I have a problem with generated confidence intervals for prevalence using survey data. We are investigating drug resistance in the country and had PSU as regions and enrolled individuals on drugs in health facilities randomly selected from these regions. The dataset is svyset and I use the following command for prevalence with the resulting table of results. Note the highlighted upper bound level
svy linearized, subpop (if GENDER_2 ==0): proportion VL_SUPPRESSED_BN
(running proportion on estimation sample)
Survey: Proportion estimation
Number of strata = 5 Number of obs = 533
Number of PSUs = 23 Population size = 284401
Subpop. no. obs = 182
Subpop. size = 96827.4
Design df = 18
------------------------------------------------------------------
| Linearized
| Proportion Std. Err. [95% Conf. Interval]
-----------------+------------------------------------------------
VL_SUPPRESSED_BN |
0 | .0234302 .0150483 -.0081852 .0550456
1 | .9765698 .0150483 .9449544 1.008185
------------------------------------------------------------------
Note: strata with single sampling unit centered at overall mean.
svy linearized, subpop (if GENDER_2 ==0): proportion VL_SUPPRESSED_BN
(running proportion on estimation sample)
Survey: Proportion estimation
Number of strata = 5 Number of obs = 533
Number of PSUs = 23 Population size = 284401
Subpop. no. obs = 182
Subpop. size = 96827.4
Design df = 18
------------------------------------------------------------------
| Linearized
| Proportion Std. Err. [95% Conf. Interval]
-----------------+------------------------------------------------
VL_SUPPRESSED_BN |
0 | .0234302 .0150483 -.0081852 .0550456
1 | .9765698 .0150483 .9449544 1.008185
------------------------------------------------------------------
Note: strata with single sampling unit centered at overall mean.
