Hi friends,
I will use a variable in the June fertility supplement to define my analytic sample. Then I will link individuals in my analytic sample to themselves in the ASEC in the same year. I will not use the longitudinal nature of my sample. I do the linking only because the variable which I use to define my sample is only available in June, and the variable that is my outcome of interest is only available in ASEC. I assume that I should use FRSUPPWT to weight my data, and there might be some bias in my sample representativeness due to the attrition from ASEC to June, which cannot be addressed as I did not find any ready-made weight on the website that can address this bias. My question is, in addition to the attrition bias, is there any other source of bias that may affect my sample representativeness that I am not aware of? Thank you!
Since you define your sample using the June supplement, the proper weight to use is FRSUPPWT. However, you may also be interested in adjusting for who can be linked across these samples (due to both the rotating nature of the CPS and other factors that affect the sample such as people moving). This requires generating your own weights using iterative proportional fitting (ipf) or raking with FRSUPPWT as the base weight. IPUMS provides sample code for doing so in Stata. You would need to substitute the provided extracts with your own sample months and change the do-file to reflect the months and years that you’re deriving weights for. Feel free to reach out if you have any trouble generating these weights. No other potential sampling biases come to mind, but I would suggest taking a look at the fertility supplement sample notes as well as the universe and comparability sections for your variables of interest to see if there are any other issues you need to consider.
1 Like