[Numpy-tickets] [NumPy] #921: numpy.random.hypergeometric: error for some cases

[NumPy]

#921: numpy.random.hypergeometric: error for some cases
--------------------------+-------------------------------------------------
 Reporter:  josefpktd     |       Owner:  somebody
     Type:  defect        |      Status:  new     
 Priority:  normal        |   Milestone:          
Component:  numpy.random  |     Version:  none    
 Severity:  normal        |    Keywords:          
--------------------------+-------------------------------------------------
 In my fuzz testing of scipy stats, I get sometimes a test failure. I think
 there is something wrong, random numbers are outside of support of the
 distribution, with numpy.random.hypergeometric for some cases:

 {{{
 >>> np.version.version
 '1.2.0rc2'
 }}}

 signature:
 hypergeometric(ngood, nbad, nsample, size=None)

 when sample size (number of draws, nsample) is small then random numbers
 look ok:
 {{{
 >>> np.random.hypergeometric(3,18,8,size=10)
 array([2, 1, 0, 2, 1, 1, 2, 1, 0, 3])
 >>> np.random.hypergeometric(3,18,9,size=10)
 array([2, 2, 0, 0, 1, 0, 1, 1, 3, 1])
 >>> np.random.hypergeometric(3,18,10,size=10)
 array([0, 2, 2, 0, 2, 2, 2, 0, 0, 2])
 }}}

 for sample size larger than 11 the random number are larger than possible
 (there are only 3 good balls in urn):

 {{{
 >>> np.random.hypergeometric(3,18,11,size=10)
 array([18, 16, 16, 17, 18, 17, 18, 17, 17, 16])
 >>> np.random.hypergeometric(3,18,12,size=10)
 array([17, 16, 16, 18, 17, 17, 18, 17, 16, 17])
 >>> np.random.hypergeometric(3,18,13,size=10)
 array([18, 17, 17, 16, 17, 17, 16, 16, 18, 16])
 >>> np.random.hypergeometric(3,18,14,size=10)
 array([16, 17, 17, 17, 17, 18, 17, 16, 18, 17])
 >>> np.random.hypergeometric(3,18,15,size=10)
 array([18, 18, 17, 16, 17, 16, 17, 17, 18, 18])
 }}}


 reversing the number of good and bad balls -> ok for small sample size:

 {{{
 >>> np.random.hypergeometric(18,3,5,size=10)    #ok
 array([5, 4, 5, 5, 4, 4, 4, 4, 5, 5])
 >>> np.random.hypergeometric(18,3,10,size=10)    #ok
 array([9, 8, 9, 7, 8, 8, 7, 8, 8, 9])
 }}}

 negative numbers for sample size >= 11:

 {{{
 >>> np.random.hypergeometric(18,3,11,size=10)
 array([-5, -5, -5, -7, -4, -5, -5, -4, -6, -6])
 >>> np.random.hypergeometric(18,3,13,size=10)
 array([-3, -5, -5, -4, -4, -4, -5, -5, -3, -4])
 >>> np.random.hypergeometric(18,3,14,size=10)
 array([-4, -3, -4, -3, -2, -4, -3, -4, -2, -4])
 >>> np.random.hypergeometric(18,3,15,size=10)
 array([-2, -2, -1, -1, -3, -2, -2, -2, -1, -2])
 >>> np.random.hypergeometric(18,3,16,size=10)
 array([-1, -1,  0, -1,  0, -2, -1, -2, -1, -2])
 }}}

-- 
Ticket URL: <http://scipy.org/scipy/numpy/ticket/921>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.