The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » Which weight to use, Household or Child?
Which weight to use, Household or Child? [message #5581] Wed, 10 June 2015 11:10 Go to next message
duke2015
Messages: 27
Registered: March 2015
Location: United States
Member
Hi there,

I am running a regression predicting child stunting from variables on the household level. I have merged the HR file and the KR (child's) recode file.

There was a recommendation that the weight to use should be the one with the highest rate of non-response, but I'm not sure which weight that is....how can I figure this out? I'm using Uganda 2011.

Thank you!
Re: Which weight to use, Household or Child? [message #5629 is a reply to message #5581] Wed, 17 June 2015 17:33 Go to previous messageGo to next message
Reduced-For(u)m
Messages: 292
Registered: March 2013
Senior Member

My thoughts in this particular instance is that you want the weight from the dataset where your outcome variable comes from (in this case, stunting). So since the observations on the left hand side of your regression come from the child recode, you should use those weights.
Re: Which weight to use, Household or Child? [message #6697 is a reply to message #5629] Mon, 29 June 2015 10:40 Go to previous messageGo to next message
duke2015
Messages: 27
Registered: March 2015
Location: United States
Member
I thought that it would make sense to use the weight from the file of the unit of analysis, but I wasn't sure if this was the correct way to do things since there really aren't any 'rules' in this respect.

Thanks in advance.
Re: Which weight to use, Household or Child? [message #6934 is a reply to message #6697] Wed, 05 August 2015 10:18 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3017
Registered: February 2013
Senior Member
Following is a response from Senior Sampling Specialist, Ruilin Ren:

If you merge the HR file with the KR file, the only weight variables you might need are the household weight HV005, and the household weight for the sub-sample of male survey HV028 (if there is a male survey and that it was conducted in a sub-sample of households). Since we usually do not calculate a child weight, you have to use the household weight for anthropometry indicators. If the anthropometry is measured in all of the households, you can use HV005. If anthropometry is measured in the male sub-sample, you can use HV028.

Please note that the survey final report might have used HV005 in the later case too (may not be systematically), so if you want to match the report figures, you may try both weights to figure out which one to use. But the difference should be very small, and will not alter any statistical conclusions.

Hope this is helpful.
Previous Topic: Compute sample weight for a dataset
Next Topic: weighting to obtain subnational summary estimates
Goto Forum:
  


Current Time: Fri Mar 29 07:30:52 Coordinated Universal Time 2024