The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Child Health » Maternal Education in Child Recode File
Maternal Education in Child Recode File [message #3982] Sun, 15 March 2015 13:22 Go to next message
jmclenna is currently offline  jmclenna
Messages: 3
Registered: October 2013
Location: Canada
Member
This is probably an obvious question but ... is maternal education (and maternal age) in the child recode file or does one have to merge datasets? In the child recode file I see "respondent's age", but I assume that not all "respondents" are necessarily the mothers. I also see educational variables V106 and V107 in child recode file which I assume refers to the "respondent" and again can't assume this is the mother.

With regard to "respondent", is there a variable in the child recode that tells you who the respondent is? I see a variable that asks relationship of the respondent to household-head but the response options are lumped (e.g., son/daughter) so not able to tease this apart on this variable. Thanks for any advice. John
Re: Maternal Education in Child Recode File [message #4008 is a reply to message #3982] Tue, 17 March 2015 09:12 Go to previous messageGo to next message
duke2015
Messages: 27
Registered: March 2015
Location: United States
Member
Hi there,

I actually have the same question. You can't really assume that the education variable in the child dataset is from the mother. I think you have to link it to the woman's individual file.

Did you ever receive a response to this? I realize it's an old thread.

Thanks!
Re: Maternal Education in Child Recode File [message #4011 is a reply to message #3982] Tue, 17 March 2015 10:42 Go to previous messageGo to next message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
Dear User,
Here is a response from one of our subject experts, Dr. Tom Pullum:
Quote:
The cases in the KR file are children born in the past five years and the v variables describe their mothers. A woman in the IR file will only appear in the KR data if she had a child in the past five years, and she will appear more than once if she had more than one child. The v variables in the KR file are the same as the v variables in the IR file--that is, they refer to the women age 15-49 who interviewed about their fertility, etc. For example, v003, "respondent's line number" is the mother's line number (in the household survey) and v012, "respondent's current age", is the mother's current age.
There is definitely room for confusion, because in the PR file, in which every person in every household is a case, hv003 is the line number of the person who provided the household information. That will be the same person and line for everyone in the household. That's the "household respondent." This is not necessarily the household head. The household is the person with hv101=1. Usually that's the person on line 1, but not always. In the PR file, each person's line number is called hvidx.
There are various ways to link the files. For example, in the KR file, the line number of the child in the PR file is b16. However, if you are using the IR or KR file and want the line number of the household respondent--I think that's what you are asking about--it's a little more complicated. You have to do a merge.
Here are the lines that you would use to merge the data for the household head or the household respondent to the data in the KR (or IR) file. Let me know if you have any questions.

* LINES TO ATTACH DATA FROM THE HOUSEHOLD HEAD TO THE KR (OR IR) DATA
* open the PR file
* save record for the household head
keep if hv101==1
* keep the PR variables you want
rename hv001 v001
rename hv002 v002
sort v001 v002
save temp.dta, replace
* open the KR (or IR) file
sort v001 v002
merge v001 v002 using temp.dta, replace 
keep if _merge==3

* LINES TO ATTACH DATA FROM THE HOUSEHOLD RESPONDENT TO THE KR (OR IR) DATA
* open the PR file
* save record for the household respondent
keep if hvidx==hv003
* keep the PR variables you want
rename hv001 v001
rename hv002 v002
sort v001 v002
save temp.dta, replace
* open the KR (or IR) file
sort v001 v002
merge v001 v002 using temp.dta, replace 
keep if _merge==3

Re: Maternal Education in Child Recode File [message #4030 is a reply to message #4011] Thu, 19 March 2015 11:01 Go to previous messageGo to next message
duke2015
Messages: 27
Registered: March 2015
Location: United States
Member
That's extremely helpful, thank you so much for the reply!

A few follow up questions:

1) How can I confirm, using the data, that the v variables in the KR file are responses of the child's mother? Is there a line number or something I can look at to say, this is definitely the mother of the child I'm analyzing?

2) If the mother can appear more than once in the KR file since the child is the unit of analysis, does that mean that the same mother is counted twice if any averages are taken? For example, if there is a mother of 2 children and she has 2 years of education, will her '2 years of education' be counted twice towards the entire average?

