The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » IPUMS Demographic and Health Surveys (IPUMS-DHS)  » Complex weights for pooling cross country IPUMS-DHS data
Complex weights for pooling cross country IPUMS-DHS data [message #17906] Fri, 12 July 2019 01:46 Go to next message
rafi.amiruddin@gmail.com is currently offline  rafi.amiruddin@gmail.com
Messages: 1
Registered: January 2018
Member
I want to do a cross-country study on child mortality using DHS data (birth recode). I am using IPUMS-DHS data (https://www.idhsdata.org/idhs/). My data extract from IPUMS-DHS consists of around 30 countries. All available waves for each country are included in the analysis.

I want to pool all the surveys across time and countries. I would be obliged if someone could tell me how to svyset the data in Stata. Some of the available methods which I could find so far suggest grouping the strata "egen stratid = group (year v024 v025), label" ( https://www.stata.com/statalist/archive/2014-01/msg00150.htm l). However, the problem with this approach is that the IPUMS-DHS data does not have region variable (v024). One obvious reason why v024 is not there in IPUMS-DHS is that regions are different in different countries.

Using the usual svyset command "svyset idhspsu [pweight=v005], strata(idhsstrata)" gives the error "Note: Missing standard errors because of stratum with single sampling unit." I am using Stata/MP 15.1.


Rafi Amir-ud-Din
Re: Complex weights for pooling cross country IPUMS-DHS data [message #17911 is a reply to message #17906] Sat, 13 July 2019 15:31 Go to previous message
boyle014 is currently offline  boyle014
Messages: 70
Registered: December 2015
Location: Minneapolis
Senior Member
The problem of the single unit in a stratum error isn't unique to IPUMS DHS--that would occur using the standard DHS data files as well.

You can deal with the error by telling Stata what to do when it encounters single units. One option is to add "singleunit(center)" to the end of the command:

svyset [pw=perweight], psu(idhspsu) strata(idhsstrata) singleunit(center)

Note that this will weight the data properly at the sample level, but if you want to aggregate across pooled samples, you'll also have to weight up to the total populations.

We're working on adding these population weights into IPUMS. So far, they are available for households--the variable is popwt. More on that at this post.


Professor Elizabeth Boyle
Sociology & Law, University of Minnesota, USA
Principal Investigator, IPUMS-DHS
Previous Topic: Next IPUMS-DHS Release
Next Topic: IPUMS DHS Tip #7 - Comparing statistics over time
Goto Forum:
  


Current Time: Sun Dec 15 09:21:45 Eastern Standard Time 2019