| Re: Pooled Datasets - Use of Svyset & regional controls [message #1274 is a reply to message #1064] |
Sun, 02 February 2014 20:08  |
Sarah B
Messages: 23
Registered: June 2013
|
Member |
|
|
Hi Lukas,
Most of these issues are addressed in a similar thread: http://userforum.measuredhs.com/index.php?t=tree&goto=93 3&S=ab4524c1bf5aed945d825720798d7ea8#page_top
For Uganda specifically, there are a few extra concerns: regions have changed over time, and not all regions of the country were included in every survey.
If you want the same stratification in every survey, I believe you could regroup the strata into the 4 regions (central, eastern, northern, and western) shown on page 4 of the UDHS 2000-2001 final report. In each final report, I believe information is included that would allow you to re-categorize the data into these 4 groups, though you would have to do this by hand.
A downside of this approach is that the geographic regions are quite large, and thus would not represent well any kind of "community." You would also lose some of the efficiency gains from stratification in estimating your sampling errors. On the plus side, though, your regions would be comparable across surveys, and you could pool the data files together -- with the caveats about weights and renaming clusters and regions mentioned in the other thread.
By the way, the same clusters are not selected in different survey years -- a different sample is drawn for each survey. So the cluster identifiers should be renumbered/renamed to be unique within your pooled data file.
I hope that is helpful -- if you have further questions, please repost.
|
|
|
|
Members
Search
Help
Register
Login
Home
