Re: st: svyset for a multistage sampling design

["Michael I. Lichter" <mlichter@buffalo.edu>]

Tomeka,

What I said was "If you don't supply a finite population correction 
option for the first stage [svyset psu, fpc()], svyset will ignore your 
second stage." When no FPC is supplied, Stata assumes that sampling was 
done with replacement or that the sampling fraction(s) were small enough 
that sampling might as well have been done with replacement and so the 
details of the second and later stages (if any) are irrelevant to 
variance calculation. The reasoning for this is explained on pp. 142-143 
of the Stata 10 Survey Data [SVY] manual. If your sampling was done 
using a "method which allows variances to be estimated as if selection 
was done 'with replacement'" then you don't need to provide an FPC and 
you don't need to specify second-stage details. In other words, you can 
safely ignore "school".

Michael

Tomeka Davis wrote:
> Thanks.  I have one other question.  In response to the question about needing an fpc in STATA, a member of the data collection team responded by saying:
>
> "STATA can incorporate a finite population correction (fpc), but does not require one. In the data, primary sampling units (PSUs) were selected using the Durbin method which allows variances to be estimated as if selection was done "with replacement." When sampling is done with replacement, an fpc is unnecessary."
>
> Is there a way to specify this in STATA without the strata being ignored?
>
> Thank you for your help.
>
>
>   
> > > > "Michael I. Lichter" <MLichter@Buffalo.EDU> 8/20/2009 7:31 PM >>>
> > > >         
> Tomeka,
>
> 1. If you don't supply a finite population correction option for the 
> first stage [svyset psu, fpc()], svyset will ignore your second stage.
> 2. What are your school strata and how many of the strata have only only 
> psu in them -- and why? It doesn't look to me like school is a 
> stratifying variable -- it's your second stage sampling unit which will 
> be ignored without the FPC.
>
> -ml
>
> Tomeka Davis wrote:
>   
> > I am trying to use the set of svy commands to analyze data collected using a multi-stage design.  The first stage of sampling involved counties that were selected proportional to their size (N~100).  Next, schools were sampled (N~400), and then students were randomly sampled within schools.
> >  
> > In this context, the county is the psu and the school is the strata. I am not sure how strata with one sampling unit were treated.  Because I must have this option in  STATA, I choose the certainty option since it contributes nothing to the standard error.  Without the option, the program does not calculate standard errors.
> >  
> > My syntax looks like :  svyset county [pweight=studentweight], strata (school) ||_n, singleunit (certainty)
> >  
> > The problem is that I am not clear if STATA is using all of the stages.  After I enter the command, I get a message that says:
> >  
> > "Note:  stage 1 is sampled with replacement; all further stages will be ignored".
> >  
> > I would appreciate any comments.
> >
> >
> > *
> > *   For searches and help try:
> > *   http://www.stata.com/help.cgi?search 
> > *   http://www.stata.com/support/statalist/faq 
> > *   http://www.ats.ucla.edu/stat/stata/ 
> >   
> >     
>
>   

--
Michael I. Lichter, Ph.D. <mlichter@buffalo.edu>
Research Assistant Professor & NRSA Fellow
UB Department of Family Medicine / Primary Care Research Institute
UB Clinical Center, 462 Grider Street, Buffalo, NY 14215
Office: CC 126 / Phone: 716-898-4751 / FAX: 716-898-3536

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/