I am calculating a lot of means on survey-data, where I have to use proportional weights.
Unfortunately, weights
are not allowed with the command ci
. I could use the mean
command but I need to have the results stored, so I subsequently can put them into a matrix.
My loop-code looks like this:
local i = 1
foreach var of varlist hinst1 - hinst34 {
ci varname if `var'==1
mat matname[`i',1]=r(mean)
mat matname[`i',1]=r(lb)
mat matname[`i',1]=r(ub)
local ++i
}
It is possible to use aweights
but I need pweights
.
Using Stata's nhanes2f
toy dataset as an example:
webuse nhanes2f, clear
svyset psuid [pweight=finalwgt], strata(stratid)
The svy
prefix can be used as follows:
matrix A = J(3,3,.)
local i = 1
foreach var of varlist sex race region {
svy: mean age if `var'==1
matrix res = r(table)
matrix A[1,`i'] = res[1, 1]
matrix A[2,`i'] = res[5, 1]
matrix A[3,`i'] = res[6, 1]
local ++i
}
matrix list A
A[3,3]
c1 c2 c3
r1 41.898222 42.508557 43.133186
r2 41.206209 41.852839 42.034576
r3 42.590235 43.164274 44.231796