| Re: Multilevel weighting issue in Bangladesh DHS data [message #30540 is a reply to message #30538] |
Tue, 17 December 2024 07:26  |
Bridgette-DHS
Messages: 3214
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
I worked with the PR file to see how hv022 is related to hv024 and hv025 (you could do it with the IR file too). I opened the PR file and reduced it to one record per household ("keep if hvidx==1"). Then I entered "tab hv024 hv025, summarize(hv022)". This shows that, as you expected, the strata are combinations of region (or governorate) and urban/rural, but there are two exceptions. The means for urban Zarqa (hv024=13) and urban Mafraq (hv024=22) are not integers. I then entered "tab hv022 if hv024==13 & hv025==1" and "tab hv022 if hv024==22 & hv025==1". These show that urban Zarqa is actually two strata (two values of hv022), 5 and 7. Urban Mafraq is also two strata, 12 and 14.
If you do "tab hv001 hv022 if hv022==5 | hv022==7" you will find that stratum 5 is clusters 199 through 266 and stratum 7 is clusters 272 through 292. You can count up how many clusters are in stratum 5 and stratum 7, and similarly for stratum 12 and stratum 14. I think that in this way you can adjust the procedure and avoid that Stata error message.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
