The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Merging data files » Merge BDSQ7RDT and BDVA7RDT with the household data file (Merging verbal autopsy file, service availability file with the household file for bangladesh 17-18)
Merge BDSQ7RDT and BDVA7RDT with the household data file [message #28573] Thu, 01 February 2024 01:04 Go to next message
farihakabir148 is currently offline  farihakabir148
Messages: 3
Registered: February 2024
Member
I am trying to merge BDSQ7RDT, BDVA7RDT, and household data file, these three files from the 2017-2018 survey of Bangladesh, but I am failing to find a unique identifier. Can you please help me merge these?

[Updated on: Thu, 01 February 2024 01:05]

Report message to a moderator

Re: Merge BDSQ7RDT and BDVA7RDT with the household data file [message #28581 is a reply to message #28573] Thu, 01 February 2024 11:04 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3000
Registered: February 2013
Senior Member
Following is a response from Senior DHS staff member, Tom Pullum:

I recommend that you think of this as merging the SQ and PR data with the VA data. When you use the VA data, your interest is in the children (up to 8) for whom the mother provides information. The lines given below will do the merge, with one record per mother, as in the VA data. Your next step would probably be to reshape this file into a file that has one record per child.

It is not necessary to include "m:1" or 1:m" in the merge command. Let us know if you have questions.

* BD7R: Merge the PR, VA, and SQ files

* Specify workspace
cd e:\DHS\DHS_data\scratch

* Prepare SQ file for merge
use "...BDSQ7RFL.DTA", clear 
gen cluster=coclust
sort cluster
save SQtemp.dta, replace

* Prepare PR file for merge
use "...BDPR7RFL.DTA", clear 
gen cluster=hv001
gen hh=hv002
gen line=hvidx
sort cluster hh line
save PRtemp.dta, replace

* Open the VA file and prepare for merge 
* The VA file has up to 8 records per mother, 1 for each reference child,
use "...BDVA7RFL.DTA", clear 
gen cluster=qncluster
gen hh=qnhnumber
gen line=qnmother 
sort cluster hh line
merge cluster hh line using PRtemp.dta

* Reduce to the mothers who are in both the VA and PR file
tab _merge
keep if _merge==3
drop _merge

* Prepare for merge with the SQ file, which has one record per cluster
sort cluster
merge cluster using SQtemp.dta
tab _merge
keep if _merge==3
drop _merge

* Save this file, which has one record per mother, and reshape for one record per child

Re: Merge BDSQ7RDT and BDVA7RDT with the household data file [message #28583 is a reply to message #28581] Thu, 01 February 2024 14:30 Go to previous messageGo to next message
farihakabir148 is currently offline  farihakabir148
Messages: 3
Registered: February 2024
Member
Thank you so much for responding in such a short notice.

I have a few other queries.

1. The last line you mentioned ' Save this file, which has one record per mother, and reshape for one record per child' how do I reshape the data for one record per child?

2. I plan to look at the impact of exposure to salinity on pregnant mothers during their utero period, on hypertension, swelling of feet, and anthropogenic features of children born to these mothers like height stunt.

I did some calculation of conception date and hypertension data and I found that hypertension data is available only for 1960? (I used the household members data file)

can you please tell me what would be the correct way to find the conception date of mothers, how to form hypertension variable?
Re: Merge BDSQ7RDT and BDVA7RDT with the household data file [message #28584 is a reply to message #28583] Thu, 01 February 2024 16:22 Go to previous messageGo to next message
farihakabir148 is currently offline  farihakabir148
Messages: 3
Registered: February 2024
Member
I followed through all your steps and yer, the attached screenshot saying "(you are using old merge syntax; see [D] merge for new syntax)
variable cluster does not uniquely identify observations in the master data)" keeps showing up.

which one is a unique identifier here? why isn't it working??

[Updated on: Thu, 01 February 2024 16:23]

Report message to a moderator

Re: Merge BDSQ7RDT and BDVA7RDT with the household data file [message #28595 is a reply to message #28584] Fri, 02 February 2024 08:07 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3000
Registered: February 2013
Senior Member

Following is a response from Senior DHS staff member, Tom Pullum:

Your study is potentially very complex and I can't provide much more help, but I'll make some comments.

The children in the VA file are children who died. You need children who survived, as well as children who died, in order to identify characteristics that are associated with child survival. It could be better to work with the KR file, which has children born in the past 5 years as the cases. The variable b5 in that file tells whether or not the child survived to the date of interview. If the child died, b7 is the child's months of age at death. The KR file includes many characteristics of the child and birth as well as most characteristics of the mother that are in the IR file. If there are some other variables in the PR or IR file that you need, and they are not in the KR file, you can get them with a merge. You can also merge the SQ file with the KR file. The Stata lines I sent earlier would only need simple modifications to do these merges.

If you want to add any variables from the VA file to the children (in the KR file) who died, please identify which variables you think are important and I can provide more detail on the reshaping I referred to.

Where will you find the data on salinity, hypertension, etc., of mothers when they were pregnant? You cannot safely assume that those kinds of biometric characteristics measured at the time of the survey can be extrapolated back to the time of the pregnancy.

DHS-8 surveys have a new variable, p20, which is the estimated duration of a pregnancy. For earlier surveys, including the 2017-18 BD survey, you would have to use 9 months, a constant, as the duration of the pregnancy, and estimate the month of conception as the month of birth minus 9.
Re: Merge BDSQ7RDT and BDVA7RDT with the household data file [message #28596 is a reply to message #28595] Fri, 02 February 2024 08:26 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3000
Registered: February 2013
Senior Member

Following is a response from Senior DHS staff member, Tom Pullum:

You are describing an annoying (to me) option in the syntax of the "merge" command. Stata wants you to put "m:1" or "1:m" or "m:m" right after the word "merge". (Omit the quotes.) For example, when you merge the SQ file with the PR file, using "cluster", you have many cases in the PR file that will match with a single cluster in the PR file. This requires "m:1" or "1:m". When I only want the cases with _merge=3, the matches, I don't want to think about WHICH of "m:1" or "1:m" I have to use!

The program ran fine for me. I am still using Stata 16. If you are using a later version, you can replace "merge" with "version 16: merge". The merge command changed well before version 16, and you will still get a warning, but version 16 does not FORCE you to specify "m:1", etc. You just have to know the structure of the data and look at the constructed variable "_merge" and keep the cases you want.
Previous Topic: How to link child health outcome with mothers' characteristics
Next Topic: merge IR MR AND PR
Goto Forum:
  


Current Time: Sun Mar 3 12:57:31 Coordinated Universal Time 2024