I have a survey design where there are regional strata, from which are selected a number of sub-regional PSUs. Within each of these there is a further selection of sites where interviews take place.
Should I use a FPC at the top level (eg in proc surveymeans)? This makes a substantial difference, as some strata only have a few PSUs in the population. However, I've seen some commentary suggesting that the FPC is only appropriate if I'm selecting the whole population within each PSU. I've looked in a few texts, but haven't found this issue discussed. Can anyone point me to suitable references?