Weighting NFHS 3 PR file [message #27546] |
Thu, 31 August 2023 16:48 ![Go to next message Go to next message](/theme/default/images/down.png) |
arkapgbha.gun
Messages: 1 Registered: August 2023
|
Member |
|
|
Hi, I am trying to calculate the self-reported prevalence of TB using the SH30 variable. I have set up the survey design using the following code.
egen cluster_ID=group(hv024 hv001)
gen wgt = hv005/1000000
svyset cluster_ID [pweight=wgt], strata(hv023)
eststo: svy: regress sh30 ibn.hv024 , nocons
to generate weighted estimates.
However, the numbers I am getting are almost 10x multiples of the ones I expect to see. This has also been estimated in https://pubmed.ncbi.nlm.nih.gov/23185241/, and the estimates are closer to what I expect to see. Any idea what I am doing wrong?
Any help would be greatly appreciated.
|
|
|
Re: Weighting NFHS 3 PR file [message #27549 is a reply to message #27546] |
Fri, 01 September 2023 09:36 ![Go to previous message Go to previous message](/theme/default/images/up.png) ![Go to next message Go to next message](/theme/default/images/down.png) |
Bridgette-DHS
Messages: 3139 Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
You are using a variable (sh30) from the PR file, describing whether anyone in the household is believed to have TB. I looked at the abstract of the article. It is apparently based on questions asked of adult women and men, s31a in the IR file and sm31a in the MR file, about whether the respondent suffers from TB. The article accurately describes the cases in the IR and MR files, although care is required in pooling the two files because the NSFH's have only a 1/6 sample of men, and men appear to have higher prevalence.
You could do a check by just tabulating s30 in the PR file, s31a in the IR file, and sm31a in the MR file, with the appropriate weights, and taking into account that the question in the household survey is about anyone in the household. The data look plausibly consistent to me.
|
|
|
Re: Weighting NFHS 3 PR file [message #28161 is a reply to message #27549] |
Mon, 20 November 2023 22:03 ![Go to previous message Go to previous message](/theme/default/images/up.png) |
dejawitting
Messages: 1 Registered: November 2023
|
Member |
|
|
The PR file contains a variable (sh30) that indicates whether there is a belief that someone in the family has tuberculosis. While perusing the piece, I noticed its abstract. Seemingly, it is based on TB-related questions answered to both men and women as adults (s31a in the IR file and sm31a in the MR file), respectively. Although the article provides an accurate description of the instances in the IR and MR files, it seems that men have a higher prevalence, and the NSFH's have only a 1/6 sample of men, therefore care must be taken when merging the two files.
[Updated on: Thu, 14 March 2024 03:28] Report message to a moderator
|
|
|