Home » Data » Merging data files » Merging children in IR with children in PR (Issues merging children in IR with children in PR)
| Merging children in IR with children in PR [message #28802] |
Mon, 11 March 2024 03:55  |
Maiwenn Meyer
Messages: 1
Registered: March 2024
|
Member |
|
|
Hi,
I am trying to merge the children in the IR files with their long term outcomes in PR files. I have 2 files, one in which I appended all the African IR datasets that I have reshaped to have 1 entry per birth and not 1 entry per woman, and 1 in which I appended all the African PR datasets.
In the IR file, I removed all the births for which the line number of the child in PR is equal to 0 or missing. I have several issues :
1)Using IR : the variables v000, v001, v002, v003, b16, iyr and hhid do not uniquely identify the entries. I have 56,026 duplicates among which 55,940 are in SL5 (23,872) and SN6 (32,068). Could you help me understand why ? For the moment I removed those entries, they represent 2.12% of my dataset.
2)Using PR : I have 9,154 duplicates in terms of all variables. After removing those, I still have 74,032 duplicates in terms of hv000, hv001, hv002, hvidx, hv007, hv112, hhid, all coming from SN6. In the 5 separate files from which SN6 individuals come from ( SNPR6IFL, SNPR6DFL, SNPR6RFL, SNPR7HFL, SNPR7IFL) those variables uniquely identify the entries. Could you help me understand why it is not the case in the appended file ? For the moment I removed those entries, they represent 0.87% of my dataset.
3)When i try merging the 2 files using the following variables I do not recover all the births I should :
-hhid which I recovered in IR using the following code : gen hhid = subset(cased,1,12)
-v000 (hv000 in PR)
-v001 (hv001 in PR)
-v002 (hv002 in PR)
v003 (hv112 in PR)
-b16 (hvidx in PR)
-iyr (hv007 in PR)
-hhid
In the PR file, the line number of the mother (hv112) is not always available depending on the phase. So I separated the dataset into 2 datasets 1 with the entries for which hv112 is missing that I merged without hv112, and 1 in which v112 is not missing that I merged with hv112. Those two databases are exclusive, no individual is in both. Then I appended them and I recovered 2,138,529 births out of the 2,581,109 births that can be merged (with a b16 not missing and different from 0) in IR. Why can't I recover more births ?
Thank you for your help !
|
|
|
|
| Re: Merging children in IR with children in PR [message #28840 is a reply to message #28802] |
Mon, 18 March 2024 08:07  |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
In other forum posts we have recommended that you do such merges survey-by-survey. Your problem is that v000 (or hv000) is not a unique identifier for the survey.
The Senegal surveys illustrate the problem, but there are other countries for which 2 (and sometimes even 3) successive surveys have the same value of v000. v000 is 3-character string consisting of the 2-character country code plus the number of the coding manual for the survey. If the 2nd character is "6", for example, that will match with the 5th character in the filename (for example the "6" in "SNPR6D" but it will also match with the "6" in "SNPR6I", etc.
I wish that DHS recode files included a truly unique identifier for survey, but they do not. Usually, v000 serves the purpose, but not always, as you have found out. You need to do these merges survey-by-survey OR you need to add more columns to v000 OR you need to add a new variable (of your own design) that serves the purpose.
Note that the 6th character in the file name could help, but it is the version number and can change if the file is updated with some corrections.
|
|
|
|
Goto Forum:
Current Time: Sun Nov 24 21:02:34 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
