I am trying to devise a model which predicts people's demand for more information.
The model I'm contemplating has 2 levels of predictors that are not nested in terms of participants (for example, students nested in classes and such), but nested conceptually.
Level 1: "information presentation": the way information is presented from an "information source" about a "target"
Level 2: "information source" & "target": people's a priori attitude toward the source of information presented to them, and the target that information refers to
So, the first level explains how the presentation method of information WITHIN a given dyad of information source-target will affect people's demand for further information...
I'm assuming that the influence of this presentation method on people's demand for more information would be different according to participants' pre-existing attitude toward the information source and target.
Thus, how information is presented, according to my assumption, is effectively nested under different information source-target dyads. I am wondering if this model structure is viable for multilevel analysis, where variables have nested structures not according to participants' membership in different groups, but purely from how associations of variables are tested.