If I want to do a chi square test and a logistic regression in Stata using the 2 variables I extracted from IPUMS Health Survey, which weight should I use if the two variables has two different weights defined for them?
Since your regression will only keep observations in which all variables have values, you should use the sample weight of the variable with the more restrictive universe statement. Considering the information from your previous question, since “sample adults” are by construction a sub-set of all persons within the NHIS sample, it sounds like you should use the SAMPWEIGHT.