[SciPy-User] Fisher exact test, anyone?
Tue Nov 16 18:10:19 CST 2010
On Tue, Nov 16, 2010 at 11:45 PM, Bruce Southey <email@example.com> wrote:
> On 11/16/2010 07:04 AM, Ralf Gommers wrote:
> On Mon, Nov 15, 2010 at 12:40 AM, Bruce Southey <firstname.lastname@example.org>wrote:
>> On Sat, Nov 13, 2010 at 8:50 PM, <email@example.com> wrote:
>> > http://projects.scipy.org/scipy/ticket/956 and
>> > http://pypi.python.org/pypi/fisher/ have Fisher's exact
>> > testimplementations.
>> > It would be nice to get a version in for 0.9. I spent a few
>> > unsuccessful days on it earlier this year. But since there are two new
>> > or corrected versions available, it looks like it just needs testing
>> > and a performance comparison.
>> > I won't have time for this, so if anyone volunteers for this, scipy
>> > 0.9 should be able to get Fisher's exact.
> All tests pass. There's only one usable version (see below) so I didn't do
> performance comparison. I'll leave a note on #956 as well, saying we're
> discussing on-list.
> I briefly looked at the code at pypi link but I do not think it is
>> good enough for scipy. Also, I do not like when people license code as
>> 'BSD' and there is a comment in cfisher.pyx '# some of this code is
>> originally from the internet. (thanks)'. Consequently we can not use
>> that code.
> I agree, that's not usable. The plain Python algorithm is also fast enough
> that there's no need to bother with Cython.
>> The code with ticket 956 still needs work especially in terms of the
>> input types and probably the API (like having a function that allows
>> the user to select either 1 or 2 tailed tests).
> Can you explain what you mean by work on input types? I used np.asarray and
> forced dtype to be int64. For the 1-tailed test, is it necessary? I note
> that pearsonr and spearmanr also only do 2-tailed.
> I have no problem including this if we can agree on the API because
> everything else is internal that can be fixed by release date. So I would
> accept a place holder API that enable a user in the future to select which
> tail(s) is performed.
It is always possible to add a keyword "tail" later that defaults to
2-tailed. As long as the behavior doesn't change this is perfectly fine, and
better than having a placeholder.
> 1) It just can not use np.asarray() without checking the input first. This
> is particularly bad for masked arrays.
> Don't understand this. The input array is not returned, only used
internally. And I can't think of doing anything reasonable with a 2x2 table
with masked values. If that's possible at all, it should probably just go
> 2) There are no dimension checking because, as I understand it, this can
> only handle a '2 by 2' table. I do not know enough for general 'r by c'
> tables or the 1-d case either.
> Don't know how easy it would be to add larger tables. I can add dimension
checking with an informative error message.
> 3) The odds-ratio should be removed because it is not part of the test. It
> is actually more general than this test.
> Don't feel strongly about this either way. It comes almost for free, and R
seems to do the same.
4) Variable names such as min and max should not shadow Python functions.
Yes, Josef noted this already, will change.
> 5) Is there a reference to the algorithm implemented? For example, SPSS
> provides a simple 2 by 2 algorithm:
Not supplied, will ask on the ticket and include it.
> 6) Why exactly does the dtype need to int64? That is, is there something
> wrong with hypergeom function? I just want to understand why the precision
> change is required because the input should enter with sufficient precision.
> This test:
fisher_exact(np.array([[18000, 80000], [20000, 90000]]))
becomes much slower and gives an overflow warning with int32. int32 is just
not enough. This is just an implementation detail and does not in any way
limit the accepted inputs, so I don't see a problem here.
Don't know what the behavior should be if a user passes in floats though?
Just convert to int like now, or raise a warning?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the SciPy-User