Broadcasting fails in hypergeom.cdf

[gibbbone]

Broadcasting arrays works in hypergeom.pmf but throws an error when I try to do it with hypergeom.cdf .

Reproducing code example:

import numpy as np
from scipy.stats import hypergeom

# this works
x = np.array([[1,2,3],[4,5,6],[11,12,13]])
M,n,N = [20,7,12]
hypergeom.pmf(x,M,n,N)
hypergeom.cdf(x,M,n,N)

# this works
M=np.array([25])
n=np.array([3,6,9])
N=np.array([2,4,6])
hypergeom.pmf(x,M,n,N)

# this doesn't work
hypergeom.cdf(x,M,n,N)

Error message:

TypeError                                 Traceback (most recent call last)
<ipython-input-3-49873b933b6b> in <module>()
     15 
     16 # this doesn't work
---> 17 hypergeom.cdf(x,M,n,N)

\Anaconda3\envs\env2.7bis\lib\site-packages\scipy\stats\_distn_infrastructure.pyc in cdf(self, k, *args, **kwds)
   2917         if np.any(cond):
   2918             goodargs = argsreduce(cond, *((k,)+args))
-> 2919             place(output, cond, np.clip(self._cdf(*goodargs), 0, 1))
   2920         if output.ndim == 0:
   2921             return output[()]

\Anaconda3\envs\env2.7bis\lib\site-packages\scipy\stats\_distn_infrastructure.pyc in _cdf(self, x, *args)
   2778     def _cdf(self, x, *args):
   2779         k = floor(x)
-> 2780         return self._cdfvec(k, *args)
   2781 
   2782     # generic _logcdf, _sf, _logsf, _ppf, _isf, _rvs defined in rv_generic

\Anaconda3\envs\env2.7bis\lib\site-packages\numpy\lib\function_base.pyc in __call__(self, *args, **kwargs)
   2732             vargs.extend([kwargs[_n] for _n in names])
   2733 
-> 2734         return self._vectorize_call(func=func, args=vargs)
   2735 
   2736     def _get_ufunc_and_otypes(self, func, args):

\Anaconda3\envs\env2.7bis\lib\site-packages\numpy\lib\function_base.pyc in _vectorize_call(self, func, args)
   2808                       for a in args]
   2809 
-> 2810             outputs = ufunc(*inputs)
   2811 
   2812             if ufunc.nout == 1:

\Anaconda3\envs\env2.7bis\lib\site-packages\scipy\stats\_distn_infrastructure.pyc in _cdf_single(self, k, *args)
   2773 
   2774     def _cdf_single(self, k, *args):
-> 2775         m = arange(int(self.a), k+1)
   2776         return np.sum(self._pmf(m, *args), axis=0)
   2777 

TypeError: only length-1 arrays can be converted to Python scalars

Scipy/Numpy/Python version information:

'0.19.1'
'1.13.1'
sys.version_info(major=2, minor=7, micro=12, releaselevel='final', serial=0)

[eneskemalergin]

Hello guys is anyone working on this could provide a solution. It is really weird that all other functions work but cdf will thrown the TypeError. I am currently using 1-sf, but cdf would be really helpful thanks.

[WarrenWeckesser]

As of version '1.1.0.dev0+c10dc21', this is not fixed. I don't know if anyone is working on it.

Until it is fixed, here's a work-around using numpy.vectorize:

In [186]: from scipy.stats import hypergeom

In [187]: hypergeom_cdf = np.vectorize(lambda x, M, n, N, loc=0: hypergeom.cdf(x, M, n, N, loc))

In [188]: x = np.array([[0, 1, 2], [3, 4, 5], [6, 7, 8]])

In [189]: M = np.array([12])

In [190]: n = np.array([2, 5, 8])

In [191]: N = np.array([6, 7, 8])

In [192]: hypergeom_cdf(x, M, n, N)
Out[192]: 
array([[ 0.22727273,  0.04545455,  0.        ],
       [ 1.        ,  0.97348485,  0.59393939],
       [ 1.        ,  1.        ,  1.        ]])

[sethtroisi]

This seems fixed at head.

I thought it might have been fixed by #10091. I syncing to 10090 and code sample passed so it seems to be something else, because I can't find the PR that fixed this...

@WarrenWeckesser Would you mind testing that the code sample passes at head and closing if so?

[WarrenWeckesser]

@sethtroisi, I suspect it was fixed by #9900.

While verifying this issue was fixed, I found a different problem: the hypergeom methods don't handle lists of numbers in many cases, e.g.

In [60]: hypergeom.pmf(2, 12, 3, [6, 7, 8])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-60-e6197be96186> in <module>
----> 1 hypergeom.pmf(2, 12, 3, [6, 7, 8])

~/mc37scipy/lib/python3.7/site-packages/scipy-1.5.0.dev0+23b6191-py3.7-macosx-10.7-x86_64.egg/scipy/stats/_distn_infrastructure.py in pmf(self, k, *args, **kwds)
   3018         """
   3019         args, loc, _ = self._parse_args(*args, **kwds)
-> 3020         _a, _b = self._get_support(*args)
   3021         k, loc = map(asarray, (k, loc))
   3022         args = tuple(map(asarray, args))

~/mc37scipy/lib/python3.7/site-packages/scipy-1.5.0.dev0+23b6191-py3.7-macosx-10.7-x86_64.egg/scipy/stats/_discrete_distns.py in _get_support(self, M, n, N)
    432 
    433     def _get_support(self, M, n, N):
--> 434         return np.maximum(N-(M-n), 0), np.minimum(n, N)
    435 
    436     def _argcheck(self, M, n, N):

TypeError: unsupported operand type(s) for -: 'list' and 'int'

I'll create a new issue for that.

Since this issue is about broadcasting, and broadcasting now works, I'll close it.

Edit: github issue for the problem noted above: #12178