1) Can someone point me to a resource that explains why the linearized S.E. and the logit transformed 95% CI are different when I do svy: proportion using the subpopcommand instead of the if var == x command?

2) Is it ok if my subpop variable consists of 1, 0, missing?

3) When calculating the proportions, Stata deleted cases that had missing data on the outcome. How does this affect the interpretation of results? Are all the things I set in svyset(weights, clustering) still valid?

4) I understand that I could do multiple imputation for the missing data, however, could I instead try to show that the cases with the missing data are not significantly different from the cases without missing data, on several demographic variables (gender, SES, etc.) and argue thus that the sample is still representative of the population? Have others taken this approach before?