The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Dataset use in Stata » Community level Poverty
Community level Poverty [message #23174] Tue, 27 July 2021 04:36 Go to next message
saifulhaq is currently offline  saifulhaq
Messages: 7
Registered: October 2019
Member
Dear expoerts,
Please give your feedback on the following coding which i have used for constructing "community level poverty" using dhs data of Pakistan. i need to know if this variable has been constructed correctly.

gen cpvt=0
replace cpvt=1 if v190==1
bysor v001:egen cmpov=mean(cpvt)
xtile compov=cmpov,nq(3)
la def compov 1 low 2 moderate 3 high
la var compov "Community Poverty Level"
la val compov compov
tab compov[iweight=wgt]

Re: Community level Poverty [message #23178 is a reply to message #23174] Tue, 27 July 2021 08:52 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3017
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

It would help if you would say what you intended to do. Otherwise, all we can say is what it appears that you did. You constructed a variable cmpov that has 3 categories, which are terciles based on the proportion of women in the cluster who are in the bottom wealth quintile. The categories are correctly labeled. There are many alternative ways in which you could define a similar measure, but this way of doing it appears to be coded correctly in Stata.
Re: Community level Poverty [message #23181 is a reply to message #23178] Tue, 27 July 2021 13:04 Go to previous messageGo to next message
saifulhaq is currently offline  saifulhaq
Messages: 7
Registered: October 2019
Member
Thank you for your feedback. I am actually trying to generate community level variable using the individual level categorical variable as given in the DHS data set. My next question is regarding the validity of this particular method of constructing community level variable. More precisely, can we take mean of categorical variable while constructing community level variable using a category ( lowest quintile for poverty in this case).
Looking forward

Saif
Pakistan
Re: Community level Poverty [message #23182 is a reply to message #23181] Tue, 27 July 2021 14:32 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3017
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

Last year we produced a report that's related to what you are doing (https://www.dhsprogram.com/pubs/pdf/AS76/AS76.pdf). It uses more than just the bottom quintile. I think your index is valid and would be useful, but there are alternatives. In my experience it's hard to interpret the results of a multi-level model with a categorical variable at the cluster level. Another option would be a continuous cluster-level measure of wealth, such as simply the percentage of women (or households) that are in the bottom quintile or in the bottom two quintiles.

Re: Community level Poverty [message #23184 is a reply to message #23182] Wed, 28 July 2021 04:40 Go to previous messageGo to next message
saifulhaq is currently offline  saifulhaq
Messages: 7
Registered: October 2019
Member
Dear Sir,
Thank you for your response. However i would appreciate if you guide me on how to construct "continuous cluster-level measure of wealth" in stata.

Thanks in advance

Saif
Pakistan
Re: Community level Poverty [message #23187 is a reply to message #23184] Wed, 28 July 2021 08:10 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3017
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

If you just use "cmpov" in the Stata code you sent, rather than "compov", you will have the proportion of women in the cluster who are in wealth quintile 1. You could treat that as a continuous measure.

As a slight modification, you could replace "if v190==1" with "if v190==1 | v190==2". This would give the proportion of women who are in wealth quintiles 1 and 2, and I think will be in a better range.

If you multiply by 100, you will have a percentage rather than a proportion.

Previous Topic: Availability of health facility
Next Topic: Multi-level mixed effect Poisson regression in Stata
Goto Forum:
  


Current Time: Thu Mar 28 19:03:22 Coordinated Universal Time 2024