The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Countries » India » Mismatch between National report and NFHS4 raw data analysis
Mismatch between National report and NFHS4 raw data analysis [message #16227] Wed, 28 November 2018 21:40 Go to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member

I am working with NFHS4 raw data. Many times my analysis is not matching with National Report. Tried to sort out this issue, discussed with various people, but couldn't succeed. For example the number given in the national report of children born in the two years preceding the survey who ever breasted itself is greater than the total number of children which includes both who breastfed and not. How come that number will be greater than this total number. This is only one example, there are many other issues as well. I have used weights and all, followed every statistical procedure explained by USAID people.
Re: Mismatch between National report and NFHS4 raw data analysis [message #16237 is a reply to message #16227] Fri, 30 November 2018 11:48 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
In Table 10.4 in the national report, the second column shows the number of children born in the last two years (97,989). that's the denominator of the estimate, not the numerator. As the table footnote indicates, that number includes births in the last two years whether the children are living or dead at the time of the interview.
Re: Mismatch between National report and NFHS4 raw data analysis [message #16239 is a reply to message #16237] Fri, 30 November 2018 21:37 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear sir
I am very gratitude andThanks for your response

If you are saying this number is (97,989) is the total number and the denominator, again it is not matching with our analysis, I have attached the stata code for the same, let me know whether I have made any mistakes.

These are my stata command used for


***********************************************************
clear all
use "/Users/muneerkalliyil/Desktop/UNICEF/NFHS4/IAKR74FL.DTA"

***APPLYING WEIGHTS************************

*Survey Weights

des v005
gen survweight = v005/1000000
gen stratum = v023
svyset v021 [iweight=survweight], strata(stratum)

****AFTER APPLYING WEIGHTS, JUST WE ARE LOOKING HOW MANY CHILDRENS ARE THERE BELOW 24 MONTHS,

**total number of children under 24 months
svy: tab hw1 if hw1 < 24
*******************************************************

The result says no of obs as 96782;

. svy: tab hw1 if hw1 < 24
(running tabulate on estimation sample)

Number of strata = 2513 Number of obs = 96782
Number of PSUs = 26472 Population size = 93053.216
Design df = 23959

NFHS4 National report gives the total number of children below 24 months as (97989), but our analysis shows as (96782 )


Please let me know, waiting for your reply

Regards
Muneer
Re: Mismatch between National report and NFHS4 raw data analysis [message #16246 is a reply to message #16239] Sun, 02 December 2018 17:42 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
It looks like you're only including living children. As indicated earlier, the denominator includes both living and dead children who were born in the last two years.
Re: Mismatch between National report and NFHS4 raw data analysis [message #16247 is a reply to message #16239] Sun, 02 December 2018 22:53 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear Sir

Thanks for your reply

I will be very grateful if you could tell me how to include both living and dead children. As per my understanding, the variable hw1 (age in months) includes both, I have gone through all other variables in Children file, but couldn't get a variable differentiating between dead and alive children. Please let me know if you know how to include.


Looking forward to hearing from you
Thanks in advance
Re: Mismatch between National report and NFHS4 raw data analysis [message #16251 is a reply to message #16247] Mon, 03 December 2018 16:25 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
You should use the kids' file (IAKR74). To obtain the number of children born in the two years before the survey (including both living and dead children), use V008 - B3 < 24.
Re: Mismatch between National report and NFHS4 raw data analysis [message #16254 is a reply to message #16251] Mon, 03 December 2018 22:36 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear Sir

very thankful to your replies

I tried using V008 - B3 < 24 for age, then I got the number of children below 24 as 101955. However, the national report says 95% of the total below 24 has been breastfed, if we take the 95% of this number (96857), again it is not matching with national report number (97989). Could you please let me know if I am making any mistakes

Looking forward to hearing from you
Regards
Re: Mismatch between National report and NFHS4 raw data analysis [message #16257 is a reply to message #16254] Tue, 04 December 2018 13:53 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
You're using the unweighted data. You'll have to weight the data to match what's in the report. Once you weight the data, you should be able to match the total number of children born in the two years preceding the survey (97,989). As mentioned earlier, the number of children in the second column of Table 10.4 is the denominator (the total number of children born in the two years preceding the survey). It is not the numerator (the number of children born in the two years preceding the survey who were ever breastfed).
Re: Mismatch between National report and NFHS4 raw data analysis [message #16259 is a reply to message #16257] Tue, 04 December 2018 23:28 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear Sir,

I have already applied survweight, tried both pweight and iweight. The number which I told earlier is coming after applying the weight. Secondly, If the number in the second column is denominator, then our estimated denominator is definitely coming larger than the given denominator. As I told before, in our estimation, the total number of children below 24 months is coming around 101955. However, the given number in the report is only around 97989.

Please let me know if I am making any mistakes.

