Dear,

We are wondering if it makes sense to allow random slopes for second level variables in a twolevel model?
Mixed in stata estimates these models (although it sometimes has trouble estimating correct SE's) but I'm not sure if this actually makes sense.
For example, students are nested within schools. It is possible to introduce random slopes only for a schoollevel variable such as schoolexpenditure?
In all the examples we have seen random slopes always concerns a level 1 variable and the focus is on the idea that level 1 variables can have different effects in each group.

So the question is basically: does it make sense in a twolevel model to introduce random slopes only for a level two variable?
If it does, could anybody suggest any literature how we can understand/conceptualize/visualize this? Since school-level variables are constant within each school I don't see how random slopes can make sense?


Kind regards,
Filip