I want to generate some averages for some occupations/industries with the monthly CPS data, and then compute means over time including replication weights implied standard errors. I understand that “svyset…” will use the replication weights as iweights by default. Will that still be working correctly in regressions after I collapse the data?
Yes, you are correct. Collapsing your data shouldn’t meaningfully impact your calculations of means and replication weighted standard errors. For more details about using replicate weights, see this page.