| Running a regression with the KR file [message #8677] |
Mon, 30 November 2015 06:45  |
geniusmuso
Messages: 11
Registered: November 2015
|
Member |
|
|
1. Please since all the variables in the KR files are related to children, I want to know if running a regression with some as dependent variables and some as independent variables is in order especially where the dependent variables are about children's immunization, choice of healthcare providers for fever and diarrhea, health status of children(fever, anemia, cough and diarrhea), vitamin A, iron supplements, deworming, insecticide nets usage for children and household iodine test.
2. Also i would want to be certain if in the 2014 Ghana GDHS, hw 57 represents anemic level of children
3. Please i would also want to know if dividing a variable say v0005 by 1000000 is the right way to create its weight in Stata?
Thank you
|
|
|
|
| Re: Running a regression with the KR file [message #8709 is a reply to message #8677] |
Thu, 03 December 2015 17:11  |
Trevor-DHS
Messages: 805
Registered: January 2013
|
Senior Member |
|
|
1. It is very common to use variables related to children both as dependent and independent variables and we do it regularly. You should be aware though that any results would just describe a relationship between variables and not provide causality. You should also be careful of temporal aspects - some variables are current status, some may relate to illnesses in the past two weeks, and others may be further back in time (e.g. vaccinations or postnatal care)
2. hw57 is the anemia status of children born to interviewed women in the last 5 years, however, table 11.7 in the Ghana DHS 2014 report is based on all children living in the household and uses hc57 which can be found in the Household members/Person's Recode (PR) file.
3. Yes, sample weights should be divided by 1000000 before applying. So you would use v005/1000000 when weighting data from the IR or KR file, or hv005/1000000 when weighting data from the PR, for example.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
