On Thu, 16 Oct 2003 23:30:44 GMT, DZ <[EMAIL PROTECTED]> wrote:

>Duncan Murdoch <[EMAIL PROTECTED]> wrote:

>> That doesn't make sense.  Why would an infinite number of
>> permutations give a "true" p-value?
>
>Consider Fisher's exact test and its permutational analogue - produced
>by sampling tables with fixed margins at random. Fisher's p-value is a
>"true" p-value of the exact test in the sense that it is obtained by
>enumerating all relevant tables. As the number of permutations
>increases, the MC p-value approaches Fisher's (true) p-value.
>
>The CI I'm talking about is about the precision of the permutation p
>with respect to this "true" p. It does depend on p. "Sample size" here
>is the number of permutations.

Okay, I get what you were saying.  Yes, that makes sense.  Those CI's
measure the uncertainty of your Monte Carlo method of estimating the
p-value.  I didn't get that from the original post.

Duncan Murdoch
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to