Re: st: Re: Confidence intervals using svymean

[Nick Winter <nw53@cornell.edu>]

Note that this solution only works when there are no missing data, or the 
"complete" option is specified with -svymean-.  If the various means are 
calculated based on different samples, then the estimates are returned in 
e(est), not e(b); the variances are in e(V_db); and the degrees of freedom 
may vary from case to case.  So:


> svymean x1 x2
> mat v=e(V)
> mat b=e(est)
svymean x1 x2
mat v=e(V_db)
mat b=e(est)
mat dof = e(_N_psu) - e(_N_str)

scalar tval1 = invttail(dof[1,1],.025)
scalar x1ci95lo = b[1,1]-sqrt(v[1,1])*tval1
scalar x1ci95hi = b[1,1]+sqrt(v[1,1])*tval1

scalar tval2 = invttail(dof[1,2],0.025)
scalar x2ci95lo = b[1,2]-sqrt(v[1,2])*tval2
scalar x2ci95hi = b[1,2]+sqrt(v[1,2])*tval2

Note that e(V_db) is NOT a diagonal matrix--it is a row-vector of 
variances, because with different samples, the covariances of the estimates 
cannot be calculated.

--Nick Winter

At 02:14 PM 7/25/2004 -0400, you wrote:

> You need to look at the var/covar matrix in e(V).  You can get the standard
> errors as the square roots of the diagonal elements. You need to calculate
> your own confidence intervals from this using info about d.f.  For example:
>
> svymean x1 x2
> mat v=e(V)
> mat b=e(b)
> scalar tval= invttail(`e(df_r)',.025)
> scalar x1ci95lo=b[1,1]-sqrt(v[1,1])*tval
> scalar x1ci95hi=b[1,1]+sqrt(v[1,1])*tval
> scalar x2ci95hi=b[1,2]+sqrt(v[2,2])*tval
> ...
> Michael Blasnik
> michael.blasnik@verizon.net
>
> ----- Original Message -----
> From: "Joao Pedro W. de Azevedo" <j.p.azevedo@ncl.ac.uk>
> To: <statalist@hsphsun2.harvard.edu>
> Sent: Saturday, July 24, 2004 8:03 PM
> Subject: st: Confidence intervals using svymean
>
>
> > Dear Stata users,
> > I would like to know if anyone knows how I could save the confidence
> > intervals after using svymean.
> > I've tried both ereturn list and return list, and neither one of them
> > presented me matrix that reseambled either the confidence intervals or the
> > std erros.
> > Am I missing something?
> > Many thanks in advance,
> > JP
> >
>
>
> *
> *   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/
--------------------------------------------------------
Nicholas Winter                           607.255.8819 t
Assistant Professor                       607.255.4530 f
Department of Government              nw53@cornell.edu e
308 White Hall            falcon.arts.cornell.edu/nw53 w
Cornell University
Ithaca, NY 14853-4601 

*
*   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/