11-13-2016 10:48 PM
I'm trying to calculate relative standard errors for weighted survey data in VA. I've used the formula:
RSE = StdErr_ByGroup_(Weighted value) / Sum_ByGroup_(Weighted Population) * 100
However, this is returning RSEs that are implausibly low. From the output, I'm assuming that the standard error function in VA uses proc means, whereas what we're trying to achieve is normally calculated through proc surveymeans.
From this, I have two questions:
1. Is anyone able to provide me with more information about how standard errors are calculated in VA? I'd like to confirm my suspicions but I haven't had any luck searching the online documentation.
2. Does anyone have any suggestions as to how proc surveymeans' RSE procedure might be replicated through VA's expression builder?