The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » General » Unable to reproduce results of age at first marriage table
Unable to reproduce results of age at first marriage table [message #9391] Wed, 23 March 2016 15:03 Go to next message
LaissezPasser is currently offline  LaissezPasser
Messages: 1
Registered: March 2016
Location: Maldives
Member
I'm unable to replicate the results of the 2009 Maldives DHS for age at first marriage (as appear in Table 6.2 (pg.67), below), and I can't figure out why. Specifically, I'm concerned with the cohort of 20-24 year-old women, and the proportion married before the age of 15 and before 18 -- one of the new, finally adopted SDG indicators (5.3.1). Here's the table as it appears in the final report:
http://imgur.com/3CGSirw.jpg
According to Table 6.2, the proportion of women 20-24 years old who married before 18yo is 3.9 percent, based on a sample of 2,161. However, the figures I'm getting are significantly higher when I run my own crosstabs (using MVIR51FL.sav): 6.7 percent, with a SE of 0.8 percent. (And when I run the same crosstab separately on the BR, KR, and CR recode files, the results are even higher.)
I have already made sure to account for the two-stage cluster design by carefully applying the complex stratified weighting procedure outlined in this video. Here's the output (also attached):
http://imgur.com/bjcGe3m.jpg
Note that the n size of the 20-24 cohort is 1,267; this isn't too far from the unadjusted figure (1,381), but it's a long way to the 2,161 women listed in Table 6.2 as comprising the 20-24 cohort. (NOTE: If I were to combine the IR, BR, KR, and CR files, I'd get 2,555 women (unadjusted) in the 20-24 year old cohort. Is this what I should be doing in order to reproduce the published results?)
Here's the SPSS syntax I used to generate this output (also attached):
DATASET ACTIVATE MVIR51FL. 
*compute weight variable*. 
compute WGT = (V005/1000000). 
EXECUTE. 
*create binary categorical for age at first marriage <18yo*. 
RECODE v511 (0 thru 17=1) (ELSE=0) INTO childbride. 
EXECUTE. 
 
* Analysis Preparation Wizard.
CSPLAN ANALYSIS
  /PLAN FILE='/Users/UNFPA/r/09MDHS.csaplan'
  /PLANVARS ANALYSISWEIGHT=WGT       
  /SRSESTIMATOR TYPE=WR
  /PRINT PLAN
  /DESIGN STRATA=V023 CLUSTER=V021 
  /ESTIMATOR TYPE=WR.

*Crosstab 5-year age cohorts with variable for <18yo first marriage*.
* Complex Samples Crosstabs.
CSTABULATE
  /PLAN FILE='/Users/UNFPA/r/09MDHS.csaplan'
  /TABLES VARIABLES=V013 BY childbride
  /CELLS POPSIZE ROWPCT
  /STATISTICS SE CIN(95)
  /MISSING SCOPE=TABLE CLASSMISSING=EXCLUDE.

*and just for kicks, let's run it unweighted".
DATASET ACTIVATE MVIR51FL.
CROSSTABS
  /TABLES=v013 BY childbride
 /FORMAT=AVALUE TABLES
  /CELLS=COUNT row
  /COUNT ROUND CELL.


Thanks in advance for your help.


Delivering a world where every pregnancy is wanted, every childbirth is safe and every young person's potential is fulfilled

[Updated on: Wed, 23 March 2016 15:24]

Report message to a moderator

Re: Unable to reproduce results of age at first marriage table [message #9401 is a reply to message #9391] Thu, 24 March 2016 10:24 Go to previous message
Trevor-DHS is currently offline  Trevor-DHS
Messages: 787
Registered: January 2013
Senior Member
The Maldives 2009 survey is an ever-married women survey (that is only ever-married women were interviewed). To get estimates for all women such as the indicator below you need to apply all women factors to the denominators to inflate them to all women. This is described in other posts on the forum - see the post here. Once you inflate the denominator to all women you will get 84.476 / 2161 = 3.9%.
Previous Topic: How to read V101/V024: Type of Region
Next Topic: Understanding the variables
Goto Forum:
  


Current Time: Thu Mar 28 05:17:48 Coordinated Universal Time 2024