Cross-sectional Data Merge & Adjust Weights [message #27330] |
Wed, 19 July 2023 09:42 ![Go to next message Go to next message](/theme/default/images/down.png) |
lijiaxuan
Messages: 2 Registered: June 2023
|
Member |
|
|
Dear everyone,
I am trying to combine SPA data from five countries in Sub-Saharan Africa, namely Kenya, Tanzania, Malawi, Congo and Senegal, to understand the quality of health services in a certain field and explore the influencing factors.
I have seen in some literature that merging datasets seems to require adjusting weights instead of ignoring them. However, in forums, there are many posts about DHS data merging, with almost no SPA, which has caused me great difficulties. Can someone help me?
The second question, about the SPA survey in Senegal from 2012 to 2019. I noticed that from 2012 to 2016, it seems that each year was not randomly sampled, how to adjust the weights if merging data from these years?
Any answer would be greatly appreciated....
|
|
|
|
Re: Cross-sectional Data Merge & Adjust Weights [message #27516 is a reply to message #27509] |
Mon, 28 August 2023 15:51 ![Go to previous message Go to previous message](/theme/default/images/up.png) |
Janet-DHS
Messages: 805 Registered: April 2022
|
Senior Member |
|
|
Following is a response from DHS Lead Statistician, Ruilin Ren:
Thanks for the question. The principles and the procedures to de-normalize DHS survey weights are addressed in many occasions and discussed a lot in the user forum. It does not depend on the type of the survey; the normalized weight should be divided by the sampling fraction for the sampling unit the weight is associated. It is true that the denormalization of SPA weight is not discussed much because it is a little bit more complex, not the procedure, but the information needed for the denormalization is usually not available, for provider weights and client weights. For the health facility weight, it is easy, and the information needed is provided in the SPA final report: the total number of health facilities in the master list and the number of health facilities interviewed. While for the provider weights and the client weights, the total number of providers and the total number of clients, by client type, are usually not available and we cannot advise the users where to find them. That is a difficult point, currently we do not have any advice.
|
|
|