Re: Preceding child sex and survival status [message #22178 is a reply to message #22144] |
Tue, 09 February 2021 10:22 |
Bridgette-DHS
Messages: 3203 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
The Stata program I sent yesterday was intended to be the basis of a program to do what you want to do. I have added to it. The attached version includes the sex of the previous sibling (b4_sib) and the survival status of the previous sibling (b5_sib). It also includes "b11_test" which shows how b11 is calculated. It matches b11 in the data file EXCEPT for multiple births. The DHS calculation of the birth interval preceding a multiple birth goes back farther in the birth history. I will not take the time to bring in those cases.
The calculation of b11 changed during DHS-7. Previously, it was just the difference between b3 (century month code of birth) for successive births. Now we calculate b11 with b18, which is the century day code of the birth. You calculate the difference in cdc's, divide by 365.25/12 (the average number of days in a month), and truncate to an integer.
As I said, this calculation does not apply to births with b0>1. However, I strongly recommend that you omit those births, as well as births with b0=1 (the first in a multiple birth) because survival is much worse for multiple births. We have found that usually the strongest single predictor of child mortality is being part of a multiple birth. They tend to be premature low birthweight and they have greater competition for maternal resources.
Note that this program runs on the BR file, not the KR file, because the siblings may have been born earlier than the reference period for the KR file. It is then reduced to be equivalent to the KR file with a restriction to b19<60. You would merge this file with the KR file to bring in the covariates that are in the KR file.
[Updated on: Tue, 09 February 2021 10:24] Report message to a moderator
|
|
|