| Re: Compensation for Omission and Displacement [message #21799 is a reply to message #21787] |
Tue, 22 December 2020 07:46   |
Bridgette-DHS
Messages: 3214
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
As I recall, that comment in MR10 was just hypothetical. We did not define such weights, we just said that they could perhaps be developed, in the same way that with post-stratification methods it is possible to re-weight a sample to match a known distribution. For example, if a survey under-represents an age or ethnic group we can "weight up" the observed cases in those under-represented groups.
The main problem with re-weighting in your context is that the true numbers of omitted neonatal deaths is not known. That's why we didn't actually try to calculate such weights in MR10. Sorry, but I don't think this strategy will help.
Another issue is potential mis-classification, rather than omission. Some reported neonatal deaths may actually be stillbirths, and some reported stillbirths may actually be neonatal deaths. The event may be in the calendar but not in the birth history. In future surveys, DHS will collect pregnancy histories rather than birth histories and this will help with estimates of perinatal mortality.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
