Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: conditional estimates for survey data


From   "Sayer, Bryan" <BSayer@s-3.com>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   st: RE: conditional estimates for survey data
Date   Fri, 25 Oct 2002 10:49:57 -0400

It is hard to generalize about the difference between the conditional and
unconditional estimates of the variances.  It depends on the within stratum
effects.  If your population of interest happens to be a subset defined by
the stratification variable, you can get away with the conditional (the
estimates are actually stratum specific, and then combined).

Why would you need to adjust the weights of a sub-pop?  And what survey are
you using that has 7.5 million observations?  If at all possible, I would
suggest limiting the number of variables in your file and trying to maintain
the full sampling structure.

Bryan Sayer
Statistician, SSS Inc.
bsayer@s-3.com

-----Original Message-----
From: rpietro@duke.edu [mailto:rpietro@duke.edu]
Sent: Thursday, October 24, 2002 2:39 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: conditional estimates for survey data


Hi,


I am analyzing a data set containing national survey data with over 7.5
million observations. According to the Stata manual and Cochran's book on
sampling, I should use my entire data set to obtain unconditional
estimates for the standard error. Any spliting, including the use of
commands such as "if", would generate conditional estimates. Given that
running the analysis with over 7.5 million observations seems to be
computationally challenging, I am now trying to understand the meaning of
the conditional estimates I would obtain by restricting my population only
to the ones with the condition of interest. Here are my questions to the
list:

1. How do I interpret conditional estimates of standard error? Are they
generalizable to the target population (individuals with the condition of
interest in the entire country)?

2. Would these estimates be smaller than the ones obtained if it were
possible to make the estimates based on the entire patient population?

3. Is a conditional standard error necessarily biased in comparison to
the unconditional?

4. Can weights be adjusted for a given subpopulation? If so, I would
appreciate any references on the subject.


many thanks,


Rick


Ricardo Pietrobon, MD
Duke University Medical Center


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index