Moises E. Rosas <[email protected]> asks about how to specify a
multistage sampling design for analyzing survey data:
> I'm wondering what the commands are for the subject, as I have found those
> for two-stage cluster sampling so far, and a wizard for multi-stage sampling
> plans in SPSS but none in Stata. I�m a kind of novice on Stata.
The linearized variance estimator for multistage survey designs is a new
feature of Stata 9. Use the -svyset- command to specify the survey design
variables, then use the -svy- prefix with the estimation command that fits
your model of interest. See -help survey- for list of estimation commands
that work with the -svy- prefix.
*** Example of -svyset-ing a multistage design:
Suppose we have a dataset from the following three-stage survey where each
stage was sampled without replacement.
Stage 1: we stratified on state and selected counties within each state.
Stage 2: we selected schools within each sampled county.
Stage 3: we selected 10% of the students within each sampled school.
In Stata 9, this would be specified by
. svyset county [pw=wgt], strata(state) fpc(ncounties)
|| school, fpc(nschools)
|| _n, fpc(nstudents)
where 'wgt' is the variable that contains the sampling weights,
Stage 1:
'state' identifies a state (e.g. "Texas")
'county' identifies a county within the state (e.g. "Brazos")
'ncounties' contains the number of counties in each state
Stage 2:
'school' identifies the schools within each 'county'
'nschools' contains the number of schools within each sampled county
Stage 3:
_n indicates that the observations represent the sampled students
'nstudents' contains the number of students within the sampled school
For more examples, see -[SVY] svyset- in the Stata 9 Survey Data Reference
Manual.
--Jeff
[email protected]
*
* 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/
