Test for differences in proportions of different subpopulations using survey weights?

[Daniel Sun]

Hello!

I'm wondering if there's a way to compare proportions across two subpopulations from a single survey-weighted sample.

For example, I want to test whether there's a difference between the prevalence of males between two non-mutually exclusive subpopulations, cohort_a and cohort_b, and generate a p-value:

. svy, subpop(cohort_a): proportion male
(running proportion on estimation sample)

Survey: Proportion estimation

Number of strata =      24      Number of obs   =       15,560
Number of PSUs   =      49      Population size =  322,324,172
                                Subpop. no. obs =          211
                                Subpop. size    = 4,763,840.19
                                Design df       =           25

--------------------------------------------------------------
             |             Linearized            Logit
             | Proportion   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
male         |
           0 |   .4747481   .0527103      .3690163    .5827937
           1 |   .5252519   .0527103      .4172063    .6309837
--------------------------------------------------------------

. svy, subpop(cohort_b): proportion male
(running proportion on estimation sample)

Survey: Proportion estimation

Number of strata =      24      Number of obs   =       15,560
Number of PSUs   =      49      Population size =  322,324,172
                                Subpop. no. obs =          520
                                Subpop. size    = 12,796,639.8
                                Design df       =           25

--------------------------------------------------------------
             |             Linearized            Logit
             | Proportion   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
male         |
           0 |    .489929   .0349094      .4187285    .5615404
           1 |    .510071   .0349094      .4384596    .5812715
--------------------------------------------------------------

Thanks in advance!

[Vivek Gupta]

Hi Daniel
You can use svy: logistic outcomevar explanatoryvars to account for survey sampling and sampling weights

thank
Vivek