| DHS 08/09 - HIV prevention variables in dataset vs. DHS report [message #4082] |
Thu, 26 March 2015 15:50  |
User_0315
Messages: 5
Registered: March 2015
|
Member |
|
|
Hi,
I'm doing analyses using HIV attitude / knowledge variables. First, I tried to recreate all values of interest to match those published in the DHS report (http://dhsprogram.com/pubs/pdf/fr229/fr229.pdf) and StatCompiler. While most of the values agree, I ran into trouble with HIV prevention questions among both men and women.
For example, looking at avoiding AIDS through condom use (in the individual recode: S908B) or limiting sex to 1 faithful partner (S908C), I get much lower frequencies than published in the sources listed above (condom use: 36% vs. 75%; 1 partner limit: 47% vs. >90%).
I'm using survey procedures in SAS and flat format files (keir52fl for individual recode).
It seems that the original question on prevention was open-ended (the questionnaires in the PDF report state "Circle all mentioned" under "908B. What can a person do? Anything else?"; the same in the male questionnaire (708B)). However, StatCompiler and several tables in the PDF report refer to the published vales as "prompted." Can it be that participants were prompted with each option after answering the open-ended question, and the open-ended answers are in the datafile, but published data used prompted responses with higher prevalence?
Thank you!
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
