I am involved in a research project examining HIV risk behaviors, pooling data from recent surveys in 14 sub-Saharan African countries.

I plan to fit a 3-level HLM model , with individuals nested in regions , which are in turn nested within countries (individuals --> communities --> countries).

DHS provides two possible options:

v021: cluster/psu
v022/v023: strata


My question is which variables to use for the level 2 (community) differences. Of both, the psu variable is more appropriate since it roughly represents the cluster from which respondents are selected.

However, since I've already used the same psu variable to weight the data using svyset, would it still be appropriate to use it in my multilevel model as a level 2 variable?

Thanks in advance for your attention.

cY