[SciPy-User] fisherexact.py returns "NA" for large #s

[josef.pktd at gmail.com]

On Fri, May 7, 2010 at 5:44 PM, Vincent Davis <vincent at vincentdavis.net>wrote:

> @ Josef, I assume you know about this reference from the wikipedia page.
> http://mathworld.wolfram.com/FishersExactTest.html
>
>
I have it in my second comment to the ticket. But from this it's still a
long
way to figure out where the zero is supposed to go in the strict or weak
inequalities in the binary search. And why does the second path work
but not the first ?

I wasn't patient enough.

Josef



> Vincent
>
> On Fri, May 7, 2010 at 2:15 PM, <josef.pktd at gmail.com> wrote:
>
>> On Fri, May 7, 2010 at 3:45 PM, Pete Shepard <peter.shepard at gmail.com>
>> wrote:
>> > Hello List,
>> >
>> >
>> > I am using "fisherexact.py" to calculate the p-value of two ratios
>> however,
>> > when large #s are involved, it returns "NA". Is there a way to override
>> > this?
>>
>>
>> You mean fisherexact in http://projects.scipy.org/scipy/ticket/956 ?
>>
>> Do you have an example? Can you add it to the ticket?
>>
>> Do you have large ratios or large numbers in each cell?
>> If you have a large number of entries in each cell, then the chisquare
>> test or similar
>> asymptotic tests should be pretty reliable.
>>
>> Last time I tried, I didn't manage to get rid of incorrect results if
>> the first cell is zero.
>> And I didn't understand the details of the algorithm well enough to
>> figure out what's
>> going on (within a reasonable time).
>>
>> If you add some print statements, you could find out if the nan comes from
>> a
>> 0./0. division or from the hypergeometric distribution.
>> Do you get the same result if you permute rows or columns?
>>
>> fisherexact works very well over a large range of values, but I'm
>> waiting for someone
>> to provide a patch for the cases that don't work.
>>
>> Josef
>>
>>
>>
>>
>>
>> >
>> > TIA
>> >
>> > _______________________________________________
>> > SciPy-User mailing list
>> > SciPy-User at scipy.org
>> > http://mail.scipy.org/mailman/listinfo/scipy-user
>> >
>> >
>> _______________________________________________
>> SciPy-User mailing list
>> SciPy-User at scipy.org
>> http://mail.scipy.org/mailman/listinfo/scipy-user
>>
>
>   *Vincent Davis
> 720-301-3003 *
> vincent at vincentdavis.net
>  my blog <http://vincentdavis.net> | LinkedIn<http://www.linkedin.com/in/vincentdavis>
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20100507/24cf0831/attachment.html>