Different values for main effect with same variables in MedMod

[::SWS::]

Newbie in MedMod question :

I've been using MedMod in Jamovi for moderation analyses and I have a question about the calculation of the main effect coefficient. I conducted 3 moderation analyses on identical IV > DV, replicating the analysis and simply changing the moderator variable. Even if there is no significant moderation effect in my context, I find it very surprising that the results are different for the main effect line in each analysis (see attached screenshots). Shouldn't this value remain constant across the 3 analyses since it is supposed to be independent of any moderator variable ? Your insights would be of great help in better understanding how moderation analysis works.

[mcfanda@gmail.com]

Hi
There results you get from medmod make perfectly sense. In any linear model with interactions, the "main" effects are the effect of one variable estimated for the other variable (the moderator) equal to zero. Medmod centers the variables to their means, so we can interpret the effect of one variable as estimated for the average value of the moderator.

When you change moderator, you change also the exact level for which the main effect of the independent variable is estimated, so the main effect may change.

[::SWS::]

Hello Marcello,
Many thanks for your answer. I have a better underdstanding of linear models with interactions now.