| Re: Mauritania [message #25392 is a reply to message #25356] |
Fri, 14 October 2022 08:47  |
Janet-DHS
Messages: 893
Registered: April 2022
|
Senior Member |
|
|
Following is a response from DHS staff member Tom Pullum:
For your first question, about education, I'll point out that hv106, hv107, hv108 are closely related. If you open the PR file in Stata and enter "tab hv107 hv106, summarize(hv108) means" you will see the relationship. If you know hv106 but don't know hv107, four possible options would be the following: (1) drop the case (good to avoid this); (2) assign hv107 to the lowest value, 0; (3) assign hv107 to the middle value, 3; (4) assign hv107 to the modal value, which for this survey is 6. I would probably choose option (3) but this sort of thing is at the discretion of the analyst and what you do with 52 cases out of nearly 35 thousand will have no effect on any conclusions. More complex imputation procedures would not be worth the trouble. A random imputation procedure would be particularly problematic because other researchers would not be able to replicate it.
Your second question is really about how to handle cases that are missing age, that is, have hv105=98. Virtually nothing can be done with such cases. They will not be eligible for the individual interviews and they will not appear as children in the BR or KR files. You can look at the report to see whether they appear in any tables at all. They should be dropped from any analysis--from denominators as well as numerators.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
