The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » Issues with weighting NEW Ethiopian DHS-2019
Issues with weighting NEW Ethiopian DHS-2019 [message #23223] Mon, 02 August 2021 15:29 Go to next message
id709nvz is currently offline  id709nvz
Messages: 42
Registered: March 2019
Member
Dear Sir/Madam
I have one question that I believe you will clarify to me:
I am using the new release of Ethiopian DHS-2019, mainly KR file.
I am interested in children's nutritional status for this I merge it HR and KR files.
Using v001 v002 as a key identifier. But When I tab anthropometry [iw=weight]
I repeatedly get no observation.

Could you please let me know how to appropriately merge the children anthropometry index from HR to the KR file
so I can tab with weights?

Kind regards
ZF
Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23276 is a reply to message #23223] Mon, 16 August 2021 08:47 Go to previous messageGo to next message
Shireen-DHS is currently offline  Shireen-DHS
Messages: 121
Registered: August 2020
Location: USA
Senior Member
Hello,

For children's anthropometry indicator you need to use the PR file and no merge is required. The code below will match the stunting indicator for example.

gen wt=hv005/1000000
//Stunted
gen nt_ch_stunt= 0 if hv103==1
replace nt_ch_stunt=. if hc70>=9996
replace nt_ch_stunt=1 if hc70<-200 & hv103==1
label var nt_ch_stunt "Stunted child under 5 years"


You can find this code and other code for DHS indicators listed in the Guide to DHS Statistics on our GitHub site: https://github.com/DHSProgram/DHS-Indicators-Stata
Specifically, this code is in Chapter 11 in the NT_CH_NUT.do file. We also have the code in SPSS here: https://github.com/DHSProgram

Please read the readme file and the main files for each chapter before using the code.

Thank you.
Best,
Shireen Assaf
The DHS Program
Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23277 is a reply to message #23276] Mon, 16 August 2021 09:07 Go to previous messageGo to next message
id709nvz is currently offline  id709nvz
Messages: 42
Registered: March 2019
Member
Dear,
Thanks for your reply.
The reason why I use KR instead of PR is, I need children below five years old and their mother only.
Can you please tell me the total number of female population base where the 2019 Ethiopian DHS IR file was sampled?
It will help me to create the correct weight: As I am intending to combine DHS 2000, 2005, 2011, 2016, and 2019 in my analysis.

Kind regards
Zecharias
Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23278 is a reply to message #23277] Mon, 16 August 2021 09:20 Go to previous messageGo to next message
Shireen-DHS is currently offline  Shireen-DHS
Messages: 121
Registered: August 2020
Location: USA
Senior Member
Hello,

In this case you can use the KR file and use the anthropometry variables hw70, hw71 and hw72. You can code stunting for instance in the same way by replacing hc70 with hw70 and remove the hv103 condition. You would use the weight v005 and divide by 1000000 as done below. There is no need to create a weight variable.

Best,
Shireen
Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23289 is a reply to message #23278] Mon, 16 August 2021 16:21 Go to previous messageGo to next message
id709nvz is currently offline  id709nvz
Messages: 42
Registered: March 2019
Member
Dear,
thank you.
But previously one of you suggested me to do the following on pooled DHS rounds:
That is re-weighting by the population base for each survey year.

*e.g dhs 2000,
gen wgt=v005/1000000
gen weight=(wgt*24949)/15683

The same code for all other surveys and I would use 'weight' for all tabulation in the following wave
tab x [iw=weight] instead of tab x [iw=wgt]

This is why I ask for the total female population base for the Ethiopian DHS /2019/

Shouldn't I take into account differences in population across survey rounds when I play with a pooled sample?

Kind regards
ZF

Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23296 is a reply to message #23289] Tue, 17 August 2021 12:37 Go to previous messageGo to next message
Shireen-DHS is currently offline  Shireen-DHS
Messages: 121
Registered: August 2020
Location: USA
Senior Member
Hello,
Can you refer me to this response about the pooled sample?
In a pooled sample you will also need to create a unique strata and PSU for each survey to use in the svyset if you are using Stata.
To give equal weight to each survey in a pooled analysis I have used previously the following code (this is using appended data from several surveys). There are many ways to deal with pooled samples so there could be another way suggested that is also fine.

* code to give equal weight to each surveys
* you will need a survey variable to identify each survey
egen N=total(v005)
egen n=total(v005), by(survey)

* In this case I had 25 surveys, change this number to the number of surveys you have.
gen wtr=v005*(N/25)/n

egen v001r=group(survey v001)
egen stratar=group(survey strata)
svyset [pw=wtr], psu(v001r) strata(stratar) singleunit(centered)

Hope this helps.
Best,
Shireen

Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23297 is a reply to message #23296] Tue, 17 August 2021 16:06 Go to previous messageGo to next message
id709nvz is currently offline  id709nvz
Messages: 42
Registered: March 2019
Member
Dear,
You can find it here, please go to the last few questions: DHS Burundi 1987 by Trevor-DHS


The idea of Trevor response is that
*Open KR 2000 (e.g)
gen wgt=v005/1000000
gen weight=(wgt*total FEMALE POPULATION)/FEMALE SAMPLE -For Ethiopian KR 2000; total FEMALE POPULATION=14619 and FEMALE sample=15367

repeat the process for all survey rounds and use tab X [iw=weight] rather than tab X [iw=weight]

I look forward to hearing from you.
Kind regards
ZF
Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23299 is a reply to message #23297] Wed, 18 August 2021 08:31 Go to previous messageGo to next message
Shireen-DHS is currently offline  Shireen-DHS
Messages: 121
Registered: August 2020
Location: USA
Senior Member
Hello,

Yes this is another method. It works better for when you you expect the population sizes to be very different from one survey to the next. I have asked Trevor where he received the population estimates and this was his response:

When I pool and re-weight the data, I use population estimates from World Population Prospects. (https://population.un.org/wpp/Download/Standard/Population/). I pick "Population by broad age groups female", which includes a column for 15-49 (column AQ), and find the Ethiopia estimates for the relevant years. For the 2019 survey I would use the 2020 estimate which is 28,513 (thousand, i.e. 28.5 million).

Thank you.
Best,
Shireen Assaf
The DHS Program

Re: Issues with weighting NEW Ethiopian DHS-2019 [message #23364 is a reply to message #23299] Sun, 29 August 2021 18:55 Go to previous message
id709nvz is currently offline  id709nvz
Messages: 42
Registered: March 2019
Member
Dear,
It helps me a lot.

Thank you!
Previous Topic: Nigeria DHS 2018: Weighting for Multilevel Model
Next Topic: How to use weights when using different samples of women
Goto Forum:
  


Current Time: Sun Jun 26 04:38:50 Coordinated Universal Time 2022