Recalculating and comparing Wealth Index for Côte d'Ivoire [message #8960] |
Tue, 19 January 2016 13:59 |
baremma2002
Messages: 20 Registered: April 2013
|
Member |
|
|
Dear DHS experts,
I'm using 1994, 1998-99 and 2011-12 Côte d'Ivoire DHS data and I'd like to calculate comparative wealth index (CWI) of Côte d'Ivoire for all those three standard DHS surveys in the trends analysis. I've three questions and I think these have not been asked yet for Côte d'Ivoire DHS data :
First, what are the values of the constant (alpha) and the coefficient (beta) that I can use to calculate the CWI for Côte d'Ivoire?
Second, the wealth index for 1994, 1998-99 and 2011-12 Côte d'Ivoire DHS data have been calculated and stored in seperate files. What are the variables used to compute the wealth index for each DHS and what were the techniques used (Principal Component Analysis, Principal Component Regression) for 1994 and 1998-99 DHS? Where can I find such files?
Third, do I have to get the same mean and the same standard deviation for CIW as found in this DHS report which is available at:
http://dhsprogram.com/publications/publication-MR9-Methodolo gical-Reports.cfm
Any help would be appreciated.
Thank you very much.
|
|
|
|
|
|
|
|
|
|
|
|
Re: Recalculating and comparing Wealth Index for Côte d'Ivoire [message #19878 is a reply to message #19864] |
Mon, 24 August 2020 17:07 |
Bridgette-DHS
Messages: 3167 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
The Stata lines listed below will do this. You should just need to change the paths. I have included some comments. The trickiest part is that you must extract v001 and v002 (or hv001 and hv002), which are numeric, from whhid, which is a string. This involves the substring and destring commands. The program shows how to do this. Let us know if you still have problems. Good luck!
set more off
use "C:\Users\26216\ICF\Analysis - Shared Resources\Data\DHSdata\CIWI3AFL.DTA" , clear
rename wlthindf v191
rename wlthind5 v190
list whhid if _n<=20, table clean
describe whhid
* hhid is a 12-column string that includes v001 and v002
* need to break them out; to do that we need to find which column is which
local li=1
while `li'<=12 {
gen col_`li'=substr(whhid,`li',1)
local li=`li'+1
}
list whhid col_* if _n<=20, table clean
* apparently v001 is columns 8-10 and v002 is columns 11-12
drop col_*
gen v001=substr(whhid,1,10)
gen v002=substr(whhid,11,2)
destring v*, replace
sort v001 v002
* must save this file somewhere
save e:\DHS\DHS_data\scratch\CIWItemp.dta, replace
use "C:\Users\26216\ICF\Analysis - Shared Resources\Data\DHSdata\CIIR3AFL.DTA" , clear
* check the number of women in this file for later comparison
codebook v000
* there are 3040 women
sort v001 v002
* I prefer the "old" version of merge but this is a many:1 merge because there can be more
* than one woman per household
merge v001 v002 using e:\DHS\DHS_data\scratch\CIWItemp.dta
tab _merge
* _merge=2 for 524 households that have no members in the IR file; drop them
* There are 3040 women with _merge=3, same as the number in the IR file
drop if _merge==2
drop _merge
save e:\DHS\DHS_data\scratch\CI3AFL_with_WI.dta, replace
tab v190
* I have checked and this file is ok, but I am surprised that the distribution of
* women across wealth quintiles is so uneven.
|
|
|