Question on Stata's svy: with suest

[David Speed]

Hi Forum,

I am using the 2013 Canadian General Social Survey, which allows its data to use the svyset feature:

svyset [pweight=wght_per], bsrweight(wtbs_001-wtbs_500) bsn(25) vce(bootstrap) dof(500) mse

I am comparing two coefficients across two different models using suest:

foreach var of varlist first_gen3 first_gen4 path2 path3 path4 path5 euro_imm1 euro_imm2 euro_imm3 euro_imm4 {
    qui svy: regress belong_canada if `var'==1
    est store can_`var'
    qui svy: regress belong_origin if `var'==1
    est store ncan_`var'
    suest can_`var' ncan_`var'
    test [can_`var']_cons = [ncan_`var']_cons
}

Stata will execute this command with no questions asked. However, the Stata documentation for svy postestimation states that:

suest can be used to obtain the variance estimates for a series of estimators that used the svy prefix. To use suest for this purpose, perform the following steps:

1. Be sure to set the survey design characteristics correctly by using svyset. Do not use the vce() option to change the default variance estimator from the linearized variance estimator. vce(brr) and vce(jackknife) are not supported by suest.

One of the things I really like about Stata is that it is pretty good with letting you know something isn't allowed. So I am puzzled about Stata "letting" me run this model, my two questions are:

1. Does my model violate Stata's recommendation?
2. If it does, what are the consequences of the violation?

Thanks everyone!

Cheers,

David.

[David Speed]

I reached out to @StataCorp and they told me that it should have produced a return code. They've fixed it from Stata 15+ I think.