Does SURVEYMEANS contain functionality to process a ratio as a mean, and BRR weights for SE as comparable ratio processing?
The SURVEYMEANS ratio functionality enables me to process household microdata to calculate a full sample annualized mean estimate as expected, where the weights for the numerator (no time adjustment) are different than the denominator (time adjusted) due to the survey design. The issue with the SURVEYMEANS ratio statement processing is that I ignored replicate weights and therefore the SEs are not as expected. When I attempt to process the replicate weights as varmethod=BRR and corresponding weight and repweights statements, the resulting means and SEs are not as expected because the numerator and denominator are weighted incorrectly.
The following code processes SEs as expected, where the NUM/DEN ratio is evaluated for the 44 replicates. Does this type functionality exist within SURVEYMEANS? If so - could you provide a reference to apply SURVEYMEANS correctly? The end goal is to determine if PROC SURVEYMEANS can replace my coded process.
sqrt(sum((1/44) * ((NUM/DEN) - &FS_RAT)**2))
Thanks- Josh
