Using Person Weights When Strata Have One Sampling Unit


I am using 5-year 2015-2019 ACS data to describe the characteristics of young adults with a bachelor’s in Chemical Engineering. I am using the command that IPUMS provides here to analyze this data using person weights: svyset cluster [pweight=perwt], strata(strata)

Here’s my problem: Since the sample size is so small (~2,000 individuals) STATA cannot calculate standard deviations for person-weighted ACS variables because there needs to be more than one sampling unit in each strata. STATA treats these sampling units as missing values unless one of the following options is added to the svyset command:

Which one of these options would be the most appropriate to use with IPUMS data? Any help would be greatly appreciated.

Which option to use is an analytical question that is beyond the scope of the user support team. I recommend consulting the existing literature on these options to make your decision. Note that the Census recommends the use of replicate weights to calculate standard errors using ACS data. Here is a user note on replicate weights. The person-level replicate weight variable in IPUMS USA is REPWTP.