| De-normalizing and weighting data for multiple countries [message #4100] |
Tue, 31 March 2015 12:25  |
bwbennett09
Messages: 3
Registered: March 2015
Location: Providence, RI
|
Member |
|
|
I am completing an analysis on eight nations in sub-Saharan Africa. I am interested in the relationship between HIV status (outcome) and measurements of women's empowerment (education, employment, etc.) I have read that I need to de-normalize my data and then weight it appropriately for population-level analyses. My questions, exactly how do I de-normalize this data (I have read a lot about multiplying weights v005*population) and then how would I weight the dataset across the 8 nations. I already have one giant dataset with the HIV data and other interested measurements for the 8 nations.
If it helps, these are the nations I am using:
DRC
Gabon
Guinea
Liberia
Mali
Niger
Namibia
Sierra Leone
Thanks for your help!
|
|
|
|
| Re: De-normalizing and weighting data for multiple countries [message #4108 is a reply to message #4100] |
Wed, 01 April 2015 15:40  |
Trevor-DHS
Messages: 805
Registered: January 2013
|
Senior Member |
|
|
If you don't need a pooled dataset, i.e. you are producing separate results for each country, then we recommend running analyses on separate datasets rather than pooling data together. In that case you don't need to worry about denormalization.
If you need a pooled dataset, then use the denormalization process to create a new weight variable, then use the new weight variable in place of the standard weight variable. Note that with pooled datasets you should continue to use svy commands (or the equivalent if using software other than Stata) and define strata codes that are unique to each survey.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
