Re: Comparison Between Birth weight Variable m19 & m18 [message #20498 is a reply to message #20488] |
Thu, 12 November 2020 09:13 |
Bridgette-DHS
Messages: 3214 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
A more efficient way to do your recode would be with "gen Birthweight=m19/1000" and "replace Birthweight=. if Birthweight>5". The way you did it is tedious, has a high risk of typing errors, and is not generalizable to other surveys because you had to run the full distribution of observed values first.
If you want to truncate the distribution at some high level, you can do so. 5 kilos (11 pounds) is certainly a high value, but I am sure there are accurately recorded values higher than that.
You also asked this: "How can I make a comparison between the measured birth weight (m19) & the perceived size of the baby at birth (m18)." You could do "tab m18 [iweight=v005/1000000], summarize(Birthweight)", for example. You can also do "graph box Birthweight, over(m18)", adding options for weighting and appearance.
You expect an association but not a high level of agreement between these two variables. Weight and size are different things. The estimate of size is subjective. It depends on how many newborn babies the respondent has seen and on her comparison group. Perhaps most of the newborns she has seen were low birthweight, in which case that could be her "average".
As you know, in the latest Pakistan survey, the one you are using, m19 is "not weighed at birth" or "DK" for about 80% of births. The children who were weighed were mostly born in a facility and are not representative of all births.
[Updated on: Thu, 12 November 2020 10:10] Report message to a moderator
|
|
|