| Binary variable of wealth index [message #23249] |
Fri, 06 August 2021 09:57  |
waqas.hameed1@gmail.com
Messages: 7
Registered: July 2021
|
Member |
|
|
Dear Sir/Madam
I'm doing DHS data analysis. According to my objecive, I need to run sub-group analysis for poor and rich. I would appreciate if you could guide me how can I create a binary variable for wealth index?
a) Should I merge poorest/poor/middle in one category and wealthier and wealthiest in other category
b) Is there a PCA score in the dataset. If so, can I take the mean/median and based on that value I split the data into half?
The problem of having middle category separate is that I do not have sufficient cell counts to run 3 sub-group analysis. Also, I tried excluding it but a reviewer put a query that in light of the nature of analysis, middle class cannot be dropped.
Would appreciate guidance in this regard.
|
|
|
|
| Re: Binary variable of wealth index [message #23273 is a reply to message #23249] |
Sun, 15 August 2021 22:26  |
Trevor-DHS
Messages: 788
Registered: January 2013
|
Senior Member |
|
|
Most users who want a binary variable for wealth create a variable that is the 60% poorest and the 40% richest, so your suggestion in a) is what we would usually do.
There is also a PCA score in the dataset (hv271 in the HR, PR files and v191 in IR, KR, BR files) and you could calculate your groups based on this and use this to split the dataset in half. Remember, though, that the wealth quintiles are quintiles of population (household members) based on weighted data, so if you decide to use this approach you will have to ensure that you are weighting the data, and basing the two groups on the population, not the number of households.
For simplicity I would use your first option.
|
|
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
