I am attempting to use the replicate weights to calculate confidence intervals for means and frequencies of CPS/ASEC variables in SAS using PROC SURVEYMEANS and PROC SURVEYFREQ procedures. However, when I do so, I am getting confidence intervals that seem unreasonably broad (e.g. +/- 3% for variables with 10,000 unweighted respondents). Please let me know if there are common misinterpretations of this replicate weights or programming errors that can lead to this issue. Thanks!
I am not sure what version of SAS you are using, but note that this document, provided by the US Census Bureau, provides SAS code that can be modified to create the replicate weight file. If you haven’t already, this may be a helpful resource. Also note that in general, the standard error of a particular estimate tends to increase with the use of replicate weights. More information about IPUMS CPS replicate weights is available here.