st: Re: question on svymean for subpopulation

["Michael Blasnik" <michael.blasnik@verizon.net>]

In general, you should not use -if- to select cases when working with survey 
data (more specifically with svy commands).  Instead, you should use either 
a -by- option when offered or else use -subpop()-.  svy commands correctly 
use information about the entire sample design even when calculating 
estimates for a subpopulation if you use the subpop option.  When you 
use -if-, then the svy command essentially ignores the sampling info for 
everything excluded and will usually give you results you don't want -- and 
often lead to the dreaded singleton psu problem.  There are some exceptions 
when -if- will work fine, but it is best to just avoid the -if- option in 
most cases when working with svy data.  You should read the manual for a 
better understanding.

Michael Blasnik
michael.blasnik@verizon.net

----- Original Message ----- 
From: "Chenghui Li" <cheli@indiana.edu>
To: <statalist@hsphsun2.harvard.edu>
Sent: Monday, December 06, 2004 12:34 PM
Subject: st: question on svymean for subpopulation



> Hi,
>
> Could anyone please help me understand the difference between
> svymean age, by(female)
> and
> svymean age  if female=1?
>
> If I use the second command, I sometimes will get the error message about
> only one psu. I never have such a problem with the first command. How the
> psus and strata are handled using the by() option? And how it is different
> from using "if" command? My goal is to estimate average age by gender.
> Does the first command assume eaqual size of male and female? Intuitively,
> I would use the second one, or think these 2 are the same. I don't
> understand why they are different.
>
> THanks in advance for your help.
>
> Chenghui Li
*
*   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/