empisoz posted on Wednesday, August 14, 2019 - 3:59 am
for a 2-2-1 twolevel mediation model I'd like to examine the presence of am exposure-mediator interaction using Mplus. Exposure-mediator interaction models have beeen described e.g. in the 'causal mediation' literature, but I really wonder whether this approach has already been applied to twolevel data? Of course, Mplus code examples would be very helpful -
I am not sure if there is literature on that. On the surface of it, it seems straightforward to do on the Between level what we do in the single-level case for this X*M interaction. It's just that Y is a latent continuous variable in the twolevel case - it is the random mean/intercept of the outcome (obtained from the decomposition of Y into Within and Between level parts). You can check if the 2 Preacher et al pdf's on our Mediation page contain such a model. On that page, you also find the single-level X*M writings we have done in our 2015 article and 2016 book.