Dear Statalist community,
I am currently estimating the effects of household clean water access on women's health using several provincially representative cross-sectional MICS datasets. My estimation model is
ivreghdfe wom_health (hh_water= cluster_water) $ controls , abs(district) cluster(PSU)..
where hh_water is a binary endogeneous regressor. and IV is cluster_water is community-level clean water access (leave-oneout mean). The controls are at individual, household, and cluster -level. I also used the time FE and district FE.
Now i am using 6 provinces data collected in different years from 2013-2018. Is there still a need to employ the survey weights ever we dont claim it a nationally representative data or claim this having all provinces in analysis. How it will work if the IV is cluster level and weights are also at each cluster level.
when i use [pweight=wmweight], the results change significantly. I response on it would be highly appreciated.
I am currently estimating the effects of household clean water access on women's health using several provincially representative cross-sectional MICS datasets. My estimation model is
ivreghdfe wom_health (hh_water= cluster_water) $ controls , abs(district) cluster(PSU)..
where hh_water is a binary endogeneous regressor. and IV is cluster_water is community-level clean water access (leave-oneout mean). The controls are at individual, household, and cluster -level. I also used the time FE and district FE.
Now i am using 6 provinces data collected in different years from 2013-2018. Is there still a need to employ the survey weights ever we dont claim it a nationally representative data or claim this having all provinces in analysis. How it will work if the IV is cluster level and weights are also at each cluster level.
when i use [pweight=wmweight], the results change significantly. I response on it would be highly appreciated.
Comment