st: Re: weights from multiple dimensions

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

I think if you have unobserved combinations of the startification variables 
in your sample, then you can't get a population estimate without making some 
added assumptions.  In the simplest example, if you are stratifying on 
gender and your sample only includes males then you can't get an overall 
estimate for a population that includes females.

You can still get a population estimate for males though.  If you use 
weights based on a standard population and apply them to the sample, your 
resulting point estimate should still be OK as an estimator for the 
restricted part of the population that your sample covers, but not for the 
overall population.

Michael Blasnik
michael.blasnik@verizon.net

----- Original Message ----- 
From: "Munro, Cameron" <munro@rand.org>
To: <statalist@hsphsun2.harvard.edu>
Sent: Monday, October 17, 2005 5:42 AM
Subject: st: weights from multiple dimensions



> I'm pondering over a rather common problem, and am wondering whether there 
> is a package within Stata to help me do this. I need to reweight my sample 
> according to income, age and gender of the whole population. Obviously I 
> can get the population proportions for each combination (income x age x 
> gender) easily enough and then determine weights from the proportions in 
> my sample, but this doesn't account for unobserved combinations when I 
> apply such weights. I understand Stata 9 has some added post 
> stratification features which may (?) be of use, but I'm using Stata 8. 
> Can anyone offer any pointers?
>
> Many thanks,
>
> Cameron.
*
*   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/