| Re: Domestic violence weight for NFHS 5 [message #26013 is a reply to message #26001] |
Fri, 27 January 2023 07:26   |
Janet-DHS
Messages: 888
Registered: April 2022
|
Senior Member |
|
|
Following is a response from DHS staff member, Tom Pullum:
All surveys that used the domestic violence module include d005. It is a modification of v005 that adjusts for any additional nonresponse that may apply to the module, beyond any nonresponse to the women's interview as a whole.
If you do a tabulation or a regression that includes a variable from the DV module, women who do not have a valid response to the DV variable will be omitted because cases with NA values for any of the variables are always dropped. In effect, d005 adjusts for those women who were eligible for the module but did not respond, to give unbiased estimates, under an assumption that the nonresponse is random (of course it is not random, but that's another story). You should use d005 in the svyset command for any tabulations, etc., that include a DV variable.
If the model does NOT include a DV variable, use the usual svyset with v005. If you accidentally used d005 in svyset when you should be using v005, you will drop women who did not respond to the DV module, because they will have NA (a dot) for d005.
In the svyset command, or with any use of pweight, you never need to divide by 1000000. You can if you want, but it makes no difference. To confirm this, do a run with that divisor and a run without, and compare the results. That's because Stata always automatically re-normalizes pweights to have a mean of 1.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
