Use of weights on subsetted data [message #23712] |
Fri, 12 November 2021 03:25 |
margovg
Messages: 3 Registered: March 2018 Location: Amsterdam
|
Member |
|
|
Hello,
I am using the Bangladesh DHS survey for 2017-18 to construct a pseudo baseline for an intervention. I have extracted the clusters which fall into my intervention area using the GPS coordinates. I wonder if, doing the analysis on this subset of clusters, requires me to adjust the weights.
Any advice or documentation on this would be appreciated!
Best, Margo
|
|
|
|
|
Re: Use of weights on subsetted data [message #23755 is a reply to message #23751] |
Wed, 24 November 2021 13:57 |
Bridgette-DHS
Messages: 3208 Registered: February 2013
|
Senior Member |
|
|
Following is another response from DHS Senior Sampling Specialist, Mahmoud Elkasabi:
Thanks for sending the details. The current survey weights are calculated so that the selected sample represent the whole division by urban/rural. As far as I understood, your target areas are the 62 unions only, and therefore your survey weight should be calculated so that the selected clusters are representing clusters on the 62 unions only and not all clusters within the division. I assume, you should be able to do so easily in the intervention survey, since it is part of the selection process. For the DHS to be used as a baseline, if you have access, it would be better if you can adjust the survey weights so it only represent the 62 unions. This might be a complicated process and prone to errors. The other option would be to use the survey weights as they are with the limitations discussed above considered.
|
|
|