I'm just resubmitting a manuscript for second review, which uses mplus for multilevel modeling with LVs. We have 2 sets of data which are unit managers nested in geographical regional units and customers who deal with different units within the region. Because we don't have a 1-1 dyad match and because customers may use more than one unit in the region, we have aggregated the unit managers to the geographical regional unit level and then matched customer responses to the geographical unit. So we have a multilevel SEM that evaluates all of the relationships.
One reviewer, who I think may be somewhat unfamiliar with multilevel techniques is asking what the unit of analysis is, which of course gets a bit complex in this case. Is it accurate to say that the unit of investigation is the 1) the individual unit managers and 2) the individual customers and therefore we have 2 units of analysis (macro and micro), those being 1) the aggregated regional unit and 2) the individual customers matched to those units. Or, since it is all modeled in one multilevel SEM is better to say that the unit of analysis is the path relationships between the aggregated managers to individual customers as represented by the latent variables.
I understand the term unit of analysis to be the number of independent observations in an analysis. If students are nested within classrooms, then classrooms would be the unit of analysis. Others may define this term differently. Perhaps you need to ask the reviewer exactly what he/she means.