3) Is it possible that there is more than 1 mother per household? From what I understand, the DHS interviews every eligible woman, so is it possible to have more than 1 family per household? Can I use the KR file to somehow verify that the mother and child are connected?

Thank you in advance!
Re: Maternal Education in Child Recode File [message #4034 is a reply to message #4030] Thu, 19 March 2015 15:31 Go to previous messageGo to next message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
Dear User, Response from one of our subject experts, Dr. Tom Pullum,
Quote:
The v variables in the KR file always refer to the mother. Apart from the height and weight measurements that are obtained during the household interview, all the information about the children in the KR file comes from the woman's interview. The mother is the respondent and she is reporting on her biological child.
If you want, say, the distribution of v106 (level of education) for all women, you should get that from the IR file, which has one case per woman. If you calculate the distribution of v106 in the KR file, it will refer to the mothers of children born in the past five years, and will repeat the mothers who had more than one child. I would say that it SHOULD do that, because the units are children and you are using the education of the mother as a characteristic of the child. However, if you did not want to repeat, and just wanted the distribution of women who had a birth in the past five years, you could insert a line such as "keep if hidx==1" or do "tab v106 if hidx==1".
The KR file can definitely include more than one mother from the same household. They will have the same values of v001 and v002 (cluster and household id) but different values of v003. To get the distribution of the number of women per household who had at least one child in the past five years, use these lines:

* open a KR file
* instead of one record per child, reduce to one record per mother
keep if hidx==1
gen mothers_per_hh=1
collapse (sum) mothers_per_hh, by(v001 v002)
tab mothers_per_hh

Quote:
For example, with the Nigeria 2013 survey, using NGKR6AFL.dta, I just got the following (unweighted) distribution. It shows that out of the 17,723 households that included at least one such mother, 87% had exactly one, 12% had two, and 1% had more than two, ranging up to a maximum of six.

mothers_per |
        _hh |      Freq.     Percent        Cum.
------------+-----------------------------------
          1 |     15,474       87.31       87.31
          2 |      2,053       11.58       98.89
          3 |        175        0.99       99.88
          4 |         19        0.11       99.99
          5 |          1        0.01       99.99
          6 |          1        0.01      100.00
------------+-----------------------------------
      Total |     17,723      100.00

Re: Maternal Education in Child Recode File [message #5652 is a reply to message #3982] Sat, 20 June 2015 05:53 Go to previous messageGo to next message
ss11 is currently offline  ss11
Messages: 2
Registered: May 2015
Member
Hello

I am having same problem. How to calculate the maternal education with diarrhea in under five children in SPSS? The variable of education in SPSS format isn't clearly defined.
Re: Maternal Education in Child Recode File [message #6795 is a reply to message #5652] Fri, 17 July 2015 10:53 Go to previous messageGo to next message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
Dear User,
Do you still need assistance with this post?
Thank you!
Re: Maternal Education in Child Recode File [message #8950 is a reply to message #6795] Mon, 18 January 2016 11:57 Go to previous messageGo to next message
duke2015
Messages: 27
Registered: March 2015
Location: United States
Member
I'm sorry to bring up an old post, but is there any chance you could explain what "keep if hidx==1" does? I see that hidx is the index to birth history, but I don't understand why you would use this line.

Thanks!
Re: Maternal Education in Child Recode File [message #9088 is a reply to message #8950] Wed, 10 February 2016 09:27 Go to previous messageGo to next message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
Dear User,
Here is a response from Dr. Tom Pullum:
Quote:
In the KR and BR files, the "b variables" give the information that is collected in the birth history. These variables are b1 through b16, bord, and bidx. bord is the birth order of the child in the birth history, starting with bord=1 for the woman's first birth, bord=2 for her second birth, etc. bidx is the REVERSE birth order of the child. That is, bidx=1 for the most recent birth, bidx=2 for the next most recent birth, etc. The maximum value of bidx and the maximum value of bord will be the same, and will be the same as v201, the number of children ever born.

