The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Child Health » Birth Weight (Recoding Birth Weight Variable)
Birth Weight [message #25722] Wed, 30 November 2022 22:03 Go to next message
Aamna is currently offline  Aamna
Messages: 11
Registered: May 2019

I am working on the Pakistan dataset (2017-2018). I am working on women's dataset (IR). I am examining children's birth weight (low) and antenatal care visits (outcomes) in one paper. I want to create a new variable for children's birth weight (m19_1 to m19_6). I want to have one birth weight variable (continuous) for all children mentioned in these six variables (m19_1 to m19_6).
Is it possible to create one variable, or do I need to work separately on the KR file for this purpose? The issue comes when the number of observations for IR and KR differs after I drop women with no birth (v208). I am not sure that I can have two different numbers of observations in one paper...
Can someone help me with coding if it is possible to create 1 variable out of these 6 variables because I am only getting error messages on my tries?
I appreciate the help.

Re: Birth Weight [message #25762 is a reply to message #25722] Tue, 06 December 2022 09:01 Go to previous message
Janet-DHS is currently offline  Janet-DHS
Messages: 773
Registered: April 2022
Senior Member
Following is a response from DHS staff member Tom Pullum:

The issue is whether you consider the child's birthweight to be a characteristic of the child or a characteristic of the mother. Either perspective is possible. If you regard it as a characteristic of the child, then yes, you would use the KR file. If you want it to be a characteristic of the mother, then it would be NA (not applicable) for women who did not have a child (or a child with a valid value of m19) in the past five years. If there are valid values of m19, you could use either the most recent one (m19_1) or an average.

If you want to take an average of the m19_* values, there's a helpful Stata command that would take into account the fact that the women had different numbers of children in the past 5 years. You could construct "bwt_1" that would recode the "not weighed" and "don't know" codes for b19_1 to a dot ("."), and repeat for _2 through _5. Then enter "egen bwt=rowmean(bwt_*)" to get the average of the valid values of bwt_1, bwt_2, etc.

Keep in mind that there may well be a bias, such that women who have valid values of m19 tend to have healthier children than women who do NOT have valid values of m19. This potential bias would also affect an analysis using children as units. Good luck!
Previous Topic: selecting variables
Next Topic: How to Merged HR dataset and KR dataset
Goto Forum:

Current Time: Mon Jul 22 06:51:08 Coordinated Universal Time 2024