Home » Countries » India » Estimation of percentage of 20-24 year old women who married before the age of 18
| Estimation of percentage of 20-24 year old women who married before the age of 18 [message #25516] |
Tue, 01 November 2022 12:30  |
cpmavelikara18@gmail.com
Messages: 3
Registered: November 2022
|
Member |
|
|
Hai All,
I have been trying to replicate the percentage of 20 to 24-year-old women who married before the age of 18, which is given in NFHS-5 factsheet for India and individual states. I used the IR (Individual Recode) file and used v511 (age at co-habitation) to replicate the results.
But I was not able to correctly match the results at national level. (Note: I weighted by iw=v005/1 million). The factsheet showed the figure as 23.3, while my estimate was 22.2.
Interestingly, I was able to replicate the state-level estimates for the same for a few states using iw=sweight/1 million.
I am adding the STATA codes I used to estimate these results.
************************************************************ ***********************************
use "C:\Users\cpmav\Desktop\NFHS 5\Individual Recoded (Women 15-49 Years)\IAIR7AFL.DTA", clear
*Dividing women's individual weight by 1 million (national weight)
gen wt=v005/1000000
*Dividing women's individual weight by 1 million (state weight)
gen swt=sweight/1000000
*Marriage Before 18 Years
recode v511 (.=0) (0/17 = 1 "yes") (18/49 = 0 "no"), gen (ms_afm_18)
replace ms_afm_18 = . if v012<18
label var ms_afm_18 "First marriage by age 18"
*Tabulations
tab ms_afm_18 if v013==2 [iw=wt] /// All India estimate
tab v024 ms_afm_18 if v013==2 [iw=swt], nofreq row /// State-level estimates.
For states like Bihar, Gujarat, Rajasthan and Meghalaya, I have experienced a deviation above 2 percentage points compared to NFHS-5 estimates.
I request scholars and DHS staff to kindly have a look at this.
[Updated on: Tue, 01 November 2022 12:32] Report message to a moderator |
|
|
|
| Re: Estimation of percentage of 20-24 year old women who married before the age of 18 [message #25531 is a reply to message #25516] |
Fri, 04 November 2022 08:55  |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS staff member, Tom Pullum:
Here is a Stata program that does what you want to do. It includes a translation of the CSPro instructions for coding age at marriage in this survey. I have included some lines (between "/*" and "*/") that give slightly different results. There were 2,487 women with a year of marriage but a missing month of marriage. Those lines assign month=6, which is often done when a month is missing, but the CSPro code treated those cases as NA. The Stata program does not give the rows for 20-49 and 25-49 but they could be added.
* Replication of Table 6.2, Age at First Marriage, in the India 2019-21 DHS (NFHS-5)
use "...IAIR7DFL.dta" , clear
gen never_married=0
replace never_married=1 if v501==0
/*
* Procedure that allocates cases with month missing to month 6 and uses the earlier of
* the reported dates at first marriage and first cohabitation (if both are given)
gen cmc_first_mar=s308c
replace cmc_first_mar=12*(s308y-1900)+6 if s308y<9998 & s308m==98
replace cmc_first_mar=. if s308y==9998
gen cmc_first_cohab=12*(v508-1900)+v507
gen afc=int((cmc_first_cohab-v011)/12)
gen afm=int((cmc_first_mar-v011)/12)
replace afm=min(afc,afm)
*/
* Procedure used in the CSPro construction of the table
gen v511x=.
replace v511x=s309 if s309>=0 & s309<=96
replace v511x=int((s308c-v011)/12) if s308c>=500 & s308c<=1500
replace afm=v511x
replace afm=99 if v501==0
* afm should be over-ridden as NA if v501==0
local lcutoffs 15 18 20 21 25
foreach lc of local lcutoffs {
gen marr_by_`lc'=0
replace marr_by_`lc'=1 if afm<`lc'
}
tab v013 marr_by_15 [iweight=v005/1000000], row
tab v013 marr_by_18 if v013>1 [iweight=v005/1000000], row
tab v013 marr_by_20 if v013>1 [iweight=v005/1000000], row
tab v013 marr_by_21 if v013>2 [iweight=v005/1000000], row
tab v013 marr_by_25 if v013>2 [iweight=v005/1000000], row
tab v013 never_married [iweight=v005/1000000], row
|
|
|
|
| Re: Estimation of percentage of 20-24 year old women who married before the age of 18 [message #25534 is a reply to message #25516] |
Sat, 05 November 2022 01:17 |
cpmavelikara18@gmail.com
Messages: 3
Registered: November 2022
|
Member |
|
|
How can a common person construct this table using information from DHS Guide to Statistics-7, without knowing this information about the coding for missing months? Where is this information in DHS Guide to Statistics-7?
The national and state level factsheets of India (NFHS-5) and Indian states contain 131 estimates. If this is the case with age at marriage, I suspect the case with all other estimates may be the same. Why should a person dedicate hours replicating these tables, when such crucial information is withheld?
So, could the DHS publish an accompanying do file/codes to replicate all these figures?
