I have data from a randomized controlled trial with two treatments (Group A and B), one control group (Group C) and four measurement points with two mediators and two outcomes. Mediators and outcomes are on the within-level, while the predictors are two-dummy coded-variables, one for each treatment, that are on the between-level. The effects of Group A are hypothesized to work via m1 on y1 and y2; the effects of Group B are hypothesized to work via m2 on y1 and y2. I am interested in the differences between A and C, and B and C, but not between A and B.

My question is, when modelling a 2-1-1 mediation in Mplus (syntax from Preacher, Zyphur, & Zhang, 2010) do I drop all cases from Group B when comparing A to C, or should I model all groups at the same time and regress both dummys on both mediators which are then regressed on the two outcomes? The goal was to conduct Monte Carlo Simulations for the indirect effects, but running the three groups together resulted in wider confidence intervals for the indirect effects than when modeling A vs C and B vs C with dropping cases. Thank you!



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.