Home » Data » Weighting data » When weights are not supported
| Re: When weights are not supported [message #3210 is a reply to message #3204] |
Fri, 07 November 2014 16:22   |
Reduced-For(u)m
Messages: 292
Registered: March 2013
|
Senior Member |
|
|
Well... we are starting to get pretty far from area of expertise now (which is more about causal inference on a single RHS variable and not multi-level modeling).
That said - I wonder if the new class of -sem- (structural equation modeling) commands in Stata might help. They support the "svy" prefix with weights and strata, and I think you can model what you want using those commands, but like I said, getting pretty far from my expertise here.
http://www.stata.com/manuals13/sem.pdf [maybe start with intro 10 -- Fitting models with survey data (sem only)]
Let me know if these work - this problem has come up in enough contexts now that I would like to have an answer for people.
Oh - and out of curiosity, are you matching your country level covariates to the survey timing (so GDP in survey year) or to the birth or age timing (so relevant to when you were born/how old you are). For schooling, I'd say there is hazard of drop-out every year, and if a kid dropped out two years ago, then current condition may not mean much as a predictor. I'm just sort of curious as to how people are thinking about these things. Thanks! Good luck.
|
|
|
|
Current Time: Fri Dec 27 05:03:37 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
