I'm trying to estimate a 1-1-1 mediation on two-levels, with a moderator at level 2. While I successfully ran a similar model with all continuous predictors, I cannot find a solution to test the model with a categorical mediator.

I have worked with syntax in Mplus for this and in MLmed (an SPSS macro for multilevel mediation). MLmed runs the model without any extra settings beyond what I can find in the user manual, but I'm not sure the results are trustworthy. Any suggestion is more than welcome! I summarize the Mplus syntax for the model with continuous mediator. How could I adapt it to the categorical one?

Usevariables are religios Parentproud JustBeat Mrelig MParentproud; (religios = x, Parentproud = m, JustBeat = y, Mrelig = cluster-mean x, Mparentproud = cluster-mean m) WITHIN are religios Parentproud; BETWEEN are Mrelig MParentproud; Missing are all (-9999); Cluster is CountryID; Analysis: Estimator = Bayes; Type = twolevel; FBiteration = 10000; Processors = 2; Model: %Within% JustBeat on Parentproud (bW); JustBeat on religios; Parentproud on religios (aW); %Between% JustBeat on MParentproud (bB); JustBeat on Mrelig; MParentproud on Mrelig (aB); Model Constraint: New (IndW IndB); IndW = aW*bW; IndB = aB*bB; Output: Tech1 Tech8 standardized sampstat cinterval(hpd);

Thank you!!