Comparing proportions using svy command

[NourElAyoun Ibrahim]

Hello,

I am a PhD student and working in stata only since 2 years.

I need to compare proportions of sexual outcomes (ever had sex, condom use at first sex...) in two independant datasets, and one of them is a survey dataset. I have troubles to combine the two datasets and to use commands that take into account the fact that one of the sample is weighted.

The first dataset (name=coverte) contains data from participants with HIV. It is not a survey. N=284

The second datasat (BS2010) contains survey data with its own weight. N=2,899

I appended the two datasets, once i've opened first the BS2010 dataset : "append using coverte07022023.dta, gen (base)"

Typing "svydescribe" shows the following :

svydescribe

Survey: Describing stage 1 sampling units

pweight: RD2TOT
VCE: linearized
Single unit: missing
Strata 1: <one>
SU 1: <observations>
FPC 1: <zero>

#Obs per Unit
----------------------------
Stratum #Units #Obs min mean max
-------- -------- -------- -------- -------- --------
1 2,899 2,899 1 1.0 1
-------- -------- -------- -------- -------- --------
1 2,899 2,899 1 1.0 1

284 = #Obs with missing values in the
-------- survey characteristics
3,183

For example, if I try to compare the proportions of participants reporting ever having sex, which is the binary variable RS_ni (RS for rapport sexuel in French).

I first used the tabulate command but the result that does not take into account the weight in the BS2010 survey, so my understanding is it's not correct :

. tab RS_ni base, col chi2

+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+

| base
RS_ni | 0 1 | Total
-----------+----------------------+----------
0 | 302 47 | 349
| 10.42 16.91 | 10.99
-----------+----------------------+----------
1 | 2,592 227 | 2,819
| 89.41 81.65 | 88.73
-----------+----------------------+----------
3 | 5 4 | 9
| 0.17 1.44 | 0.28
-----------+----------------------+----------
Total | 2,899 278 | 3,177
| 100.00 100.00 | 100.00

Pearson chi2(2) = 25.8040 Pr = 0.000

And if I try to use the svy command for proportions, it ignores the coverte sample. (number of observation : 2,899). I tried the svy: logit command and it is the same issue.

svy: proportion RS_ni, over (base)
(running proportion on estimation sample)

Survey: Proportion estimation

Number of strata = 1 Number of obs = 2,899
Number of PSUs = 2,899 Population size = 3,311.8753
Design df = 2,898

--------------------------------------------------------------
| Linearized Logit
| Proportion Std. Err. [95% Conf. Interval]
-------------+------------------------------------------------
RS_ni@base |
0 0 | .1227894 .007692 .108486 .1386854
1 0 | .8750966 .0077625 .8590646 .8895394
3 0 | .002114 .0012226 .0006795 .0065567
--------------------------------------------------------------


I would be very grateful if someone can help me. I have checked the stata survey data reference manual, but the only section referring to combining samples from multiples surveys did not help me.

[NourElAyoun Ibrahim]

I just want to complete with some important information before anyone is willing to answer :

The proportions in the first dataset (coverte) have been standardized to adjust for the distribution of age and education level of the second dataset (BS). BS is a French representative sample.
That is to say that actually I need to compare :
- standardized proportions in dataset n°1
- and weighted proportions in dataset n°2.

To compute the standardized proportions in the first dataset, I did svyset the data doing that :
g pop = 1
svyset [pweight=pop]

An then to compute the standardized rate of RS_ni in the first dataset, i did the following :
svy, subpop (female) : proportion RS_ni, stdize (age_dip) stdweight (std_wt)
svy, subpop (male) : proportion RS_ni, stdize (age_dip) stdweight (std_wt)

Thank you very much.