Peter Dalgaard wrote:
Daniel Brewer wrote:
Hi,
I am trying to sort out a discrepancy between power calculations results
between me and another statistician. I use R but I am not sure what she
uses. It is on the proportions test and so I have been using
pwr.prop.test. I think I have tracked the problem down to pwr.prop.test
not using the continuity correction for the test (I did this by using
the java applet from
http://stat.ethz.ch/R-manual/R-patched/library/stats/html/power.prop.test.html).
So I was wondering whether:
1) Someone could confirm that pwr.prop.test does not use a continuity
correction in its calculation.
2) Someone could tell me either how to use pwr.prop.test or another
function to get the power of a prop.test with continuity correction.
The reason I want this is that I would normally apply the correction
when I actually used the test.
Many thanks
Dan
power.prop.test (sic) is relying heavily on asymptotic normality, as do
similar formulas. It doesn't use continuity correction, but if you're
working with such small group sizes, I suspect that the correction term
is the least of your worries and that direct simulation would be better.
(Another source of discrepancy, sometimes seen in textbooks, is that
authors use the null variance of p1-p2 also under the alternative. This
simplifies the formulas considerably, but it does assume that the actual
difference is rather small.)
R is Open Source. If you want a correction term, it is just a matter of
figuring out where to modify expressions like
p.body <- quote(pnorm(((sqrt(n) * abs(p1 - p2) -
(qnorm(sig.level/tside,
lower.tail = FALSE) * sqrt((p1 + p2) * (1 - (p1 +
p2)/2))))/sqrt(p1 *
(1 - p1) + p2 * (1 - p2)))))
by adding or subtracting 0.5 or 0.5/n in the appropriate places.
In addition to what Peter said, the continuity correction is in effect
an attempt to make the proportion test behave like Fisher's exact test
which is known to be conservative. We don't usually desire P-values
that are too large, so I don't recommend the continuity correction.
See the bpower.sim function in the Hmisc package for a simulation-based
method, and the reference below.
Frank
@Article{cra08how,
author = {Crans, Gerald G. and Shuster, Jonathan J.},
title = {How conservative is {Fisher's} exact test? {A}
quantitative evaluation of the two-sample comparative binomial trial},
journal = Stat in Med,
year = 2008,
volume = 27,
pages = {3598-3611},
annote = {Fisher's exact test; $2\times 2$ contingency table;size
of test; comparative binomial experiment;first paper to truly quantify
the conservativeness of Fisher's test;``the test size of FET was less
than 0.035 for nearly all sample sizes before 50 and did not approach
0.05 even for sample sizes over 100.'';conservativeness of ``exact''
methods}
}
--
Frank E Harrell Jr Professor and Chair School of Medicine
Department of Biostatistics Vanderbilt University
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
