Re: Pooled data analysis from 25 countries [message #16753 is a reply to message #16618] |
Wed, 27 February 2019 09:06 ![Go to previous message Go to previous message](/theme/default/images/up.png) |
boyle014
Messages: 78 Registered: December 2015 Location: Minneapolis
|
Senior Member |
|
|
Dear Mayank,
DHS weights are not adjusted to take into account sample or population size in comparative analyses. There are separate discussions of when and how to do that in the User Forum though.
We are corresponding privately about your other questions, but I wanted to clarify my earlier post for other DHS users.
As to the specifics of your model, there are multiple ways to go. Including a survey fixed effect is the simplest approach and a good place to start. IPUMS DHS already includes a survey variable (SAMPLE), so you could use i.sample to do this with an IPUMS DHS data extract.
My post was providing the commands to do a multilevel model using IPUMS DHS variable names.
With respect to those commands, apologies--they included typos. The multilevel commands should have been "melogit" rather than "logit" and "mixed" rather than "regress". Further, one might try these without the weights in brackets first. Specifically:
melogit depvar || idhspsu:
estat icc
I hope your research is advancing well.
Professor Elizabeth Boyle
Sociology & Law, University of Minnesota, USA
Principal Investigator, IPUMS-DHS
|
|
|