[SciPy-dev] scipy.stats: sf for symmetric distributions.

josef.pktd@gmai... josef.pktd@gmai...
Wed Sep 16 16:13:13 CDT 2009


On Wed, Sep 16, 2009 at 3:53 PM, David Warde-Farley <dwf@cs.toronto.edu> wrote:
> On 16-Sep-09, at 3:17 PM, josef.pktd@gmail.com wrote:
>
>> Did you check for which distributions this would apply?
>> I remember there was some problem that pymvpa had with the r
>> distribution because the numerical integration close to the lower
>> bound didn't work.
>
> I haven't gone through the list exhaustively but I'm fairly certain it
> will make a positive difference for the t-distribution and the
> standard normal, especially since these are used in statistical tests
> where people care about their p-values.

good, your changes are similar to what Per proposed in
http://projects.scipy.org/scipy/changeset/5602/trunk/scipy/stats/distributions.py

There might be a few more cases like these.


>
> I just committed a fix for these two distributions, we can weigh the
> worth of subclassing or not further (incidentally, this breaks two
> ttest tests because they differ in the 17th decimal place, how sure
> are you of those hard-coded numbers?).
>

I got busy and just saw your follow up.
If there is a request for higher than standard precision of
floating point calculationss, then it should be a bug in the
test. I guess, I missed adding decimal if it is in one of
my tests.

Which ttest test is it?

Josef


> David
> _______________________________________________
> Scipy-dev mailing list
> Scipy-dev@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-dev
>


More information about the Scipy-dev mailing list