| Re: Children under 5 anthropometric failure and maternal bmi [message #28031 is a reply to message #28013] |
Fri, 03 November 2023 09:41   |
Janet-DHS
Messages: 855
Registered: April 2022
|
Senior Member |
|
|
Following is a response from DHS staff member, Tom Pullum:
The short answer to your question is yes. The conditions required for v445 in the IR file would apply to v445 in the KR file (v445, like most of the variables in the IR file, is copied into the KR file).
However, because your outcome is the nutritional status of the child, rather than the mother, I recommend that you keep v445 as it is coded, and if you are doing multivariate analysis, you include the pregnancy status of the mother as a covariate. (You could also include whether the woman had a birth in the past two months.)
For example, I just did a logit regression, with HTKR71, of "underweight" on the mother's BMI, with controls for the mother's pregnancy status and age group. The coefficient for v445 is negative and highly significant, meaning that mother's who weigh more are less likely to have underweight children. You could go into more detail on this relationship, of course, but I would argue that the inclusion of pregnancy status (as a control) makes it unnecessary to adjust BMI for whether the woman is pregnant.
gen WAZ=100*hw71 if hw71>-600 & hw71<500
gen underweight=0 if WAZ<.
replace underweight=1 if WAZ<-2
svyset v001 [pweight=v005], strata(v023) singleunit(centered)
svy: logit underweight v445 v213 i.v013 if v445>=1200 & v445<=6000
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
