The DHS Program User Forum: India » Inconsistencies between recode number of deaths and birth history

Home » Countries » India » Inconsistencies between recode number of deaths and birth history

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

Inconsistencies between recode number of deaths and birth history [message #25822]

Wed, 14 December 2022 23:23

cportner
Messages: 20
Registered: September 2014
Location: Seattle

Member

Hello,

It looks like there are 142 women in NFHS-4 for whom v206+v207 (sons and daughters who have died) does not match the count of deaths across b7_*. Is there a reason for this mismatch, and which one should I trust more?

All the best,

Claus Pörtner

[Updated on: Thu, 15 December 2022 10:06]

Report message to a moderator

Re: Inconsistencies between recode number of deaths and birth history [message #25826 is a reply to message #25822]

Thu, 15 December 2022 11:38

Bridgette-DHS
Messages: 3038
Registered: February 2013

Senior Member

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

I used the following Stata code to check whether there are any discrepancies between the v20* variables and the b4* and b5* variables in terms of the numbers of sons and daughters who are alive or dead. I get perfect agreement. I used b5 rather than b7, but b5 and b7 should agree, in that b7 is <. If the child died (b5=0). However, b7=. could mean either that the child survived (b5=0) or there was no birth for the specified value of bidx (b5=.) Can you check again?

use v20* b4* b5* using "...IAIR52FL.dta", clear
rename *_0* *_*

gen v20sons_alive=v202+v204
gen v20dtrs_alive=v203+v205

gen sons_alive=0
gen dtrs_alive=0
gen sons_died=0
gen dtrs_died=0

forvalues li=1/16 {

scalar si=`li'
scalar list si

replace sons_alive=sons_alive+1 if b4_`li'==1 & b5_`li'==1
replace dtrs_alive=dtrs_alive+1 if b4_`li'==2 & b5_`li'==1
replace sons_died=sons_died+1   if b4_`li'==1 & b5_`li'==0
replace dtrs_died=dtrs_died+1   if b4_`li'==2 & b5_`li'==0 
}
  
tab v20sons_alive sons_alive
tab v20dtrs_alive dtrs_alive
tab v206 sons_died
tab v207 dtrs_died

gen d1=v20sons_alive-sons_alive
gen d2=v20dtrs_alive-dtrs_alive
gen d3=v206-sons_died
gen d4=v207-dtrs_died

tab1 d1 d2 d3 d4

Report message to a moderator

Re: Inconsistencies between recode number of deaths and birth history [message #25827 is a reply to message #25826]

Thu, 15 December 2022 13:22

cportner
Messages: 20
Registered: September 2014
Location: Seattle

Member

The data file that I used was IAIR74FL.DTA. The 52 one is NFHS-3.

If I run Tom's code on IAIR74FL.DTA I get:

. tab1 d1 d2 d3 d4

-> tabulation of d1

d1 | Freq. Percent Cum.
------------+-----------------------------------
-3 | 1 0.00 0.00
-2 | 4 0.00 0.00
-1 | 7 0.00 0.00
0 | 699,600 99.99 99.99
1 | 66 0.01 100.00
2 | 5 0.00 100.00
3 | 2 0.00 100.00
4 | 1 0.00 100.00
------------+-----------------------------------
Total | 699,686 100.00

-> tabulation of d2

d2 | Freq. Percent Cum.
------------+-----------------------------------
-4 | 1 0.00 0.00
-2 | 1 0.00 0.00
-1 | 10 0.00 0.00
0 | 699,601 99.99 99.99
1 | 67 0.01 100.00
2 | 6 0.00 100.00
------------+-----------------------------------
Total | 699,686 100.00

-> tabulation of d3

d3 | Freq. Percent Cum.
------------+-----------------------------------
-4 | 1 0.00 0.00
-3 | 2 0.00 0.00
-2 | 5 0.00 0.00
-1 | 66 0.01 0.01
0 | 699,599 99.99 100.00
1 | 8 0.00 100.00
2 | 4 0.00 100.00
3 | 1 0.00 100.00
------------+-----------------------------------
Total | 699,686 100.00

