The DHS Program User Forum: Weighting data » Survey weight for a Merged IR file of NFHS-4 and NFHS-5

Home » Data » Weighting data » Survey weight for a Merged IR file of NFHS-4 and NFHS-5 (Calculating survey weight for a vertically merged Dataset. )

Show: Today's Messages :: Show Polls :: Message Navigator

Survey weight for a Merged IR file of NFHS-4 and NFHS-5 [message #28038]

Sat, 04 November 2023 03:14

dnameispaone
Messages: 19
Registered: February 2019

Member

Hi There,

I am undertaking an analysis that requires me to vertically merge the NFHS-4 and NFHS-5 datasets.
The merged data has about 1.3 million observations (0.7 million (NFHS-5) and 0.6 million (NFHS-4).

I have no idea how to calculate the survey weight for the merged dataset.
I searched for older posts on the forum, but most of them were concerned with merging datasets from multiple countries.
I am not sure the same methodology applies for merging two datasets from the same country.
Moreover, I would appreciate it if anyone could provide a reference for citing in my research project.
Or tag a post on this forum that provides guidance on the same topic.

Thank You!.

Report message to a moderator

Re: Survey weight for a Merged IR file of NFHS-4 and NFHS-5 [message #28050 is a reply to message #28038]

Mon, 06 November 2023 09:41

Bridgette-DHS
Messages: 3043
Registered: February 2013

Senior Member

Following is a response from Senior DHS staff member, Tom Pullum:

The forum discussions about combining surveys from different countries would also apply to combining surveys from the same country. I believe your question about the weights is only relevant if you plan to produce estimates or tables that pool the two surveys--for example, to calculate the mean of v201 (children ever born) with the two surveys combined, rather than separate. We advise against doing that. But if you want to calculate the difference between the mean of v201 in the first survey and the mean of v201 in the second survey, and test the statistical significance of that difference, for example, you can indeed do that a little more easily with a combined file, and for that you would leave the weight variable (v005) as it is. No need to re-scale it.

The NFHS-4 and -5 are huge surveys. A disadvantage of combining them is that programs will take a little longer to run.

Report message to a moderator

Previous Topic:	weighting urban rural divide , multiple countries years
Next Topic:	Correct use of weights for subsamples

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sat Apr 27 02:57:02 Coordinated Universal Time 2024