Re: st: proportion confidence intervals <0 and >1 in MI SVY

[Steve Samuels <sjsamuels@gmail.com>]

I don't have an -mi- example, but here is an example with plain -svy: prop-
-Steve

************CODE BEGINS***********
capture program drop _all
program backlogit
nlcom log($parm/(1-$parm))
scalar lparm = el(r(b),1,1)
scalar se  =  sqrt(el(r(V),1,1))
scalar bound = invttail(e(df_r),.025)*se
scalar ll = exp(lparm -bound*se)
scalar ul = exp(lparm +bound*se)
scalar ll = ll/(1+ll)
scalar ul = ul/(1+ul)
di %9.4f $parm   %9.4f ll  %9.4f ul
end

sysuse auto,clear
svyset _n
svy: prop rep78
global parm _b[1]
backlogit
********CODE ENDS*********************

On Tue, May 11, 2010 at 8:44 AM, Paul Shattuck <pshattuck@wustl.edu> wrote:
> Hi,
>
> I am analyzing imputed survey data using the MI SVY estimation commands. I am estimating proportions and requesting confidence intervals (syntax below). However, many of the confidence intervals are falling outside of 0,1 bounds. The tab confidence intervals available in plain SVY (without MI) are generated using a logistic transform that prevents this problem. However, the tab procedure is not available in MI estimation.
>
> Can anyone suggest a work-around that will let me obtain proportion confidence intervals in MI SVY estimation that remain within the 0,1 bounds?
>
> mi estimate: svy, subpop(if var1==1): proportion var2
>
> Thank you,
>
> Paul
>
>
>
>
>
> *
> *   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/
>



-- 
Steven Samuels
sjsamuels@gmail.com
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783

*
*   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/