Looking forward to hearing from you

Thanks
Re: Mismatch between National report and NFHS4 raw data analysis [message #16262 is a reply to message #16259] Wed, 05 December 2018 11:30 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
The Stata code below matches the denominator in the NFHS-4 report:

gen age = v008 b3
gen xweight = v005/1000000
tabulate age if age < 24 [iw=xweight]

age | Freq. Percent Cum.
------------+-----------------------------------
0 | 1,629.0776 1.66 1.66
1 | 3,670.6754 3.75 5.41
2 | 3,935.1411 4.02 9.42
3 | 4,126.8893 4.21 13.64
4 | 4,405.3671 4.50 18.13
5 | 4,464.9622 4.56 22.69
6 | 4,501.2567 4.59 27.28
7 | 4,415.1894 4.51 31.79
8 | 4,631.844 4.73 36.51
9 | 4,300.8026 4.39 40.90
10 | 4,114.4416 4.20 45.10
11 | 3,905.3321 3.99 49.09
12 | 4,205.2019 4.29 53.38
13 | 4,226.0275 4.31 57.69
14 | 4,074.5033 4.16 61.85
15 | 4,041.3627 4.12 65.97
16 | 4,252.8293 4.34 70.31
17 | 4,150.9773 4.24 74.55
18 | 4,489.1452 4.58 79.13
19 | 4,189.1484 4.28 83.41
20 | 4,192.2816 4.28 87.69
21 | 4,265.4102 4.35 92.04
22 | 4,003.8333 4.09 96.12
23 | 3,797.2187 3.88 100.00
------------+-----------------------------------
Total | 97,988.918 100.00
Re: Mismatch between National report and NFHS4 raw data analysis [message #16264 is a reply to message #16262] Wed, 05 December 2018 20:57 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear Fredarnold Sir,

Thank you so much, the code is working.

I was looking for this solution for many days.

Will keep in touch with

Regards
Muneer
Re: Mismatch between National report and NFHS4 raw data analysis [message #16343 is a reply to message #16264] Tue, 18 December 2018 00:46 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear Fredarnold Sir,

First of all thank you so much for earlier clarification regarding mismatch in NFHS national report, Your solution was working perfectly.

However, now I have come across state-level reports. For example, I have taken the Jharkhand report. Table 61 of initial breastfeeding has been attached for your reference.


The table says that Number of last born children below 24 months is given as 4723.

However my estimation is not matching with state report, Stata coding is given below;

************************************************************ *******
****Jharkhand****
clear all
use "/Users/muneerkalliyil/Desktop/UNICEF/NFHS4/IAKR74FL.DTA"
gen survweight = v005/1000000
gen stratum = v023
svyset v021 [iw=survweight], strata(stratum)

gen age = v008 - b3

numlabel, add
****keeping only Jharkhand****
tab v024

keep if v024 == 15

****giving weight****
tabulate age if age < 24 & midx==1 [iw=survweight]

****without weight*****
tab age if age < 24 & midx==1
************************************************************ ********
I have estimated number of last born children below 24 months, giving weight and without giving as well.
Result says the number of children with weight = 2864, without weight = 4705.
Therefore, in both cases, the result is not matching with state report. It says the number is 4723. And the gap becomes large when we use weight.

My queries are;
Is it correct to use weight in the state level, if yes, is it different from national weight?
Why the gap is large with and without weight
Could you please sort out my problem as soon as possible

Looking forward to hearing from you

Thanks in advance

Re: Mismatch between National report and NFHS4 raw data analysis [message #16346 is a reply to message #16343] Tue, 18 December 2018 11:44 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
For tables for individual states, you should use the state weight variable (SV005) instead of V005. SV005 is the weight used in all the state report for NFHS-4.
Re: Mismatch between National report and NFHS4 raw data analysis [message #16348 is a reply to message #16346] Tue, 18 December 2018 20:21 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Thank You so much, sir
It is working perfectly
Re: Mismatch between National report and NFHS4 raw data analysis [message #16350 is a reply to message #16346] Wed, 19 December 2018 11:09 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear Sir,

Sorry for spamming your inbox again and again.
This time again with earlier same table, Table 61 Initial breastfeeding (attached below).
I am looking to the second column, Percentage who started breastfeeding within one hour of birth. The table says that 29.9% of Urban and 33.8% of Rural children have been put into breast within one hour of birth.

myStata code;
recode m34 (0/100 = 1) (101/236 = 0), gen(bf1hour)
svy: tab bf1hour if age < 24 & midx==1 & v102 == 1
svy: tab bf1hour if age < 24 & midx==1 & v102 == 2

and our results say

. svy: tab bf1hour if age < 24 & midx==1 & v102 == 1
(running tabulate on estimation sample)

Number of strata = 24 Number of obs = 828
Number of PSUs = 282 Population size = 858.55407
Design df = 258


RECODE of
m34 (when
child put
to
breast) proportions

0 .6929
1 .3071

Total 1

Key: proportions = cell proportions

. svy: tab bf1hour if age < 24 & midx==1 & v102 == 2
(running tabulate on estimation sample)

Number of strata = 72 Number of obs = 3735
Number of PSUs = 917 Population size = 3725.5765
Design df = 845


RECODE of
m34 (when
child put
to
breast) proportions

0 .6518
1 .3482

I agree it is a small difference, 30.71 in place of 29.9% and 34.82 in place of 33.8%, still, I am asking you because I got an exact matching solution from you for earlier problems.

