I’m trying to accurately calculate margins of error at the 90% confidence level using ACS data in SAS.
My code (which also includes chi square) looks like this, and is working great.
PROC surveyfreq DATA=data1 VARMETHOD=JACKKNIFE;
tables variable1*(variable2)/ expected row col chisq lrchisq wchisq wllchisq CLWT CL ALPHA = 0.1;
Question: There are two ways I’ve calculated the MOE, and results differ. If I multiply the resulting standard error by 1.645, I get a slightly different result than if I subtract an estimate from its upper limit. I think I’m confusing myself or missing something, but shouldn’t these be the same? If not, which one would be the correct one drawing on the jackknife/replicate weight method?
Percent | 4.132
Std Err ofPercent | 0.085
90% Confidence Limits for Percent | 3.991, 4.274
MOE= 1.645*0.085 = 0.1397
MOE= 4.274 - 4.132 = 0.1414