Robert Bozick <rbozick@jhu.edu> wants to fit a random effects model using
survey data:
> I am working on an analysis that will require the use of a random effects
> model on a data set that includes two observations per individual. The
> general form of the code I am using to estimate the model is:
>
> xtreg y x1 x2 x3, re i(id)
>
> (id = the case id for each individual)
>
> The sample uses a stratified cluster design and I therefore have to use the
> svy commands when analyzing this data set.
>
> I am using Stata 8.0 and could not find any documentation on the application
> of svy commands with estimating a random effects model.
>
> Is there a svy command or some other procedure available so that I can
> obtain the correct standard errors for this model?
There are no estimation commands (in Stata 9 or earlier) that will fit a
random effects model using the linearized variance estimator for survey data.
If Robert were using Stata 9 and was willing to use -xtreg, mle- to fit the RE
model, he could use the -svy jackknife- prefix command to get design-based
variance estimates via the jackknife.
NOTE: The -xtreg, re- command does not accept weights, but
-xtreg, mle- allows -iweight-s.
If Roberts survey design variables were named swgt (sampling weight), strid
(strata id variable), and psuid (PSU id variable), he would type
. svyset psuid [iw=swgt], strata(strid)
. svy jackknife _b: xtreg y x1 x2 x3, mle i(id)
I'm assuming that -id- is nested within -psuid- (panels are nested within the
PSUs).
--Jeff
jpitblado@stata.com
*
* 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/