| Re: Decile Share using Wealth Index Factor by Urban and Rural (Stata) [message #2529 is a reply to message #2528] |
Tue, 08 July 2014 00:17   |
Reduced-For(u)m
Messages: 292
Registered: March 2013
|
Senior Member |
|
|
Well... my bad - a few alternative possibilities:
1 - Use xtile itself (which supports survey weights), doing rural/urban separately, and then generate a new variable that has the decile for both. So:
xtile dec_rural = hv271 [pweight = weightvariable] if rural==1, nq(10)
xtile dec_urban = hv271 [pweight = weightvariable] if rural==0, nq(10)
gen decile = dec_rural
replace decile = dec_urban if rural==0
2 - You can use "qreg" - quantile regression - to find the cut-off points, and assign the deciles yourself with a series of "if wealthindx > cut_a & wealthindex < cut_b" type commands. I am pretty sure that the qreg command supports weights, but I don't think it supports "svy" (which you wouldn't need - you aren't trying to get standard errors here)..
3 - I believe that in the newer DHS rounds, they have begun to separately compute the wealth index for rural and urban areas. Or better said - I think they try to correct in some way, when they are computing the asset scores, for differences in rural and urban areas, and therefore the wealth deciles could be computed from the aggregate data giving you a single wealth decile that spans rural/urban households, and not from the rural/urban groups separately. I think there was a thread on the change in the asset/wealth index computation, but I don't have it handy.
I think (1) is probably what you want, but I also think it depends on what you want to do with these categories - a case could be made for (3) in some circumstances; (2) is just another way to do (1).
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
