[SciPy-User] scipy.stats.linregress

[josef.pktd at gmail.com]

On Tue, Feb 12, 2013 at 11:45 AM, Jerome Kieffer <Jerome.Kieffer at esrf.fr> wrote:
> On Tue, 12 Feb 2013 10:48:33 -0500
> josef.pktd at gmail.com wrote:
>
>> On Tue, Feb 12, 2013 at 10:33 AM, Jerome Kieffer <Jerome.Kieffer at esrf.fr> wrote:
>> > On Tue, 12 Feb 2013 07:54:38 -0500
>> > josef.pktd at gmail.com wrote:
>> >
>> >> Volunteers for a pull request ?
>> >> and for checking that the test suite has a case with negative r, so
>> >> this doesn't get changed by accident.
>> >
>> > I can ... but I am not used to the tests in scipy
>>
>> Usually I just have to search the test folder by function.
>>
>
> the patch for the code is trivial...
> but get a valid test-case (currently breaking) is not trivial, especially that numpy.cov is casting to double
precision (I have seen R<-1 only with single precision arithmetic)

We should have a test for negative r, as backwards compatibility check
for the future,
but if it's too difficult to get one with r<-1, then we can leave two
lines untested.

I don't remember if I have ever seen a case with abs(r)>1.

Josef

>
> I am preparing a pull-request.
>
> Cheers,
> --
> Jérôme Kieffer
> On-Line Data analysis / Software Group
> ISDD / ESRF
> tel +33 476 882 445
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user