| RWANDA dataset merge [message #4205] |
Thu, 16 April 2015 17:23  |
cne2109
Messages: 3
Registered: April 2015
Location: United States
|
Member |
|
|
Could anybody provide some help merging child datasets from Rwanda 1992, 2000 and 2005 with the new standards of WHO (datasets with recoding available) for stunting and wasting
1992 is fine, we have the same number of rows in each dataset and matches perfectly, but 2000 is different, the child dataset has more rows than the WHO one, and if we drop the not measured kids or dead (from both) doesn't match either. We merge the kids on hwhhid whline in the who one and do a substring for the child with caseid and b16 to merge the children with those two variables, we ended up with above 400 cases from the USING (WHO) dataset not matched. Plus we compared kids with the same age and hight and they were some differences in the stunting z-scores...
Any leads on this??
Thank you in advance!!
Cecilia
[Updated on: Thu, 16 April 2015 17:23] Report message to a moderator |
|
|
|
| Re: RWANDA dataset merge [message #4206 is a reply to message #4205] |
Thu, 16 April 2015 17:30  |
Reduced-For(u)m
Messages: 292
Registered: March 2013
|
Senior Member |
|
|
Quick thoughts, but I suspect a DHS person will chime in with some details:
1 - There is a sampling frame issue regarding "kids born to survey mothers" or "kids in the household", which are not always the same. Could that explain some of it?
2 - the WHO sample would likely drop a few different kids than the old standards, if in one or the other the kid is less than the "acceptable range" (-5 or -6 HAZ).
3 - You can just re-calculate your own z-scores using the WHO measures using the Stata package zscore06 and using gender, raw height, and age-in-fractions-of-months (if there is a given birth day and not just birth month). Then you should be able to tell why some are there and some aren't, or at least test (2) above.
Hope one of those helps a little bit, or at least gives a real DHS person some concrete ideas to shoot down.
|
|
|
|
| Re: RWANDA dataset merge [message #4210 is a reply to message #4206] |
Thu, 16 April 2015 23:42 |
Trevor-DHS
Messages: 789
Registered: January 2013
|
Senior Member |
|
|
Reduced-For(u)m is right in point 1 (actually in all 3, but point 1 is where the difference comes in). In the early DHS surveys (including Rwanda 1992) anthropometry was collected for children of interviewed mothers ("kids born to survey mothers"), but in later surveys anthropometry was collected for children listed in the household schedule. The children measured in the later surveys will match with children listed in the PR file, but will not match completely with children in the KR file for 2 reasons - 1) there are children in the KR file who do not live in the household, but are children of the respondent, and 2) there are children living in the household whose mother does not live in the household.
There is a variable in the HW datasets that indicates where the original data were collected. hwlevel contains this information, and code 1 is used for data collected for children of the household, and code 2 for children of interviewed women.
[Updated on: Thu, 16 April 2015 23:43] Report message to a moderator |
|
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
