The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » NFHS-5, when to calculate new weights? (variables that form a subset of individuals, do they need a special weight variable?)
NFHS-5, when to calculate new weights? [message #25726] Thu, 01 December 2022 05:08 Go to next message
Parashara is currently offline  Parashara
Messages: 3
Registered: December 2022
Member
For India's NFHS-5 I try to analyze women's autonomy data (from the women's data set IAIR74FL). However some relevant questions (e.g.: if allowed to go to the market and under which circumstances) were only posed to a fraction of all individuals (about 1/7).
But every individual in the 'domestic violence' module is included.

I want to aggregate the data on a district level (using the collapse command in stata) and therefore weighting the data correctly is important. Also I want to include as much data as possible.

So do I need to calculate a seperate weight for that situation and variable or do the original individual or domestic violence weights still hold their value?
Furthermore if data from only the domestic violence module is considered some data would distort the final result, due to 3rd party listeners during the interview. Would I need to consider a new response rate (and hence calculate a new weight) for that as well?

Thank you very much for your attention and help.
Re: NFHS-5, when to calculate new weights? [message #25727 is a reply to message #25726] Thu, 01 December 2022 09:30 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 2664
Registered: February 2013
Senior Member
Following is a response from Senior DHS staff members, Tom Pullum and Fred Arnold:

You do not need to calculate new weights. You will be ok with the dv weights as they appear in the IR file. This survey sampled men in only 1/6 of the households (every 2nd household in 1/3 of the clusters) and the DV module was only used in those households. The necessary adjustments for subsampling and nonresponse have been incorporated into the construction of d005.

Third party listeners are not allowed for the domestic violence module under any circumstances. If complete privacy for the DV module is not possible, the entire domestic violence module is skipped.

[Updated on: Thu, 01 December 2022 09:52]

Report message to a moderator

Re: NFHS-5, when to calculate new weights? [message #25730 is a reply to message #25727] Thu, 01 December 2022 13:49 Go to previous messageGo to next message
Parashara is currently offline  Parashara
Messages: 3
Registered: December 2022
Member
Thank you very much for the clarification regarding the privacy aspect!

However I still have one question, for there are women that answered question but are not in the dv module (maybe the're in the same household though?). So for those that dont have a dv weight, what do I do? Just leave them out since they can't be weighted properly?

I made a screenshot of those cases where individuals have a variable response but no dv weight.
Re: NFHS-5, when to calculate new weights? [message #25739 is a reply to message #25730] Fri, 02 December 2022 11:09 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 2664
Registered: February 2013
Senior Member

Following is a response from Senior DHS staff member, Fred Arnold:

The NFHS-5 sample was designed to provide information on sexual behaviour; HIV/AIDS, knowledge, attitudes and behaviour; life-style indicators; other health issues; and domestic violence only at the state level and the national level, while most other indicators are reported at the district level. Households that were not selected for the state module skip the section on domestic violence. In accordance with international ethics on confidentiality, only one woman in the eligible age group in a household can be selected for the domestic violence module. This is required to make sure that nobody else in the household even knows that these questions on domestic violence are being asked. If there is more than one woman in the household in the eligible age range (age 18-49) for the domestic violence module, only one woman in that household is randomly selected to be asked the questions on domestic violence.

Table 15.1 in the NFHS-5 national report shows that there are 72,056 women age 18-49 who were eligible for the domestic violence section. If you find more than that number of women who answered the DV questions, if they do not have a DV weight, you can leave them out of your analysis.

Re: NFHS-5, when to calculate new weights? [message #25754 is a reply to message #25730] Mon, 05 December 2022 13:01 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 2664
Registered: February 2013
Senior Member

Following is a response from Senior DHS staff members, Tom Pullum and Fred Arnold:

At the beginning of the domestic violence section, filter question WQ1102 determines whether or not privacy for the DV section was obtained. If privacy was not possible, the entire DV section was skipped. If privacy was obtained, the interviewer proceeded with the DV section. For those women who answered the DV questions, at the end of the DV section, the interviewer completed filter question WQ1138 which states DID YOU HAVE TO INTERRUPT THIS SECTION OF THE INTERVIEW BECAUSE SOME ADULT WAS TRYING TO LISTEN OR CAME INTO THE ROOM, OR INTERFERED IN ANY OTHER WAY? If the answer is YES, then the interviewer indicates who the person(s) was and for each type of person whether the person interrupted this part of the interview once or more than once. If there was any interruption, it does not mean that the person actually did listen to any of the questions since the interviewer was instructed to stop the domestic violence section completely until the person(s) interrupting or trying to interrupt the interview left. Presumably, none of the DV questions were ever asked when a person other than the respondent was trying to listen to any of the questions. During the interviewer training, it was made clear that no question in the DV section could be asked when someone else came into the room or tried to listen to a question.

It is DHS policy to omit responses that have a weight of 0. We would strongly recommend against ever constructing a weight that is not already in the data files.
Thank you [message #25755 is a reply to message #25754] Mon, 05 December 2022 13:10 Go to previous message
Parashara is currently offline  Parashara
Messages: 3
Registered: December 2022
Member
Thank you for these responses, they clarified a lot of things!

Previous Topic: Weighting for the Pakistan Special 2019 DHS on Stata
Next Topic: analysing change over time and weights normalisation
Goto Forum:
  


Current Time: Tue Jan 31 22:46:32 Coordinated Universal Time 2023