[NumPy-Tickets] [NumPy] #2031: numpy.cov behaviour change when called with 2 arrays (cross-covariance)

[NumPy Trac]

#2031: numpy.cov behaviour change when called with 2 arrays (cross-covariance)
-------------------------+--------------------------------------------------
 Reporter:  pierreh      |       Owner:  somebody   
     Type:  enhancement  |      Status:  new        
 Priority:  normal       |   Milestone:  Unscheduled
Component:  numpy.lib    |     Version:  1.6.1      
 Keywords:               |  
-------------------------+--------------------------------------------------

Comment(by bsouthey):

 I disagree with this ticket because it contains many obvious errors.
 [[BR]]

 1) It is very obvious from basic linear algebra that ''cov(X,X) would
 become equivalent to cov(X) '' is incorrect! [[BR]]

 2) "cov should only compute the *cross*-covariance of the two arrays" -
 this is in fact '''exactly''' what the cov function is doing so there is
 no problem.[[BR]]

 3) cov(X,Y) is just an equivalent of cov(concatenate(X,Y)) is incorrect as
 easily shown:
 {{{
 >>> X
 array([-2.1, -1. ,  4.3])
 >>> Y
 array([ 3.  ,  1.1 ,  0.12])
 >>> np.cov(np.concatenate([X,Y]))
 array(5.845666666666666)
 >>> np.cov(x,y)
 array([[ 11.71      ,  -4.286     ],
        [ -4.286     ,   2.14413333]])
 }}}

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/2031#comment:1>
NumPy <http://projects.scipy.org/numpy>
My example project