The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Merging data files » Which DHS File should I use? (Urgent)
Which DHS File should I use? [message #19716] Wed, 05 August 2020 04:37 Go to next message
Chery87 is currently offline  Chery87
Messages: 6
Registered: November 2019
Member

Dears,

I am writing a research paper using two different DHS series (Earliest and latest)to study socioeconomic related inequalities (measured by wealth index) of individuals. as I am employing the regression-based decomposition of variables, my questions are presented as follows.

1) which files am I supposed to use in calculating u5mr and linking it with mother's age, household wealth index, mother's level of education, sex of children, sex of household head, residence type, and region?

2) which files am I supposed to use in calculating the stunting rate for children aged 5 years and below and linking it with mother's age, household wealth index, mother's level of education, sex of children, sex of household head, residence type, and region?

3) I want to study the changes of those variables between the earliest and latest DHS and thus, Do I have to merge the DHS files for two different periods or estimation must be done separately?


Looking forward to hear from any one of the experts.
Re: Which DHS File should I use? [message #19717 is a reply to message #19716] Wed, 05 August 2020 08:31 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3214
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

The survival of a child, and whether a living child is stunted, etc., are characteristics of the child. Child survival is indicated by b5 in the KR or BR file. Stunting, etc. is calculated from hw70-hw72 in the KR file because it is only available for (surviving) children age 0-4. It can also be calculated from hc70-hc72, in the PR file. The PR file includes all children age 0-4 identified in the household survey. The PR file includes anthropometry for some children who are not in the KR file. The KR file includes those children in the PR file whose mother was alive and in the same household.

To get a rate you have to pool the survival information about, say all the children in urban areas and all the children in rural areas. A rate is not multivariate. DHS reports include child mortality rates for sub groups, and that is done by calculating the mortality rate over and over within each sub group. Similarly, the percentage stunted is calculated within subgroups.

For multivariate analysis you can just work with a binary (0 or 1) outcome--for survival or stunting, for example, and logit regression. That's what I would recommend.

If you want to look at change between two surveys in the same country, or you want to look at differences between two surveys, you can put them into the same data file, but have codes such as survey=1 and survey=2 when you want to describe or test a difference. This is appending, not merging.
Re: Which DHS File should I use? [message #19751 is a reply to message #19717] Sat, 08 August 2020 08:35 Go to previous messageGo to next message
Chery87 is currently offline  Chery87
Messages: 6
Registered: November 2019
Member
Thank you very much for the briefing, found it helpful. here is another related request for technical support.

Do anyone of you have "do-file" with special emphasis on U5MR and stunting and be explained by socioeconomic covariates like mother's age, household wealth index, mother's level of education, sex of children, sex of household head, residence type, and region?


Looking forward to hearing from you.


Regards,

Chery
Re: Which DHS File should I use? [message #19757 is a reply to message #19751] Mon, 10 August 2020 09:15 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3214
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

Rates such as the U5MR describe an aggregate. You can calculate the rate separately within categories of a covariate. That's all you can do with a rate.

If you want to do multivariate analysis, you have to construct a binary outcome for each child based on survival and analyze with hazard models or survival models or logit regression.
Re: Which DHS File should I use? [message #19760 is a reply to message #19751] Mon, 10 August 2020 11:57 Go to previous messageGo to next message
Chery87 is currently offline  Chery87
Messages: 6
Registered: November 2019
Member
Thank you very much once again.

In addition to the previous questions, I want to estimate:

1) stunting prevalence across urban and rural, wealth quintiles, mother's education groups

2) U5MR in Urban and rural, wealth quintiles, mother's education groups.


how can I proceed with stata commands to estimate in those sub-groups? I can only estimate those rates at national level.


Thanks in advance.


Regards,

Chery
Re: Which DHS File should I use? [message #19787 is a reply to message #19760] Fri, 14 August 2020 10:26 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3214
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

Most of what you want to calculate is already provided in the main report on each survey, and in StatCompiler for each survey. Programs to calculate them in Stata are given with the GitHub link that is now on the landing page for the DHS Forum. Let us know if you cannot find the GitHub programs.
Re: Which DHS File should I use? [message #19799 is a reply to message #19787] Mon, 17 August 2020 01:44 Go to previous message
Chery87 is currently offline  Chery87
Messages: 6
Registered: November 2019
Member


Dear,


Thank you very much for the recommendations, I already know and accessed them. I want to estimate myself and finally I have figured out how to do it.

I may come back in the future if I need your technical support.


regards,

Chery
Previous Topic: Merging DHS data in R
Next Topic: How to merge IR, KR and HR files in stata
Goto Forum:
  


Current Time: Sun Dec 22 21:55:31 Coordinated Universal Time 2024