Dear all, I have a example from library survey about estimating in subpopulations and I need to reproduce it step by step. This is the example (from paper: Estimates in subpopulation of Thomas Lumley and just instead of svymean I used svytotal) that I did in R: library(survey) data(fpc) dfpc<- svydesign(id=~psuid, strata=~stratid, weight=~weight, data=fpc, nest=TRUE) dsub<- subset(dfpc, x>4) svytotal(~x, design=dsub) total SE x 123.9 33.004 svyby(~x, ~I(x>4), design = dfpc, svytotal) I(x > 4) x se.x FALSE FALSE 23.2 17.01764 TRUE TRUE 123.9 33.00409 and what I can't figure out is which formula or approximation is used for standard error calculation. I tried with some famouse literature written by Cohran, Sarndal, Lohr but I still can't figure out this. I would really appreciate if anyone can explain me in few steps formula which is used to obtain value of 33.00409 for standard error. Best regards, Andrija [[alternative HTML version deleted]]