I am trying to estimate a threelevel model where I have students (level 1), classrooms (level2) and countries (level3). I have no variables measured at level 2 but I would like to use L2 cluster averages for level1 variables. In Type=TWOLEVEL this works with latent variable decomposition but this seems not to work for Type=THREELEVEL. So I used the cluster_mean function to compute cluster means. My question is how I can make sure that Mplus really computes the L2 cluster means and does not include L3 in some way. I used the following syntax:
VARIABLE: CLUSTER IS ctry idclass; BETWEEN = (idclass) classses ; DEFINE: classses = CLUSTER_MEAN (ses);
Which cluster does Mplus use to compute the cluster mean?