| Senegal men's recode urban sample [message #23588] |
Fri, 15 October 2021 07:45  |
numbers2life
Messages: 2
Registered: October 2021
Location: Cape Town
|
Member |
|
|
Hi DHS
I have been working with men's recode data on home ownership in Senegal 2019 DHS, and I have run into some figure mismatches in the Statcompiler and the raw data. I can reproduce the figures at the national level for men who own a house solely, jointly, and solely and jointly, but I cannot do the same at the urban level. After digging deeper, I suspect that this is a sampling issue: at the national level, there are 3,365 observations in the raw data and in the Statcompiler (shown when unweighted observations is enabled). However, at the urban level there are 1,181 observations in the Statcompiler but 1,285 observations when filtering the raw data (mv025 == "urban"). Further complicating the issue, if I correctly interpreted the French, in the final report the urban sample size is shown as 1,248 observations.
My calculations are being run in R, pulling the raw data off the API using the rdhs package.
Please could you advise on how to resolve the sample mismatches? Is there some other filter than the Statcompiler applies to men's recode data that changes the sample size at the urban level? Why are the urban sample sizes different between the raw data, Statcompiler and final report, yet the same at the national level?
Cheers
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
