|
|
| Re: the early child mortality and diseases [message #29608 is a reply to message #29586] |
Wed, 10 July 2024 07:27  |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
The VA file for the Bangladesh 2017-18 survey (BDVA7RFL.dta is the Stata version) was prepared in Bangladesh. It should be possible to merge it with the KR file, as you are trying to do, but I have been only partially successful in an effort to do that. In the VA file, you can do the following 4 lines:
gen v001=qncluster
gen v002=qnhnumber
gen v003=qnmother
gen bord=qnline
save this file as BDVAtemp.dta and then open the KR file, restrict to cases with b5=0, and do
merge 1:1 v001 v002 v003 bord using BDVAtemp.dta
This will successfully merge 357 out of the 471 cases in the VA file but there are another 114 cases that I can't merge.
In the VA file, I expected that qnline, which is labelled "line number of child" would be equivalent to bidx, but it is not. It is actually bord, the birth order of the child.
Also, in the VA file, qn302, the sex of the child who died, is not the same as b4, the sex of the child in the KR file. qn302 has 1 for female, 2 for male, the reverse of b4. There are 2 conventions on the coding of sex, and you just need to know which is being used.
I see that qnresidence is coded differently from v025. Lesson: you need to check the coding of covariates. Note also that the VA file has its own weight, vaweight, which may differ from v005--I did not check that.
We have not received other questions about this file, and it appears that DHS staff have never used it. I encourage you to continue, and regret that I cannot give a complete answer.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
