Modeling of interaction terms as normal constructs

[Pat79]

Hello,

A question concerning the integration of interaction terms in a normal path model.

I want to test an organisational concept based on the interaction of two subelements. The two subelements are measured both by formative constructs. To check the impact of the organisational concept I designed a model in which their is only the interaction term of the subelements, but not the subelements themselves (making the comparison to moderating effect, I let out predictor and moderator).
Is it feasible to model these "standalone" interaction constructs?

Would also be feasible to make an interaction term of the endogenous variables (in my case short and long term performance; by making the interaction term I would like to have a construct measuring combined scores)

If not what would be an alternative to these interaction approach?

Thanks a lot in advance for your help,

Patrick

[Diogenes]

HI,

firstly you have to save the scores of the LVs
then you could multiple these scores to have the interaction term (LV_score_preditor * LV_score_moderator).
You could do this before input the data to SmartPLS or do it directly as regression model in Excel, SPSS, SAS...

in viewtopic.php?t=479 I have explained some ways to have the LV scores computed.

I hope this help.

Bido