Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Extrapolated inferences and confidence intervals


From   Stas Kolenikov <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Extrapolated inferences and confidence intervals
Date   Wed, 24 Jun 2009 10:54:14 -0500

What exactly is your extrapolation procedure? Do you use any of the
demographic variables (for either the physician or the patient)? If
you just say, "Among the responding physicians, the side effect was
verified for 27% of the patients, and I will simply use that figure as
the overall prevalence" -- then in the simplest manner, you would have
the standard error of sqrt( 0.27(1-0.27)/ # physicians ) on top of
that (or 0.27*proportion reporting any side effects instead of 0.27 if
you are generalizing to the population of the patients). If you want
to make it a little fancier, you can think about finite population
aspects of your study and clustering of patients into physicians, in
which case you would want to -svyset physician- your data set and
compute the estimates of the total using poststratification: the
patients who did not report any side effects will have a
post-stratification weight of 1, and the patients who did, the weight
of 1/proportion contacted physicians.

On Wed, Jun 24, 2009 at 9:51 AM, Fred
Wolfe<fwolfe@arthritis-research.org> wrote:
> I wonder if anyone on the Stata list may be able to suggest a method
> to solve this problem using Stata.
>
> I am trying do determine the prevalence and confidence intervals of a
> side effect to a treatment. Several thousand treated subjects have
> taken a drug. They are asked if they had a specific side effect to the
> drug. Those that say no are classified as being side effect negative.
> Those that say yes are contacted to obtain the name of their
> physician, and the physician is asked to verify the side effect. Only
> about one third of physicians can be contacted, and from them we can
> get the proportion positive and negative for the side effect. Assume
> that the non-responses are unrelated to the outcome. I can extrapolate
> from the responding physicians to the non-responding physicians and
> calculate an overall prevalence of the side effect. What I would like
> to do is to obtain confidence intervals for this prevalence estimate,
> but I do not see an easy way to do this in Stata. I wonder if anyone
> might suggest a solution.
>
> Thanks,
>
> Fred
>
> --
> Fred Wolfe
> National Data Bank for Rheumatic Diseases
> Wichita, Kansas
> NDB Office  +1 316 263 2125 Ext 0
> Research Office +1 316 686 9195
> fwolfe@arthritis-research.org
>
> *
> *   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/
>



-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/



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