Re: st: svy commands & weights with xtmixed

[Steven Samuels <samplerx@earthlink.net>]

On Dec 17, 2006, at 7:26 PM, Sarah Gill wrote:

> I am attempting to carry out multi-level modelling using Stata 9.0  
> xtmixed regression functions using longitudinal household panel data.
>
> I have found that 1) The svy function that I have used for  
> regression analyses in the past with cross-sectional datasets does  
> not appear to support any of the newer xt functions, and 2) xtmixed  
> does not appear to support weights.
>
> Has anyone come across a way to successfully analyse stratified  
> data and/or use pweights with xtmixed?
>
> Any informative will be greatly appreciated
Sarah:

The xt commands DO support pweights, at least in Stata 9.2.  The  
estimates should therefore be unbiased for the population values,  
even if strata are not specified. However standard errors will be  
model-based.  GLLAMM supports pweights and a cluster option for a  
variable that does not appear in the model. In XTMMIXED you could  
model PSU as the highest level random effect.

This still leaves the problem that standard deviations may vary  
between strata.  In GLLAMM you can apparently specify that variances  
differ between categories that need not appear as fixed effects  
(Section 1.1.1 of the GLLAMM manual). You may also be able to do this  
in XTMIXED by including a different random effect for each stratum at  
the appropriate level of the multi-level model.


Steve
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/