| Re: Append IPUMS_Merge to DHS Data [message #17828 is a reply to message #17810] |
Wed, 19 June 2019 15:14   |
boyle014
Messages: 78
Registered: December 2015
Location: Minneapolis
|
Senior Member |
|
|
Dear Yawo,
Apologies for my delay. The question about denormalizing the samples depends on your research question.
For descriptive statistics aggregated to the country level, use the appropriate IPUMS DHS weighting variable (perweight for women; dvweight for the domestic violence module, etc.). If you want to create confidence intervals, you'll want to use the svyset command in Stata. There is no reason to denormalize weights for this type of research.
If you are running a regression analysis on the pooled data, you will want to use a multilevel or mixed model, or possibly use country or region fixed effects. Weights are less essential since you're controlling for the characteristics (like urban/rural) that went into the construction of the weights.
If you want to calculate the total number of women in all of these countries who are experiencing something, then you need to apply the regular weights and inflate the sample sizes so that they reflect the number of women aged 15-49 in each country. In other words, this is when denormalization is necessary.
Let me know if you have more questions. Feel free to share your research question if you want more specific guidance.
Professor Elizabeth Boyle
Sociology & Law, University of Minnesota, USA
Principal Investigator, IPUMS-DHS
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
