Calculation of standard error of stunting at small subpopulation such as district [message #12583] |
Sat, 17 June 2017 13:06 |
|
Hi,
I wish to calculate proportion of stunted children for the micro-level administrative units like district and sub-districts with their standard errors. I followed the usual method of complex survey method in SPSS. The problem I found that for the sub-population having single cluster the design effect is zero and also the standard error is zero. The SPSS plan code is given below.
COMPUTE WGT=V005/1000000.
EXECUTE .
* Analysis Preparation Wizard.
CSPLAN ANALYSIS
/PLAN FILE='C:\Users\dell\Downloads\BDHS2011.csaplan'
/PLANVARS ANALYSISWEIGHT=WGT
/SRSESTIMATOR TYPE=WR
/PRINT PLAN
/DESIGN STRATA=V023 CLUSTER=V001
/ESTIMATOR TYPE=WR.
May be I need the exact way for obtaining the standard error. If I avoid the sample design and use sampling weight only, I obtain the same estimates as I found using sampple design but the standard error is not now zero. Thus the issue is the design effect for the small sub-population. Is there any way to avoid such computation problem.
If I am working correctly and the results are theoretically reasonable, can you please inform me the exact reasons.
If there is any other technique, please suggest e hw can I solve the problem.
Regards,
Sumonkanti Das
Sumonkanti Das
|
|
|
|
|
Re: Calculation of standard error of subpopulation such as district [message #26594 is a reply to message #26591] |
Wed, 05 April 2023 09:08 |
Bridgette-DHS
Messages: 3167 Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
A standard error is calculated for an estimate, such as the estimated mean value of an indicator. The procedure to calculate a standard error is the same for any subpopulation (unless you use Bayesian methods for nested populations).
Are you trying to match a number in a DHS report? Are you interested in some specific indicator?
|
|
|