[Scipy-tickets] [SciPy] #1634: scoreateprecentile return the worng value when used on pandas.Series

SciPy Trac scipy-tickets@scipy....
Tue Mar 27 09:57:06 CDT 2012


#1634: scoreateprecentile return the worng value when used on pandas.Series
----------------------------------------+-----------------------------------
 Reporter:  imrisofer                   |       Owner:  somebody   
     Type:  defect                      |      Status:  new        
 Priority:  normal                      |   Milestone:  Unscheduled
Component:  scipy.stats                 |     Version:  0.10.0     
 Keywords:  scoreateprecentile, pandas  |  
----------------------------------------+-----------------------------------

Comment(by josefpktd):

 I'm not sure what the answer is.

 My impression: it's a Pandas or user problem

 It looks like the function is written for np.asanyarray and not for
 asarray. It returns the same array subclass as the input for matrix and
 masked arrays.

 (I never checked this function in detail, and I just saw that limit
 doesn't work with 2d arrays.)

 As basic policy we need to be able to assume something about array
 subclasses. For example when I rewrote stats.zscore, I used np.asanyarray
 and made sure it works for matrices and masked arrays. I would think that
 we can assume that basic algebraic operations and indexing/slicing works
 in the same way as for numpy arrays. (besides matrix multiplication.

 as aside
 stats.zscore returns different values if it is called with an ndarray or
 with a pandas series (because of different default ddof)

 {{{
 >>> stats.zscore(a).std()
 1.0
 >>> stats.zscore(b).std()
 1.0
 >>> type(stats.zscore(b))
 <class 'pandas.core.series.Series'>
 >>> np.std(np.asarray(stats.zscore(b)))
 0.99498743710661997
 }}}

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1634#comment:1>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list