Negative binomial sampling bug when p=0

Negative binomial sample when p=0:

np.random.negative_binomial(1, 0)

I believe the output should be 0 in this case.

1.18.2 3.7.3 (default, Mar 31 2019, 11:50:42) [Clang 10.0.1 (clang-1001.0.46.3)]

[WarrenWeckesser]

P.S. @adamgayoso, thanks for reporting the issue!

[WarrenWeckesser]

@adamgayoso, the pull request gh-15914, in which the negative_binomial method of numpy.random.Generator is changed to disallow p=0, has been merged. With the master branch of numpy (and using the new numpy.random API), we have:

In [1]: import numpy as np

In [2]: np.__version__
Out[2]: '1.19.0.dev0+e78d076'

In [3]: rng = np.random.default_rng()

In [4]: rng.negative_binomial(1, 0)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-4-d61ac7f10aa5> in <module>
----> 1 rng.negative_binomial(1, 0)

_generator.pyx in numpy.random._generator.Generator.negative_binomial()

_common.pyx in numpy.random._common.disc()

_common.pyx in numpy.random._common.check_constraint()

ValueError: p <= 0, p > 1 or p contains NaNs

Note, however, that the function numpy.random.negative_binomial has not also been updated. That function is part of the “legacy” random API. The legacy code is usually only updated if a bug causes Python to crash. If you are writing new code, you should use the new API.