On Wed, 28 Jun 2000 15:22:06 +0200, Bernd Genser
<[EMAIL PROTECTED]> wrote:
BG >
" I have a number of studies who estimate an unknown proportion
(incidence of a rare disease). What is the correct way to estimate a
global proportion by a meta-analytic approach from different
independent studies?"
For an intelligent meta-analysis of two studies, you would possibly
consider them carefully and then plunk for one, or for the other.
If there were differences in diagnostic subtypes, then you might want
to specify totals in terms of subtypes. This could give the effect of
looking at "union" and "intersection" of the definitions, and
(correspondingly) of the counts.
BG >
"1) Weighting the studies' estimators by n (assuming a common
proportion but neglecting the properties of the binomial
distribution).
"2) Weighting the different estimators by the reciprocal of its
variances."
I have not looked at disease estimates very often. But from when I
have, I don't remember being TEMPTED to compute weighted estimates.
It might be different if I was a Public Health Officer, trying to
coordinate reports from a bunch of agencies, a bunch of sections, etc.
Still, I think I would have to be mainly focused on definitions and
competence, and not on weighting.
Maybe someone else will have a more relevant experience.
By the way, if I understand (1) and (2), the difference arises when
the two studies have estimators that are QUITE DIFFERENT. For
instance, if there are two measures in the same N, N_c or number of
cases *is* the variance, if you don't pool the estimates. This is a
pretty dubious practice: to compute and report an average of two
estimators, when you know that they are vastly different.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
