Home » Countries » Other countries » Discrepancies in Observations (Children's File 2014 Ghana DHS)
| Re: Discrepancies in Observations (Children's File 2014 Ghana DHS) [message #10091 is a reply to message #10062] |
Fri, 24 June 2016 09:51  |
Bridgette-DHS
Messages: 3230
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Stata Specialist, Tom Pullum:
Yes, the Guide to DHS Statistics does recommend against using weights in in regressions and correlations, etc. This is an old recommendation and it will not appear in the next revision of the Guide. We now recommend that all models take account of the weights, clustering, and stratification with the svyset and svy commands (or whatever is appropriate if you are not using Stata).
For your question about analyzing the employment status of women, I think your research question is whether the number of living children under five affects a woman's employment status. If so, your cases would be women rather than children and should probably include women who have NO children under five as a comparison group. It may be easier to work with the IR file. The KR file will include women who had a birth in the past five years (with repetitions of women with more than one such birth, as you say). The BR file will include women who have had ANY births, with repetitions of women for each birth.
If you use the IR file, you will get all women, regardless of whether or when they ever had children. v208 is number of births in the past five years. v220 is number of living children (plus current pregnancy) and v213 is current pregnancy status. I don't think there is a variable that is specifically the number of living children under five, although maybe I am missing it. You could construct such a variable, either working from the b variables that are on the woman's record or working from a merge with the KR or BR file. Please think about specifically what it is about the children that you want, and then try to construct it. Let me know if you have difficulty constructing it. I strongly recommend, however, that you include women with NO children, whether living or under 5, in your model. I could imagine other relevant categories, such as no children under five but children OLDER than five, or no children ever born, etc. Hope I have correctly understood what you are trying to do.
|
|
|
|
Current Time: Wed Feb 5 15:43:14 Coordinated Universal Time 2025
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
