| Number of strata in 1999/2000 and 2004 BD DHS [message #1565] |
Thu, 13 March 2014 16:11  |
annadinnyc
Messages: 9
Registered: March 2014
|
Member |
|
|
Hello,
I cannot find the number of strata used in the BD DHS 1999/2000 and 2004. I checked the final reports and they only provide the number of Enumeration Areas (341 and 361, respectively). The reports just mention choosing EAs from the list of EAs in the 1991 or 2001 census, respectively, but it is unclear how the two-stage selection was done.
When I tabulate the stratum variables (hv022 or v022, labeled "sample stratum number"), I get 168 for the 1999/2000 data and 177 for the 2004 data. In contrast, the 2007 and 2011 BD DHS have 22 and 20 strata, respectively.
I am trying to merge data from four BD DHS surveys and need to make sure that I get the survey design, weights, etc. correct.
Any suggestions would be appreciated. Thank you.
Anna
|
|
|
|
|
|
|
|
|
|
| Re: Number of strata in 1999/2000 and 2004 BD DHS [message #1569 is a reply to message #1568] |
Thu, 13 March 2014 18:19  |
Reduced-For(u)m
Messages: 292
Registered: March 2013
|
Senior Member |
|
|
Anna,
I'm not a DHS person, but I've used the BD surveys a good bit. I used the code from the DHS FAQ: gen strata = group(v024 v025)
BUT - looking at the documentation for the 1996 survey (https://www.dhsprogram.com/pubs/pdf/FR88/FR88.pdf) page 3, they actually appear to have stratified at a different level ("Each division was stratified into three groups: 1 ) statistical metropolitan areas (SMAs) 2,
2) municipalities (other urban areas), and 3) rural areas." - so not just urban/rural-X-region, but urban/SMA/rural-X-region. So that would look to me like you should use:
gen strata = group(v023 v025) for 1996.
It looks like this system changed from round to round. Further, v023 is all 0 for 2004, meaning it isn't representative at sub-national levels, and for that one you'd need to use the group(v024 v025).
I'd suggest looking at the sampling design section (like the one I quoted) from each survey, generate the appropriate strata for each survey (by, say, tacking on the survey round, so 1_1996, 2_1996, 1_1997 - however you want to do that)and then use that new strata var in your svyset command. But I'm pretty sure v022 isn't the right one ever (something about for use with Taylor expansions? I dunno).
So I learned something, and I'd been doing it wrong. And that's always nice (well, the learning part, not the embarrassment part).
Hope this helps a bit. It helped me.
|
|
|
|
|
|
|
|
| Re: Number of strata in 1999/2000 and 2004 BD DHS [message #1604 is a reply to message #1573] |
Mon, 17 March 2014 10:57 |
annadinnyc
Messages: 9
Registered: March 2014
|
Member |
|
|
Hello Sarah,
I would like you to clarify please. I do not use round 1996 (which the other user refers to below). The DHS reports for 1999/2000 and 2004 are not explicit about stratification (only about the number of EAs), though the 2004 report seems to suggest a stratification procedure is similar to the 2007 data (the 2007 report was explicit).
Thanks,
Anna
|
|
|
|
|
|
|
|
|
|
|
|
| Re: Number of strata in 1999/2000 and 2004 BD DHS [message #2798 is a reply to message #2777] |
Mon, 25 August 2014 21:25 |
Trevor-DHS
Messages: 789
Registered: January 2013
|
Senior Member |
|
|
From the Uganda 2011 report (appendix A), the sample was stratified by urban rural within each of the 10 regions. Thus you can create the strata variable as follows:
egen strata=group(v024 v025)
gen wt=v005/1000000
svyset v021 [pw=wt], strata(strata)
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