Looking forward to hearing from you
Thanks in advance



Re: Mismatch between National report and NFHS4 raw data analysis [message #16363 is a reply to message #16350] Wed, 26 December 2018 12:15 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
It looks like you are trying to match the results for Jharkhand. In that case, you need to change the weight from v005 to sv005, which is necessary every time you are analyzing an individual state. As you can see below, the urban percentage for Jharkhand (29.9 percent) comes out exactly the same as the estimate in the Jharkhand state report.

. svy: tab bf1hour if age < 24 & midx == 1 & v102 == 1 & v024 == 15
(running tabulate on estimation sample)

Number of strata = 24 Number of obs = 852
Number of PSUs = 286 Population size = 881.91211
Design df = 262

-----------------------
RECODE of |
m34 (when |
child put |
to |
breast) | proportions
----------+------------
0 | .7011
1 | .2989
|
Total | 1
-----------------------
Key: proportions = cell proportions
Re: Mismatch between National report and NFHS4 raw data analysis [message #16364 is a reply to message #16363] Thu, 27 December 2018 09:07 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Thanks so much
The problem was in recoding, now it is matching perfectly


Thank
keep in touch with
Re: Mismatch between National report and NFHS4 raw data analysis [message #16686 is a reply to message #16363] Sun, 17 February 2019 10:34 Go to previous messageGo to next message
Pooja Arora is currently offline  Pooja Arora
Messages: 3
Registered: February 2019
Member
Dear fredarnold,

When I use svyset command with state level weights for doing state level analysis for India, I use the weight variable sv005. In that, does my strata remain the same? That is v022?

Thank you,
Pooja




fredarnold wrote on Wed, 26 December 2018 12:15
It looks like you are trying to match the results for Jharkhand. In that case, you need to change the weight from v005 to sv005, which is necessary every time you are analyzing an individual state. As you can see below, the urban percentage for Jharkhand (29.9 percent) comes out exactly the same as the estimate in the Jharkhand state report.

. svy: tab bf1hour if age < 24 & midx == 1 & v102 == 1 & v024 == 15
(running tabulate on estimation sample)

Number of strata = 24 Number of obs = 852
Number of PSUs = 286 Population size = 881.91211
Design df = 262

-----------------------
RECODE of |
m34 (when |
child put |
to |
breast) | proportions
----------+------------
0 | .7011
1 | .2989
|
Total | 1
-----------------------
Key: proportions = cell proportions

[Updated on: Sun, 17 February 2019 10:37]

Report message to a moderator

Re: Mismatch between National report and NFHS4 raw data analysis [message #16692 is a reply to message #16686] Tue, 19 February 2019 08:58 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
yes
Re: Mismatch between National report and NFHS4 raw data analysis [message #16694 is a reply to message #16692] Tue, 19 February 2019 11:44 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3199
Registered: February 2013
Senior Member

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

Yes, v022 is the stratum id at the state level as well as at the national level. The strata that are outside the state will be ignored.
Re: Mismatch between National report and NFHS4 raw data analysis [message #16707 is a reply to message #16694] Fri, 22 February 2019 08:27 Go to previous messageGo to next message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Dear fredarnold,

As there are two different weights for the country and state, is there any third weight for district analysis in case of India. If not, what should we do, should we apply state weight or do without any weight.

Looking forward to hearing from you

Thanks
Re: Mismatch between National report and NFHS4 raw data analysis [message #16708 is a reply to message #16707] Fri, 22 February 2019 09:00 Go to previous messageGo to next message
fredarnold is currently offline  fredarnold
Messages: 45
Registered: March 2014
Member
For district level, or any sub-domain, either national weights or state weights can be used. Both will result in the same indicator values but not the same weighted N's. The national weights can be used for any analysis. The state weights can be used for data analysis on the state or lower levels.

Re: Mismatch between National report and NFHS4 raw data analysis [message #16709 is a reply to message #16708] Fri, 22 February 2019 09:57 Go to previous message
Muneer Kalliyil is currently offline  Muneer Kalliyil
Messages: 13
Registered: November 2018
Member
Thanks
Previous Topic: HIV 2015 dataset and weight
Next Topic: Multilevel modeling in DHS-Sri Lanka
Goto Forum:
  


Current Time: Mon Nov 25 07:05:38 Coordinated Universal Time 2024