I recently purchased Mplus in order to fit a multilevel path model with ordered categorical variables both as a mediator and as an outcome variable. Individuals are clustered within countries.
What I would like to examine is the effect of both occupation and industry sector (measured at the individual level) on job security (and further on childbearing intentions). Because I have a large number of occupational groups (for now I have aggregated them in 25 groups but this may even be too coarse), it is cumbersome to treat them as fixed effects (and thus add 24 dummies with only about 650 individuals in total). Furthermore I hypothesize the effect to vary between countries which would require 24 random slopes.
Therefore I thought it might be feasible to include a random effect for occupation and possibly sector instead of a fixed effect and examine the variance components and EB estimates in more detail to see whether some significant variation of job security across occupations actually exists and whether this covaries with the random intercept at the country level. This, I think could be achieved by a cross-classified random effect model (please correct me if I'm wrong).
If I understand correctly, such models are not possible in Mplus 5.21 (yet)? Do you have any suggestions on how to work around this? Thanks already for your thoughts, Mieke
Thank you for your quick reply! Ideally I would also like to include at least 1 country-level variable as an explanatory variable. This is why I was considering multilevel path models to start with. Unfortunately it then looks like I will have to treat either occupation or country as a fixed variable in my model.