Re: st: fishers exact test

[Marcello Pagano <pagano@hsph.harvard.edu>]

No problem.  It is saying that all tables that satisfy
these marginal constraints are as likely or less
likely than the table you have observed, within
round-off.  Round-off also explains the
1.00000000012.

A better name than exact, since it rarely is, would
be something like permutation.

m.p.

Richard_Lenhardt@rush.edu wrote:

> Hello,
>
> I have a 2x2 table:
>
> first row:               41   12
> second row:         6      1
>
> Running tabi gives a p-value for Fisher's exact test of 1.0 (two sided) and
> 0.52 (one sided).  The hypothesis suggests using two sided value.
>
> A reviewer commented that p-values, unless the data is quite unique, should
> not be 1.0
>
> Why is the p-value for Fisher's exact test exactly 1.0?  Does this make
> sense?
> Moreover, after viewing "return list," the p-value isn't 1.0 but
> 1.00000000012.  How can a p-value exceed 1.0?
>
> Also, looking on the web, some sites calculate the following:
>
> Two sided p-values for p(O>=E|O<=E)
>  p-value= 1.0000000000* (the sum of small p's)
> Two sided p-value for p(O>E|O<E)
>  p-value= 0.6385656450 (the sum of small p's)
>
> Should I be using a different test?
>
> Thank you,
> Richard Lenhardt
>  
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/