Re: duplicated almost blank variables [message #18669 is a reply to message #18666] |
Fri, 24 January 2020 10:40 |
Bridgette-DHS
Messages: 3207 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
You are probably using the IR file, which has women as units. Each woman has m70_1 through m70_6 for up to 6 children born in the past five years. If a woman had, for example, two children in the past few years, then only m70_1 and m70_2 will be filled in, and m70_3 through m70_6 will be blank. The blank just means "not applicable". For some of these subscripted variables, only the first one (for the most recent birth) will have any data in it, because the question was only asked about the most recent birth. I strongly recommend that you switch to the KR file, which includes almost all of the mother's data but has a separate record for each child. The variable m70_1, for example, becomes m70, with a new variable on the record called bidx; for the most birth, bidx=1. For the next most recent birth m70 will be the same as m70_2 and will have bidx=2.
Using the KR file, you could have, for example,
syvset [pw=weight], psu(v021) strata(v023) singleunit(centered)
svy: regress h3 m70
|
|
|