www.cmm.bristol.ac.uk/forum

Just want to make sure I'm setting up REALCOM correctly. Say I have a mixed model with an outcome y and a location variable, site, under which individual patients are clustered. I'm treating site as random, and as the subject in the analysis. As such, I will declare it as a level 1+2 variable in my set up. Now, assume I also have a continuous variable that I wish to consider as random representing a patient characteristic (e.g., weight, temperature, etc.). I want to treat this variable as a random slope which is the same as saying that it is random at level 1 (it describes the patient, not the site). So, I would declare this as a level 1 variable in my set up, correct? My goal would be to conduct an analysis with random intercepts and slopes. I think this is the correct analysis if site is both random and the subject, and the covariate describing level 1 units is also random.

As this is a slightly conceptual question you might want to ask it on the missing data forum (https://groups.google.com/forum/#!forum/missing-data) if you don't get an answer here.