I have a question about what to do with (dis)aggregated data when multilevel is not an option. More specifically, I measured individual achievement and then averaged these scores to create measures of school and class average achievement to test the so-called BFLP effect on self-concept. Because of the clustering of students in classes and schools we thought to use the multilevel or the complex function in Mplus. However, because of several reasons these techniques do not seem appropriate for our data (low ICC, few clusters). Still, we are working with variables at different levels. My question then is if, despite all reasons not to do it, multilevel analysis is the only solution for this model? Or are there other ways in Mplus to handle aggregated data?