I am working with a multi-level dataset that is akin to a repeated-measures design with no time component. Each subject (n=134) rated a set of 25 attributes (e.g., concentration, mood) on 7 continuous variables (e.g., fundamentalness to identity, desirability). Subjects also indicated whether they would take a pill to improve each attribute (definitely would not, probably would not, probably would, definitely would).
I am struggling to find examples of multi-level models that do not use traditional nesting (e.g., students within classes within schools). The 25 attributes could be considered repeated measures of the IVs and DVs; however longitudinal models do not seem appropriate because the attributes are not ordered.
Ideally, I’d like to use a latent class/profile analysis approach to identify subgroups of the 25 attributes that share common features (based on the continuous variables) that predict the DV, willingness to take a pill to improve the attribute. I’d also like to estimate and compare the portions of variance in the DV attributable to the subject doing the rating versus the attribute being rated.
I apologize for the vague question. I’m hoping for some direction regarding how to set up the analyses (e.g., would you consider attribute level-1 and participant level-2, or vice versa?), or perhaps recommended reading/examples.