The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » HIV » Svyset in Stata for HIV results and ART use data
Svyset in Stata for HIV results and ART use data [message #637] Thu, 18 July 2013 10:20 Go to next message
petermacpherson is currently offline  petermacpherson
Messages: 1
Registered: July 2013
Location: UK

I am conducting a study investigating the prevalence of HIV-negative status on HIV testing among individuals who report that they are currently taking antiretrovirals. I plan to conduct this analysis using data from DHS surveys in sub-Saharan Africa, but have started with the Malawi 2010 DHS to get a feel for the data. I am using Stata 12.1 for analysis.

So far, with the Malawi 2010 DHS, I have appended the men's individual records to the women's individual records and merged with the HIV testing records, following instructions in a previous forum post.

However, I am not completely sure about how to procedure from here. I am stuck on three main areas:

1) I am not sure about how to correctly "svyset" the data in Stata. This is my code:
gen psu=v021
egen strata = group(v024 v025), label
gen weight=hiv05/1000000
svyset psu [pweight=weight], strata(strata)

Is this correct? I am relatively unfamiliar with analysing survey data.

2) Estimating the prevalence of HIV-negativity among individuals who report taking ART. Again, I am not sure if this is correct syntax
svy: tab hiv03 s1318

3) Finally, I want to construct a logistic regression model to investigate potential risk factors for HIV-negative status in individuals who report taking ART (e.g. sex, age group etc)
My intention is to generate a new variable (hivneg_art) with 1=HIV negative and taking ART, and 0=HIV positive and taking ART.

My code for analysis would then be:

xi:svy: logistic hivneg_art

Is this correct?

Thanks so much for the assistance.

Re: Svyset in Stata for HIV results and ART use data [message #905 is a reply to message #637] Thu, 31 October 2013 16:43 Go to previous message
Messages: 1516
Registered: February 2013
Senior Member
Dear User,
Your specification of svyset is ok. Unfortunately, however, DHS data will not get you very far with this topic. First, only two surveys are available that include self-reports of ART--Malawi and Uganda. Second, we have evidence that these self-reports are not reliable. Substantial proportions of users of ART will not acknowledge such use. Respondents are not asked about ART use unless they disclose that they are HIV positive, and there is evidence suggesting that self-reported HIV-status among people who are HIV-positive is poor. Third, the number of respondents who say they are using ART but are HIV negative is very small. There are only about 20 to 30 cases in each survey, certainly not enough for statistical analysis.

Let us know if you have other questions.

[Updated on: Fri, 01 November 2013 16:39]

Report message to a moderator

Previous Topic: How to use Weights for HIV and Women data
Next Topic: Replicating HIV testing coverage rates
Goto Forum:

Current Time: Mon Jun 27 14:04:36 Coordinated Universal Time 2022