Hi,
I was facing a question regarding the svyset command that is adjusting for weights , stratification, and clustering. With the IPUMS- USA data, I see many studies only accounting for sampling weights, and not adjusting for stratification and clustering. My interpretation from the IPUMS USA website has led me to come to the conclusion that controlling for sampling weights is necessary as well as sufficient. Am I correct in my interpretation? Thank you
When calculating statistics with IPUMS USA data, you only need to use the weights to get correct point estimates. However, in order to calculate correct standard errors, you need to account for clustering and stratification.
For decennial census samples, the correct way to do this in Stata is using the STRATA and CLUSTER variables with svyset, as detailed on the IPUMS USA Variance Estimation page. However, for ACS samples, the Census Bureau releases and recommends the use of replicate weights (available in the variables REPWT and REPWTP for households and individuals, respectively). These can also be used with Stata’s svyset command, as detailed at the IPUMS USA Replicate Weights page.