I have a survey data and I think I need to use pweight=wt to run the regression. The wt is constructed as the inverse of the probability a household to be selected =1/(prob county being selected*prob city being selected*prob any household being selected). I just wonder when I add fixed effect, which level of the fixed effect do I need to add in this case? Or the fixed effect has nothing to do with the sampling process？
Thanks