Posted by
Anthony Babinec on
URL: http://spssx-discussion.165.s1.nabble.com/Revision-control-for-SPSS-source-code-tp4267835p4268583.html
In my previous post, I am touting the Cytel Software historical "party
line." Mehta, Patel, and
colleagues made a lot of progress by working in the fixed marginals case.
They make
some arguments that try to justify that approach even which the sampling
model/research design
by which the table is thought to arise is not the fixed totals situation.
They have done
some work to move beyond this in a couple areas. One is exact power and
sample size for comparing
two binomials by Barnard's unconditional exact test for difference and
ratio. Yes, I am
aware of work that shows that Fisher's Exact Test for the 2x2 can be
improved on;
think that Agresti talks about it. I believe that Hirji, cited on the
Campbell page, was
as Cytel for a time. He subsequently published his own book; I don't own it.
My sense is that
there is no justification for the Cochran-Yates rule of thumb, which is what
I was responding
to in the original post that as I read it was not about the 2x2 case.
Tony Babinec
[hidden email]
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
Bruce Weaver
Sent: Monday, March 28, 2011 5:01 PM
To:
[hidden email]
Subject: Re: exact versus asymptotic p-value under normal circumstances
Hi Tony. I disagree about exact tests always trumping asymptotic tests.
Take the 2x2 table, for example. Fisher's exact test was designed for the
situation where both the row and column totals are fixed in advance (Model
I, as Barnard 1947 called it). If that is not the case, FET is known to be
too conservative. When the marginal totals are not fixed in advance, the
N-1 chi-square test (an asymptotic test) is a much better choice, IMO.
For more details about this particular example, take a look at Ian
Campbell's website, and his article in Statistics in Medicine.
http://www.iancampbell.co.uk/twobytwo/background.htmThe same thing applies to certain confidence intervals. E.g., exact
(Clopper-Pearson) confidence intervals for binomial proportions are
considered by many to be inferior to other methods. Here's a note from the
GraphPad site that makes this argument:
http://www.graphpad.com/articles/CIofProportion.htmCheers,
Bruce
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD