| FGC--The Gambia: DHS 6--2013. [message #10532] |
Wed, 03 August 2016 03:01  |
Jawla
Messages: 7
Registered: August 2016
|
Member |
|
|
I am conducting analysis on attitudes towards FGC using the women data set with Stata. I have a question on weight.
Primary sampling unit v021
Sample strata v022
Women individual sample weight or the pw v005/1000000
fpc/10,233
Hence, this is a two stage stratified cluster sampling. Should I be worried about the secondary sampling unit and which variable is that? Do I need the househould dataset for this? How relevent is the v020 in the women data set. Should I use the single stage or multi stage stata syntax.Thanks!
Best, Jawla.
|
|
|
|
| Re: FGC--The Gambia: DHS 6--2013. [message #10560 is a reply to message #10532] |
Mon, 08 August 2016 09:43  |
Liz-DHS
Messages: 1516
Registered: February 2013
|
Senior Member |
|
|
A response from Shireen Assaf:
Quote:
Hello Jawla,
In DHS we have a multi stage stratified sample design with one strata variable v022 or v023 as you indicated. There is no need to use fpc. You can set up the svyset for your analysis as follows:
gen wt =v005/1000000
svyset v021 [pw=wt], strata(v022) singleunit(centered)
Thank you.
Best,
Shireen Assaf (DHS Senior Research Associate)
|
|
|
|
|
|
| Re: FGC--The Gambia: DHS 6--2013. [message #10584 is a reply to message #10563] |
Fri, 12 August 2016 22:42 |
Jawla
Messages: 7
Registered: August 2016
|
Member |
|
|
Shireen,
I played with the data after trying:
svyset v021 [pw=wt], strata(v022) singleunit(centered)
(Did not need to use -- gen wt =v005/1000000 -- as wt is already defined.)
after playing with some tabulation, the "population size" is still coming out approximately the same as the "number of observations." I think there is a problem -- I may be wrong. Evidence that I could be wrong is that a simple tab and a tab that includes the probability weights yield different answers. For example:
tab g119 [iweight=wt]
Female |
circumcision: |
continue or |
be stopped | Freq. Percent Cum.
--------------+-----------------------------------
1. Continued | 6,594.629 64.96 64.96
2. Stopped | 3,392.317 33.42 98.38
8. Don't know | 161.738423 1.59 99.97
9 | 3.19361798 0.03 100.00
--------------+-----------------------------------
Total | 10,151.878 100.00
. tab g119
Female |
circumcision: |
continue or |
be stopped | Freq. Percent Cum.
--------------+-----------------------------------
1. Continued | 6,305 62.12 62.12
2. Stopped | 3,674 36.20 98.32
8. Don't know | 165 1.63 99.94
9 | 6 0.06 100.00
--------------+-----------------------------------
Total | 10,150 100.00
However, as the Stata manual says better than I could:
pweights, or sampling weights, are weights that denote the inverse of the probability that the observation is included because of the sampling design.
So it seems to me that the population size reported in tabulations should be much higher than the number of observations used to create that tabulation.
Why is that when you do a two-way tabulation (for example),the number of observations and the population size are reported to be very similar to one another?
Thank you and I look forward to hearing from you at your earliest convenience.
Best, Jawla.
Best, Jawla.
|
|
|
|
|
|
| Re: FGC--The Gambia: DHS 6--2013. [message #10685 is a reply to message #10678] |
Mon, 29 August 2016 22:11 |
Jawla
Messages: 7
Registered: August 2016
|
Member |
|
|
Dear Shireen,
Thank you so much for your follow up on this matter. This is very much appreciated! I will keep you posted if I encounter any further issue with the DHS data.
Best, Jawla,
Best, Jawla.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
