Contiuous moderating effect in SmartPLS 3

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
drege
PLS Junior User
Posts: 1
Joined: Tue Mar 08, 2016 11:49 am
Real name and title: Thomas Remieg

Contiuous moderating effect in SmartPLS 3

Post by drege »

Hi all,

I am currently conducting an analysis for my master thesis and have a few problems with the moderating effect in SmartPLS 3.
Regarding the context: I gathered the data from a survey conducted among students. All of the answers are based on a 7-point Likert Scale. Meaning that the smallest value is 1 and the highest 7.
And here already comes the first question: Did I understand it right that if there are no zero values in the moderating variable then one should use the mean-centered product term generation? At least that is how I understood the advice in "A primer on partial least squares structural equation modeling" on page 260.

Second: In this particular setting I am talking about my moderating and independent variable are both reflectively measured, but my dependent variable is formatively measured. I tried both calculation methods of the moderating effect, namely two-stage and product indicator, and both yield different results. When I use the two-stage approach with mean-centered product term generation, the moderating- as well as the direct effect are significant. But using the product indicator approach they are not anymore. Is using the two-stage approach a valid move here? As understood in the aforementioned book you can use this calculation method with reflective and formative indicators, but in the original article from Henseler and Fassot 2010, they recommend using product-indicator when the independent and moderating variable are both reflectively measured?
And thirdly: the book says that testing the hypothesis whether the independent variable has an effect on the dependent variable (main effect) should be done without the moderating effect. In my case this link is significant. But after incorporating the moderating effect, do I then still use the path coefficient of the model without the moderating effect (so from the main effect) to describe the effect of my independent on dependent variable or is then the path coefficient with the moderating effect incorporated the one to go respectively the one to present in the results section of my thesis?

Hope my explanations are not too confusing! And sorry for the long post.
But thanks in advance for your help. Looking forward to the answers!!

Best,

drege
Post Reply