luhua Zhao posted on Thursday, August 07, 2008 - 1:43 pm
hello, I am trying to run a multi-level logistic regression on BRFSS data (telephone survey by CDC). BRFSS has strata, psu and a single-levlel sampling weight variable (sampling component variables are included in datafile). My level 1 unit is individual and level 2 unit would be MSA, which is differnt from survey strata or psu. Do you think I can calculate the sampling weight for multilevel regression in this case? Also can you commment why the single level sampling weight can't be used for multilevel analysis? Thanks.
Typically the single level weight variable is the product of the weight variables on the within level and the weight variable on the between level and you would have to split it up for multilevel analysis. In some case the between level weight is 1 for all between level units (if they were equally likely to be selected in the sample). In that case you would have the single level weight variable as the within level weight variable.