The KR file is a child file like the BR file but is limited to the children born in the last five years or whatever the interval was for the health questions. It has all the b variables (including bidx) but also includes hidx, which is an exact copy of bidx for the children in the KR file. If you do "tab hidx bidx" in a KR file you will see that they are the same and you will usually see that there are only a few children with hidx>2 and very few children, but some, at hidx=5. (Few surveys will have any children with hidx=6.)

Many of the health questions are only asked about the youngest child born in the past five years, that is, the child with hidx=bidx=1. For those questions, children with bidx>1 will be given the "not applicable" code, "." in Stata. You originally asked why so many children were missing on such a question. I was simply replying that almost all of the children you thought were missing were actually children other than the youngest child, and the question was not asked about them. In this case, as is general policy for DHS, "." does not mean "missing". It means "not applicable".

To analyze data about the youngest child, you do not need to impose a restriction to children with hidx=1, but you do need to know that the responses only apply to those children, and for those variables there is a potential bias. There is a tendency for such children to be the ONLY child born in the past five years, and to be born to women who have lower fertility. Their mothers tend to have the characteristics of family planning users (among other things). Thus, for example, the children with hidx=1 are usually more likely than other children to have been born in a health facility than children with hidx>1. I recommend that you look at a recent methodological report by Shea Rutstein regarding this potential bias: https://www.dhsprogram.com/pubs/pdf/MR14/MR14.pdf.


Re: Maternal Education in Child Recode File [message #10909 is a reply to message #9088] Tue, 04 October 2016 23:21 Go to previous messageGo to next message
Olutosin is currently offline  Olutosin
Messages: 5
Registered: October 2016
Location: Nigeria
Member
Alright sir

Olutosin
Re: Maternal Education in Child Recode File [message #10910 is a reply to message #4034] Tue, 04 October 2016 23:31 Go to previous messageGo to next message
Olutosin is currently offline  Olutosin
Messages: 5
Registered: October 2016
Location: Nigeria
Member
Got that sir because am from Nigeria it might be useful for my project work thanks sir

Olutosin
Re: Maternal Education in Child Recode File [message #14560 is a reply to message #4008] Sat, 21 April 2018 19:01 Go to previous messageGo to next message
kingx025 is currently offline  kingx025
Messages: 95
Registered: August 2016
Location: Minneapolis. Minnesota
Senior Member
I think it is helpful to keep in mind how the information in the KR file (and BR file) was collected. The women's survey is administered to women of childbearing age, and as part of their survey, the woman answers questions about HER OWN children under age 5 and HER OWN births, for up to 20 births. The KR file is really a reordering of information that was collected from the child's own mother, about her own children, so if there is a variable relating to mother's (respondent to women's survey) education, it is indeed for for surviving biological mother of the child in the KR file. Much of the information about the mother (who answered questions about her own children) is carried over to the KR file.

However, some information collected from mothers--such as their experience with intimate partner violence--doesn't get copied over to the KR files. If you want all the information about a child's mother as part of the child record, you can avoid doing the merging to the IR file yourself by getting your data from IPUMS-DHS (which has merged the mother's data and the household data to the child records from the KR files). IPUMS-DHS (available at idhsdata.org) currently covers 28 countries.

If you are using data on children from the household file (PR file), the respondent would be whoever answered questions for the household survey, and orphans whose mothers are dead and not in the IR files are included. But you don't need to worry about getting information (say on mother's education) for someone other than the child's own mother if you restrict your analysis to children included in the KR child recode files.

Let me know if this answer isn't clear.

Miriam King



Dr. Miriam King
IPUMS-DHS Project Manager (www.idhsdata.org)
Re: Maternal Education in Child Recode File [message #14782 is a reply to message #14560] Sat, 05 May 2018 08:36 Go to previous message
Hassen
Messages: 111
Registered: April 2018
Location: Ethiopia,Africa
Senior Member
Dear Dr.Tom Pullum,Liz-DHS,Bredgette-DHS Thank you very much for building up of our Brain Architecture to think critically and doing more!!

Hassen Ali(Chief Public Health Professional Specialist)
Previous Topic: NUMBER OF CHILDREN UNDER-FIVE
Next Topic: Child Health Indicators in Eastern African Countries
Goto Forum:
  


Current Time: Fri Apr 16 15:23:37 Coordinated Universal Time 2021