[SciPy-Dev] Owen's T function

[Evgeny Zhurko]

Maybe someone have borrowed constants before me but i didn't found it. I
included all constant and test data from Boost in last commit.

But i have new questions:

I implemented T1-T6 methods and tried test function on Boost's test data.
All tests passed when i use assert_almost_equal with decimal=1. About 30 of
400 examples have deviation 0.01 from real value, 40 of 400 examples have
deviation about 1e-10 and the others less then 1e-18. Boost have
improvement for some methods (
http://www.boost.org/doc/libs/1_52_0/boost/math/special_functions/owens_t.hpp).
Is it enough to insert copyright in source code if i'll implement this
algorithm into Scipy?

Test data have a very big size(about 100 character per number). I inserted
data into test_basic.py . May I move test data into txt or csv file and
read it in tests? Is it requires including copyright into file with test
data?

Test data from Boost not cover all methods(only T2, T4) as a result I want
to use
Owen, Donald B. "Tables for Computing Bivariate Normal Probabilities." (
http://projecteuclid.org/download/pdf_1/euclid.aoms/1177728074).

Pull request: https://github.com/scipy/scipy/pull/7120
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20170304/317015a9/attachment.html>