Chao Yawo posted on Friday, July 24, 2020 - 9:47 pm
I am fitting a multilevel model examining HIV risk behaviors using a pooled data from the DHS, in which individuals (level 1) are nested into clusters (level 2).
DHS provides only one total weight variable (perweight=v005 or hiv05). I have heard that one way to handle this issue is to set the level 2 weights to a constant, but this is not optimal in the case of DHS since the probabilities for selection are not constant but unequal.
If I want to fit a multilevel model, what other options do I have to apply the appropriate weights?