Home » Countries » India » Weighting variables in DHS India data (1992 and 1998)
| Re: Weighting variables in DHS India data (1992 and 1998) [message #8185 is a reply to message #8124] |
Tue, 08 September 2015 09:07   |
user_rm
Messages: 11
Registered: August 2015
|
Member |
|
|
Thanks for your email. I suppose my confusion just boils down to 1 point - if each child in 1 cluster has a weight of "1.148330" as you explain, then when I am collapsing observations by district (such that I have 1 averaged datapoint for every district), do I apply the 1.148330 weight to the whole district?
e.g if there are 5 kids in 1 district, each with a weight of 1.148330. When I collapse the HAZ score, I get 1 average value for the district. Now when running regressions on average data, do I apply a weight for 1.148330 to that district or would the weight by 1.148330*5?
A.
The confusion arises in this context because I am trying to calculate the proportion of villages with Anganwadi centres in each district, weighted by the number of kids treated in that district. I am confused as to which command I should use in step 2:
1. gen weight = v005/1000000
2. collapse(mean) weight, by(District Village kids) or collapse(rawsum) weight, by(District Village kids)?
3. collapse(mean) kids[pweight=weight], by(District)
where kids treated = 1 if the village in which the kid lives has Anganwadi centre.
If I use the collapse (mean) command, then within a district, an average kid in 1 village and the treated/untreated kid get the same weight
Village kids District weight
25 0 AHMADNAGAR 2.060619
37 1 AHMADNAGAR 2.060619
40 1 AHMADNAGAR 2.060619
53 1 AHMADNAGAR 2.060619
56 1 AHMADNAGAR 2.060619
132 1 AHMADNAGAR 2.060619
If I use the collapse(rawsum) command, then the weight is different according to the number of kids treated in each village, which is kind of what I would like.
Village kids District weight
25 0 AHMADNAGAR 30.90929
37 1 AHMADNAGAR 26.78805
40 1 AHMADNAGAR 26.78805
53 1 AHMADNAGAR 35.03052
56 1 AHMADNAGAR 18.54557
132 1 AHMADNAGAR 10.3031
It's tricky because for other variables, I think a normal mean collapse command would work
e.g collapse(mean) HAZ WAZ [pweight=weight], by(District)
B. I then use the weighted treated proportion variable and HAZ WAZ, etc in a regression. Now then again, I would be using the survey weights? Is that right? I am very confused about this.
e.g gen surveyweight = v005/1000000
collapse(mean) surveyweight, by(District)
reg HAZ WeightedProp MothEducYrs, [aweight= surveyweight]
C. As a separate point, I wanted to include some summary statistics (ie mean and sd) tables for individual children related data (not averaged). I used the estpost tabstat command. But they don't let me use 'iweight' or 'pweight' I had to use 'aweight' Do you think that is okay?
Please clarify.
Many thanks
R
[Updated on: Tue, 08 September 2015 10:11] Report message to a moderator |
|
|
|
 |
|
Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Thu, 06 August 2015 13:20 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Fri, 07 August 2015 11:00 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Wed, 12 August 2015 19:01 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Fri, 14 August 2015 14:07 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Sun, 16 August 2015 12:14 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Sun, 16 August 2015 17:17 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Tue, 18 August 2015 13:06 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Tue, 01 September 2015 08:34 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Tue, 08 September 2015 09:07 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
By: user_rm on Fri, 11 September 2015 06:12 |
|
|
Re: Weighting variables in DHS India data (1992 and 1998)
|
Goto Forum:
Current Time: Sat Jul 13 07:16:43 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
