In preparing for a research study I'm unsure about how many levels to plan on including my my model. This is research in an educational setting so multilevel modeling fits (the data are nested). My question is whether or not to differentiate between a classroom, teacher, or school in the model.
I will have students in each class. But a few classes may be taught by each teacher. And there may be a few teachers at each school.
My inclination is that because there are few classifications at these levels, and they are pretty much synonymous, I would combined them in the model. (I.e., the variance explained by what class a student is in will be only slightly different from the variance explained by what teacher they have so it isn't necessary to include both.)
Is my reasoning sound? Or are there other strategies/rules for determining when to include levels in the model?
I did see this question but my assumptions about the variance explained make mine different. I also saw this question but I'm committed to multilevel modeling.
