The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Service Provision Assessment (SPA) » sick child observation - client weights
sick child observation - client weights [message #4407] Thu, 21 May 2015 13:54 Go to next message
emilywj is currently offline  emilywj
Messages: 19
Registered: September 2013
Member
Hi,

For the sick child observation, how were the client weights constructed?

Thanks
Emily
Re: sick child observation - client weights [message #4411 is a reply to message #4407] Fri, 22 May 2015 10:41 Go to previous messageGo to next message
Sarah-DHS is currently offline  Sarah-DHS
Messages: 54
Registered: February 2013
Senior Member
Dear Emily,

Here is a response from Senior Sampling Statistician, Ruilin Ren, and Data Processing Specialist, Claudia Marchena.

For sick child interviews or any other type of client, the total number of clients present at the facility on the day of the interview and the total number of clients that ended up being interviewed are taken in consideration to calculate weights. Both questionnaire of the interviewed client (the observation of the consultation and the exit interview) must be completed in order to have a weight for the client, otherwise 0 is assumed. The base weight for the clients is the provider's weight because the clients were selected among those their provider was observed (interviewed ).

Please let us know if this answers your question.
Thanks,
Sarah-DHS
Re: sick child observation - client weights [message #5439 is a reply to message #4407] Wed, 27 May 2015 08:18 Go to previous messageGo to next message
emilywj is currently offline  emilywj
Messages: 19
Registered: September 2013
Member
Thanks, this was very helpful information.

Based on this response, frequencies across sick child observations should be weighted (pweight=clientwt) to account for unequal probabilities of selection given differences in client volumes on the interview date. Standard errors also need to account for data clustering at the provider level (based on a unique provider ID created by joining facil and provno numbers). Finally, the sample should also include only children/caregivers participating in both the observation and exit interview.

Is this a correct understanding?

Thanks again for your guidance on these various questions. It is greatly appreciated. SPA is such a wonderful resource!

With many thanks
Emily
Re: sick child observation - client weights [message #5471 is a reply to message #5439] Thu, 28 May 2015 09:07 Go to previous messageGo to next message
Sarah-DHS is currently offline  Sarah-DHS
Messages: 54
Registered: February 2013
Senior Member
Dear Emily,

Here is a response from Senior Sampling Statistician, Ruilin Ren, and Data Processing Specialist, Claudia Marchena.

From a data processing perspective, frequencies could be weighted or un-weighted depending on what do you need them for. Regarding the weighting, it seems that you are trying to produce estimations of frequencies with sampling errors, so the use of sampling weight is recommended. Only the weighted estimations are unbiased. Regarding the question on standard error estimation counting for clustering, it is the health facility which should be declared as cluster, not the provider. In all the statistical software which can take clustering effects into account in variance estimation, the "cluster" means the "primary sampling unit", the PSU. In the SPA settings, the PSU should be the health facility, not the providers. Regarding the units should be included in the sick child analysis, it is indeed correct.

Please let us know if this answers your question.
Thanks,
Sarah-DHS

Re: sick child observation - client weights [message #5547 is a reply to message #5471] Thu, 04 June 2015 11:49 Go to previous messageGo to next message
emilywj is currently offline  emilywj
Messages: 19
Registered: September 2013
Member
Thanks, apologies for my confusion.

Final question on weights = similar to DHS, the SPA client weights (clientwt) needs to be divided by 1,000,000 before applying it to generate frequencies among sick children. Is this correct?

Re: sick child observation - client weights [message #5552 is a reply to message #5547] Thu, 04 June 2015 16:51 Go to previous message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
Dear Emily,
Here is a response from Dr. Ruilin Ren:
Quote:
The answer to the question is yes. We multiplied the sampling weight by 1,000,000 in both DHS and SPA recode files. Thanks.
Next Topic: Release of Ethiopia SPA?
Goto Forum:
  


Current Time: Thu Mar 28 10:59:11 Coordinated Universal Time 2024