The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » Use of weight on subsetted data (subsetted data, weight)
Use of weight on subsetted data [message #18008] Thu, 15 August 2019 14:14 Go to next message
FLW is currently offline  FLW
Messages: 3
Registered: August 2019
Member
Based on my understanding, the sampling weight in DHS is provided based on the entire survey sample.
If I would like to study only a subset of population i.e. men with children.
Can the original sampling weight be applied?
Will the applied weight ensure national respresentativeness of subsetted data?

Or I will need to calculate the new sampling weights based on the subsetted data?

Hope someone can guide me for this.

Thanks :)
Re: Use of weight on subsetted data [message #18009 is a reply to message #18008] Thu, 15 August 2019 15:38 Go to previous messageGo to next message
schoumaker is currently offline  schoumaker
Messages: 44
Registered: May 2013
Location: Belgium
Member
Yes, you can use the weights, but their sum should equal to the sample size for your subset. So you may have to rescale them. This is automatic with some commands and/or some software packages, but not all.
Best,
Bruno


Bruno Schoumaker
Centre for Demographic Research
Université catholique de Louvain
Re: Use of weight on subsetted data [message #18011 is a reply to message #18009] Fri, 16 August 2019 02:52 Go to previous messageGo to next message
FLW is currently offline  FLW
Messages: 3
Registered: August 2019
Member
I have seen this in the internet regarding subsetting data.

https://stylizeddata.com/how-to-use-survey-weights-in-r/

is it appropriate to do so?
Use svydesign in R to declare surveydata and weights, then only i subset the data.
Will it fix the rescaling issue?
Re: Use of weight on subsetted data [message #18060 is a reply to message #18011] Tue, 03 September 2019 14:40 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1748
Registered: February 2013
Senior Member

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

Are you asking whether it is appropriate to use weights with DHS data (it is!) or are you asking how to use weights with R? The link should provide all the information you need about the latter.
Re: Use of weight on subsetted data [message #18325 is a reply to message #18008] Sun, 10 November 2019 15:52 Go to previous messageGo to next message
LizavG is currently offline  LizavG
Messages: 1
Registered: November 2019
Member
This is a really relevant question and I would like to extend it.I am also sub-setting data based on the availability of child measures and ages. I would like to rescale the data because I am working with STATA comments that do not allow for svy.

I know that v005 are the weights usually used for weighing in DHS but I am unsure how it is constructed, and thus how I could rescale the data. What factors are used to create v005?
Re: Use of weight on subsetted data [message #18359 is a reply to message #18325] Mon, 18 November 2019 07:33 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1748
Registered: February 2013
Senior Member

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

There are some Stata commands that do not allow svyset. For some of those commands, however, you CAN use weights, or weights and clusters, so long as you don't include strata, which seem to be the most complicated part of the svyset adjustments. You can use the weights with, for example, "[pweight=v005]" or "[iweight=v005/1000000]". Those inserts always go before the comma in the estimation command. You can try to use the cluster adjustment by putting "cluster(v021)", for example, with the options AFTER the comma. That may work even if svyset is not accepted.

If you are using pweight, which is the type of weight that is required in svyset, you do not need to re-scale v005 in any way. In fact, with pweights you CANNOT rescale v005. Whatever variable you enter as a pweight will always be rescaled so the values add to 1. For that reason, for example, "[pweight=v005]" and "[pweight=v005/1000000]" will always give you the same results. Many users go through the step of defining the weight to be v005/1000000 but that's not necessary (for the pweight option). Hope this answers your question.
Previous Topic: Pooling 3 rounds of DHS Nepal -- weights?
Next Topic: Using normalized weight in analysis
Goto Forum:
  


Current Time: Tue Feb 18 15:27:40 Eastern Standard Time 2020