You should generally use subpop() and not -if- to select observations in a
svy analysis. If you're having problems with singleton psu's, you should
collapse strata in whatever way makes the most sense for your data --
sometimes there is no way that seems to make much sense, but just dropping
the data is often less sensible than some arbitrary strata re-assignment.
Michael Blasnik
[email protected]
----- Original Message -----
From: "David J. Bernstein" <[email protected]>
To: <[email protected]>
Sent: Thursday, April 08, 2004 8:04 PM
Subject: st: svylogit question / with a hypothetical
> Hello All,
>
> I have a svylogit question.
>
> Here is my hypothetical:
>
> Out of 20,000 patients I have a survey sample of 1000.
> In this sample I have weights that represent the
> racial mix of the complete population.
>
>
> Here is the program:
>
> svyset [pweight=weight], psu(ssn) strata(hospital)
>
> svylogit recovered race age _Iyear* (year dummies)
>
>
> Now if I decide to drop the hosiptals that have
> complete failure (i.e. recovered=0 for all)
> does re-running the svylogit still yield valid
> results? In other words, do I have a weighting
> problem from dropping, say, 50 observations because
> they are from hospitals that had 0 recovery? By doing
> this I eliminate stratid's with 1 observation (and
> crashing my run).
>
> What do you guys think?
>
> -David
>
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! Small Business $15K Web Design Giveaway
> http://promotions.yahoo.com/design_giveaway/
> *
> * 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/
*
* 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/
