The DHS Program User Forum: General Data Questions

Home » Data » General Data Questions » similar variable

Show: Today's Messages :: Show Polls :: Message Navigator

similar variable [message #10390]

Sun, 24 July 2016 16:05

sultanatonni@gmail.com is currently offline

sultanatonni@gmail.com
Messages: 4
Registered: July 2016

Member

Sir/Madam

In DHS dataset,2014, household member file (PR) has women body mass index-HA40 variable and individual record file (IR) also has maternal body mass index-V445. Are variables HA40 and V445 same?

I understood the variable HV270-wealth index. what does it mean by HV271-wealth index factor score?

Thanks

TANWI

Report message to a moderator

Tue, 26 July 2016 09:45

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Senior DHS Stata Specialist, Tom Pullum:

Yes, those two BMI variables are the same. The height and weight measurements for women and children are taken during the household survey. The anthropometry indicators are therefore in the HR and PR files, but for convenience they are also in the IR file. The values for children are also in the KR file.

hv270 gives the wealth quintiles, with 1 for lowest and 5 for highest. hv271 is the continuous wealth index. The wealth quintiles are constructed from hv271 by identifying 4 cut points or values such that an equal number of weighted de jure household members are in each quintile. For most purposes you will only use hv270. But if, say, you wanted to construct terciles (three equal sized categories), you could do that using hv271.

Report message to a moderator

Wed, 02 May 2018 15:33

Hassen
Messages: 121
Registered: April 2018
Location: Ethiopia,Africa

Senior Member

Dear Bridgette-DHS Thank you very much for your Investment to Our Brain Architecture in order to make us motivated to read and do more!!
Having said this,my Question is How can I construct terciles (three equal sized categories from HV271 Using SPSS?
Thank you in Advance!!

Hassen Ali(Chief Public Health Professional Specialist)

Report message to a moderator

Thu, 03 May 2018 08:42

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Senior DHS Stata Specialist, Tom Pullum:

I cannot give you the specific steps in SPSS, but here is the strategy. First, the wealth quintiles are constructed by identifying four values of the continuous wealth index, hv271, such that the weighted number of de jure (hv102=1) household members in the PR file will be the same in each of the five intervals. (This is one of only a few instances in which de jure residence is used, rather than de facto residence). I suggest that you open the PR file, reduce it to the cases with hv102=1, sort on hv271, and identify the value of hv271 where hv270 shifts from 1 to 2, from 2 to 3, from 3 to 4, and from 4 to 5. Then apply the procedure to identify the TWO points that will give you THREE equal-sized categories, i.e terciles. Note that everyone in the same household will have the same value of hv271. There will be ties at the breakpoints. You will see how those ties are resolved for the quitiles, hv270, and apply the same principle for constructing terciles.

Report message to a moderator

Sat, 05 May 2018 02:01

Hassen
Messages: 121
Registered: April 2018
Location: Ethiopia,Africa

Senior Member

Thank you very much!!

Hassen Ali(Chief Public Health Professional Specialist)

Report message to a moderator

Mon, 04 June 2018 06:19

Mayank_Ag
Messages: 22
Registered: May 2018
Location: Hyderabad

Member

I am confused regarding the variables ha40 and v445. I checked if they are similar. I found that only 2 obs were matching. Can you please explain this ?
I am using a merged file of PR and KR for India.

TIA

Report message to a moderator

Fri, 08 June 2018 07:35

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Senior DHS Stata Specialist, Tom Pullum:

HA40 is BMI for women age 15-49 in the PR file. v445 is BMI for the same women in the IR file. They have to match. Perhaps you did a merge of the IR and PR files and the merge was not successful? They would not normally appear in the same file. In any case, in all surveys they are the same variable.

All of the "ha" variables refer to women in the household files. All of the "v" variables refer to women in the IR files. The v variables in the IR file also appear as v variables in the BR and KR files.

Report message to a moderator

Fri, 08 June 2018 09:03

Mayank_Ag
Messages: 22
Registered: May 2018
Location: Hyderabad

Member

I did a merge of KR (v001 v002 b16) and PR (hv001, hv002, hvidx) with PR as my base.

I did a little digging so as to why they are not matching. All the variables containing the child and household characteristics
(HW70 with HC70; HW71 with HC71 etc.) are matching. But the variables containing mother's characteristics are not matching. Further it looks like as if they have got shifted by one place. I have no idea how is that happening. After this i checked the merged file with my base files. It looks as if the problem is with the base PR file itself.

I have attached an image that can tell the whole problem. I have selected 3 variables to demonstrate the issue. But the trend is all over.

I used this same file(merged) to generate all the estimates as presented in the final table 10.1. I got them all matching except the ones that involve mother's characteristics (schooling, interview and BMI).

Attachment: issue.JPG
(Size: 88.81KB, Downloaded 539 times)

Report message to a moderator

Wed, 20 June 2018 05:23

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Guillermo Rojas, Senior DHS Data Specialist:

KR files contain one entry for every child born in the last 5 years according to the woman's birth history. For that reason only variables from children will be properly linked when merging KR with PR. If you want the woman's variables you need to merge PR with IR and for men PR with MR. In the last two cases the merging variables for women will be V001, V002 and V003 and for men MV001, MV002 and MV003.

Report message to a moderator

Wed, 20 June 2018 06:36

Mayank_Ag
Messages: 22
Registered: May 2018
Location: Hyderabad

Member

Thanks a lot for the reply. That really helped. But i have 2 more queries.

1. I found 3 duplicates in KR file with same values for all the variables except for the identity variables (V001, V002, V003). Can you please explain this?

2. I have been trying to match the estimates presented in the table 10.0 of the final report. As mentioned earlier i have not been able to do so for Mother's nutritional status. I was given some code by another specialist on the forums. The code did generate the percentages with good accuracy but the total no. of obs. didn't match as presented in the final reports (Image attached) and the conditions mentioned in the report were not taken into account (Images attached). Can you please mention how to correctly generate the figures ?

Estimates

Underweight - 45.9; 51103
Normal- 38.2; 128260
Overweight 27.1; 32318

TIA.

Attachment: BMI 2.JPG
(Size: 15.46KB, Downloaded 552 times)
Attachment: BMI3.JPG
(Size: 11.33KB, Downloaded 479 times)
Attachment: MN.JPG
(Size: 64.36KB, Downloaded 503 times)

[Updated on: Wed, 20 June 2018 06:37]

Report message to a moderator

Previous Topic:	Recode Manual for DHS-VII
Next Topic:	Senegal health insurance variables

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Thu Dec 26 20:13:23 Coordinated Universal Time 2024