[Updated on: Sat, 05 November 2022 01:29] Report message to a moderator |
|
|
|
|
|
| Re: Estimation of percentage of 20-24 year old women who married before the age of 18 [message #25547 is a reply to message #25540] |
Tue, 08 November 2022 00:59 |
cpmavelikara18@gmail.com
Messages: 3
Registered: November 2022
|
Member |
|
|
Dear Bridgette/Tom Pullum,
First of all, my whole-hearted gratitude for your timely responses.
I understand the automated process of CSPro, which creates ready-made tables. However, most of the users of NFHS-5 (India DHS) who I know, use STATA. Therefore, we request the concerned team at DHS to publish just the STATA do files/codes for replicating the factsheet figures (excluding the putdocx command etc. to create tables). I use the word 'we', because I am writing this, on behalf of many of my friends in various departments of anthropology, sociology and other humanities stream at University of Hyderabad, who have faced difficulties, replicating these figures.
The factsheet contains only 131 indicators. I only require the codes only for these. The data for NFHS 5 and DHS Guide to Statistics are available. But we get stuck while preparing tables. In the previous messages, regarding the variable concerning child marriage, you will realize that it is because of certain adjustments in cmc (month, year) variables. This information is not available in DHS Guide to Statistics, which are just general instructions to generate the results.
My goal is to replicate the exact factsheet with extreme accuracy, taking into account all such minute details. I acknowledge that the revisions to data etc. may create deviations.
I insist on these do files for India and states in India, because in spite of being the largest DHS in the world in terms of sample size, it not widely used. One of the major reasons for this may be such data replication issues, which makes NFHS a black box for many scholars, requiring copious amount of time, to replicate even one figure.
The nodal agency in India, coordinating the survey, IIPS Mumbai, conducts only occasional data workshops for NFHS. There are also no videos or helpful material, uploaded by IIPS, to replicate these figures. Some researchers, at IIPS, working in certain domains, possess the knowledge to replicate such figures. But they are inaccessible to students from other universities, due to their busy schedules.
So, in the interest of widening the use of NFHS data, I request the DHS team to disseminate the STATA do files, containing only the replication codes - the bare minimum (not tabulation codes) for all 131 indicators in the factsheets for India and Indian states.
[Updated on: Tue, 08 November 2022 01:42] Report message to a moderator |
|
|
|
|
|
|
|
|
|
| Re: Estimation of percentage of 20-24 year old women who married before the age of 18 [message #26308 is a reply to message #26290] |
Mon, 06 March 2023 11:13 |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS staff member, Tom Pullum:
Below I will paste the full Stata code that I wrote to calculate table 6.2 in the NFHS-5 report. It borrows from the original CSPro code for this table. There are some small discrepancies from the published table (mostly only one-tenth of a percentage point) that I cannot account for.
In the procedure, afm (for age at first marriage) is set to 99 for women who are never-married. In your code you do this by setting v511x to 99. The point is that you cannot have "." or NA for afm for any women, because then you would be omitting unmarried women from the calculation. The never-married women have to be given a numerical value for afm that is greater than their current age. This is the only way you can calculate median age at marriage or the % married by an age that is less than current age. I don't know what you mean by "they will anyway get a value of 0 when we run the loop". Binary variables (such as "marr_by_18") are constructed, based on afm, but afm itself is NOT set to 0.
use "....IAIR7DFL.dta" , clear
gen never_married=0
replace never_married=1 if v501==0
* Procedure that allocates cases with month missing to month 6 and uses the earlier of
* the reported dates at first marriage and first cohabitation (if both are given)
gen cmc_first_mar=s308c
replace cmc_first_mar=12*(s308y-1900)+6 if s308y<9998 & s308m==98
replace cmc_first_mar=. if s308y==9998
gen cmc_first_cohab=12*(v508-1900)+v507
gen afc=int((cmc_first_cohab-v011)/12)
gen afm=int((cmc_first_mar-v011)/12)
replace afm=min(afc,afm)
* Procedure used in the CSPro construction of the table
gen v511x=.
replace v511x=s309 if s309>=0 & s309<=96
replace v511x=int((s308c-v011)/12) if s308c>=500 & s308c<=1500
replace afm=v511x
replace afm=99 if v501==0
* afm should be over-ridden as NA if v501==0
local lcutoffs 15 18 20 21 25
foreach lc of local lcutoffs {
gen marr_by_`lc'=0
replace marr_by_`lc'=1 if afm<`lc'
}
tab v013 marr_by_15 [iweight=v005/1000000], row
tab v013 marr_by_18 if v013>1 [iweight=v005/1000000], row
tab v013 marr_by_20 if v013>1 [iweight=v005/1000000], row
tab v013 marr_by_21 if v013>2 [iweight=v005/1000000], row
tab v013 marr_by_25 if v013>2 [iweight=v005/1000000], row
tab v013 never_married [iweight=v005/1000000], row
|
|
|
|
|
|
|
|
|
|
Goto Forum:
Current Time: Mon Dec 2 00:26:29 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
