Interaction for constructs embedded in nomological network

[saddas]

Dear all,

Suppoe you have the following model:

1) Exogenous formative construct A
2) Exogenous reflective construct B
3) Reflective Moderator M
4) Endogenous reflective construct D

Suppose I want to model the interaction effect between A and M on D. Now since one of the two main effects is formative, I understand I need to use the 2-step approach where in the 1st step the latent variable scores are obtained and then used in the 2nd step as indicators in the structural model. Now my question is: in order to model the interaction effect and report its significance and effect size (f2), do we do the 2-step approach with construct B in the model or not? In other words, construct B is not part of the main effects or interaction term, but it is another variable which happens to be in the overall nomological network.

Thanks in advance!

[Hengkov]

Hi Shamel,
Yes.
For effect size construct B you calculate with formula: R2 Inc-R2 Exc : 1-R2 Inc.
For effect size interaction: R2 with moderator-R2 without moderator : 1-R2 moderator.
Note: Evaluation outer model construct before save LV score.
Test Two-Stage Approach with construct B => simultan.
Regards,
Hengky

[saddas]

Thank you Hengky!

[lailakasem]

Hi Shamel,

I recommend you read this post, viewtopic.php?t=180&start=0&postdays=0& ... ardization

Basically, there is no fixed answer to this, Prof Ringle and Dr Stefan Behrens discuss this in detail in the post.

Good luck