a reasonable approach.
u-ing wrote: ... get the data back to smartPLS and implement the Moderator as LV with one Indicator?
You can do that. It's what Eggert/Helm recommend in their contribution to "Bliemel/Eggert/Fassott/Henseler: Handbuch PLS-Pfadmodellierung".
However, I would recommend to use SPSS (or, if you prefer, the analysis tool within excel), take the factor scores as estimates of the variables you are looking at, and do the regressions you need.
Post Reply