The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Nutrition and Anthropometry » Stunting rate with accurate observation number
Stunting rate with accurate observation number Thu, 31 May 2018 11:46
 anikhpg42@gmail.com Messages: 38Registered: December 2017 Location: Bangladesh Member
Hi,
In order to find the stunting rate among children under age 5 using BDHS 14 data, I have used the following Stata command.
I used this command both in KR & PR dataset. But the total observation number is not exactly matching with the BDHS-14 report.

Prevalence of Stunting in 2014 (Children under age 5)

My estimation,
Using KR file, 36.24 % (N = 6965)
Using PR file, 36.09 % (N = 7256)

From BDHS'14 Report
36.1 % (N = 7318)

Here, my question is, how can I obtain the same prevalence rate and same observation for calculating child stunting rate?
Where is my mistakes, that I've made in my Stata command?
I have to get the exact observations (N = 7318) and the exact stunting rate (36.10%).

Stata command for stunting (just for the KR file)

*SVY command
gen strata=v023
gen psu=v021
gen sampwt=v005/1000000
svyset psu [pw=sampwt], strata (strata)

//child stunting calculation
codebook hw70
tab hw70 if hw70>9990,m
tab hw70 if hw70>9990, m nolabel
gen HAZ=hw70
replace HAZ=. if HAZ>=9996
*****
gen stunted=.
replace stunted=0 if HAZ~=.
replace stunted=1 if HAZ<-200

svy: tab stunted

Best regards,
Anik

ASIBUL ISLAM ANIK
Re: Stunting rate with accurate observation number [message #15136 is a reply to message #15090] Thu, 07 June 2018 10:45
 Trevor-DHS Messages: 787Registered: January 2013 Senior Member
We estimate stunting based on the data in the PR file, but restricted to the de facto children. the following code produces the estimate you want:
recode hc70 (min/-201 = 1 "Stunted") (-200/9990 = 0 "Not stunted") (else=.), gen(stunting)
tab stunting [iw=hv005/1000000] if hc70 < 9990 & hv103==1
Re: Stunting rate with accurate observation number [message #15291 is a reply to message #15136] Mon, 25 June 2018 14:25
 hamidine Messages: 1Registered: June 2018 Location: Niger Member
Dear Trevor,
I have used PR file for DHS Niger 2012 . I found for Global acute malnutrition 18,8% against 18.0% reported but N=5701 match with the reported
Weight / age, I found (prevalence = 38% vs 36% reported and observations mismatch N= 5911). height/age (N= 5743 and prevalence =42.9% Vs 44% reported).
I have recalculate Zscore with using syntaxe enclosed. could you check why could I not have the same result as reported.

could you mind to give your syntaxe for SPSS Users "recode hc70 (min/-201 = 1 "Stunted") (-200/9990 = 0 "Not stunted") (else=.), gen(stunting)
tab stunting [iw=hv005/1000000] if hc70 < 9990 & hv103==1"

Hamidine Hassane
Re: Stunting rate with accurate observation number [message #15293 is a reply to message #15291] Mon, 25 June 2018 15:27
 Trevor-DHS Messages: 787Registered: January 2013 Senior Member
If you are using the igrowup syntax you will not be able to match exactly what DHS produces for a number of reasons. See message 892 for a discussion of the differences in the flagging. Principally DHS used to flag and exclude cases that were flagged on any of the three anthropometric Z-scores, and flags cases if either the month or year of birth was not given (hc33 > 1) (and, if I remember correctly, the igrowup routines included these cases for Height-for-age or weight-for-age, which we disagree with).

Below is code in SPSS to calculate stunting. I tested this with Niger 2012 PR file and found 43.9% stunted as in the report:
recode hc70 (lo thru -201 = 1) (-200 thru 9990 = 0 ) (else=sysmis) into stunting.
variable labels stunting "Stunting".
value labels stunting 0 "Not stunted" 1 "Stunted".
formats stunting (f1.0).

compute wgt = hv005/1000000.
weight by wgt.

compute filter_\$=(hc70 < 9990 & hv103=1).
variable labels filter_\$ 'hc70 < 9990 & hv103=1 (FILTER)'.
value labels filter_\$ 0 'Not Selected' 1 'Selected'.
formats filter_\$ (f1.0).
filter by filter_\$.
execute.

frequencies stunting.

Re: Stunting rate with accurate observation number [message #15726 is a reply to message #15293] Thu, 06 September 2018 20:20
 dhsforum Messages: 4Registered: September 2018 Member
Dear Trevor

I'm trying to merge height and weight data (HW datafile) based on the new WHO standards to children's recode for Ghana DHS 2003. However, the identifiers provided in the HW dataset does not seem to match with what are in the children's recode file. Additionally, there are more observations in the HW recode file than there are in the children's recode file. Any help please?

I have been able to do that perfectly for 1998 and 1993 datasets though

Thanks
Re: Stunting rate with accurate observation number [message #15727 is a reply to message #15726] Fri, 07 September 2018 00:00
 Trevor-DHS Messages: 787Registered: January 2013 Senior Member
This HW file is probably designed for merging with the PR file, rather than the KR file. Early survey collected anthropometry as part of the women's questionnaire, and analyzed using the KR file. Later survey switched to collecting the data as part of the household questionnaire (and now as part of the biomarker questionnaire), and the data are analyzed with the PR file. See the .doc file included with the HW file for more information.
Re: Stunting rate with accurate observation number [message #15730 is a reply to message #15727] Fri, 07 September 2018 11:14
 dhsforum Messages: 4Registered: September 2018 Member
Dear Trevor
Many thanks for your response. As you advised, I have been able to merge it with the PR file but following the directions in the .doc file to merge the newly merged file with the children's recode file again seems not to work. Am I missing something?

Thanks once again
Re: Stunting rate with accurate observation number [message #15731 is a reply to message #15730] Fri, 07 September 2018 11:19
 Trevor-DHS Messages: 787Registered: January 2013 Senior Member
We analyze the anthropometry for children in these surveys using the PR file, not the KR file. If you want to match results in the report they are based on the PR file.

If you still want to use the KR file, you need to merge the PR to the KR file. The IDs from the PR file are HV001 HV002 HVIDX, and from the KR file they are V001 V002 B16. Note that they will not all match as any child of the women who does not live in the household will be missing on B16. Similarly, there will be a number of children listed in the PR file who are not children of interviewed women so they will not appear in the KR file at all.

[Updated on: Fri, 07 September 2018 11:19]

Report message to a moderator

Re: Stunting rate with accurate observation number [message #15732 is a reply to message #15731] Fri, 07 September 2018 11:29
 dhsforum Messages: 4Registered: September 2018 Member
Thanks for that.
I am not trying to match the result in the report though.
I'm doing an analysis based on children as the unit of analysis rather.