-> tabulation of d4

d4 | Freq. Percent Cum.
------------+-----------------------------------
-2 | 6 0.00 0.00
-1 | 66 0.01 0.01
0 | 699,603 99.99 100.00
1 | 9 0.00 100.00
2 | 1 0.00 100.00
4 | 1 0.00 100.00
------------+-----------------------------------
Total | 699,686 100.00

Report message to a moderator

Re: Inconsistencies between recode number of deaths and birth history [message #25834 is a reply to message #25827]

Fri, 16 December 2022 08:19

Bridgette-DHS
Messages: 3038
Registered: February 2013

Senior Member

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

Apologies for checking the wrong survey! I have revised that Stata program to list the inconsistent cases and loop through the IR files from the five India surveys. I will paste it below. It's pretty general and could be used on any DHS surveys. I use a prefix "v" for subtotals coming from v202-v207 and a prefix "b" for subtotals coming from the b4_* and b5_* variables. The number of children "living with the mother" could be checked against the number with b9_* equal to 0.

There are no inconsistent cases in the NFHS-2, -3, and -5 surveys. There is one inconsistent case in NFHS-1, and it is easily found, because it is the only case with v201=. . In NFHS-4, I find 128 cases with 2 inconsistencies and 21 cases with 4 inconsistencies. I suspect that the inconsistencies occur in pairs, for example with a difference of +1 for sons alive and -1 for sons dead. I did not take the time to check that.

Such inconsistencies should not occur. If you need to give priority to one over the other, the b variables should be more reliable. In the interview, there is probing during the collection of the birth histories. The totals and subtotals in v201-v207 should be forced to agree with the b variables. Thank you for identifying this issue with the NFHS-4.

* this program checks v202-v207 against the birth histories

**************************************************************************************

program define check_subtotals

keep v024 v001 v002 v003 v20* b4_* b5_*
rename *_0* *_*

gen vsons_alive=v202+v204
gen vdtrs_alive=v203+v205
gen vsons_died=v206
gen vdtrs_died=v207

gen bsons_alive=0
gen bdtrs_alive=0
gen bsons_died=0
gen bdtrs_died=0

* Compare the maximum observed value of v201 with the maximum allowed value of the birth history index.
summarize v201

scalar sbstop=0
forvalues li=15/25 {
  if sbstop==0 {
  capture confirm numeric variable b4_`li', exact 
    if _rc>0 {
    scalar sbstop=1
    scalar sbmax=`li'-1
    } 
  }
}

scalar list sbmax
local lbmax=sbmax

quietly forvalues li=1/`lbmax' {
scalar si=`li'
replace bsons_alive=bsons_alive+1 if b4_`li'==1 & b5_`li'==1
replace bdtrs_alive=bdtrs_alive+1 if b4_`li'==2 & b5_`li'==1
replace bsons_died=bsons_died+1   if b4_`li'==1 & b5_`li'==0
replace bdtrs_died=bdtrs_died+1   if b4_`li'==2 & b5_`li'==0 
}
  
*tab vsons_alive bsons_alive
*tab vdtrs_alive bdtrs_alive
*tab vsons_died  bsons_died
*tab vdtrs_died  bdtrs_died

gen d1=vsons_alive-bsons_alive
gen d2=vdtrs_alive-bdtrs_alive
gen d3=vsons_died -bsons_died
gen d4=vdtrs_died -bdtrs_died

tab1 d1 d2 d3 d4

gen D=0
label variable D "case has D inconsistencies"
replace D=D+1 if d1~=0
replace D=D+1 if d2~=0
replace D=D+1 if d3~=0
replace D=D+1 if d4~=0

sort v024 v001 v002 v003

end

**************************************************************************************
**************************************************************************************
**************************************************************************************
**************************************************************************************
**************************************************************************************
* Execution begins here

local lpvs "23 42 52 74 7D"

foreach lpv of local lpvs {
scalar spv="`lpv'"
use "...\IAIR`lpv'FL.dta", clear
quietly check_subtotals
scalar list spv
tab D if D>0
list v024 v001 v002 v003 v201 *alive *died if D>0, table clean abbrev(11)
}

Report message to a moderator

Previous Topic:	Telangana State discrepancy in numbers of women with a live birth
Next Topic:	TV exposure decline and v159 coding error?

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed Apr 24 05:15:53 Coordinated Universal Time 2024