[Scipy-tickets] [SciPy] #86: Statistics Review: sem

SciPy scipy-tickets at scipy.net
Wed Nov 15 10:09:10 CST 2006


#86: Statistics Review: sem
-------------------------+--------------------------------------------------
 Reporter:  rkern        |        Owner:  rkern                   
     Type:  review       |       Status:  assigned                
 Priority:  normal       |    Milestone:  Statistics Review Months
Component:  scipy.stats  |      Version:  devel                   
 Severity:  normal       |   Resolution:                          
 Keywords:               |  
-------------------------+--------------------------------------------------
Comment (by smoerz):

 Replying to [ticket:86 rkern]:
 > The function {{{sem}}} in file source:trunk/Lib/stats/stats.py needs
 review.
 >
 > Please look over the StatisticsReview guidelines and add your comments
 below.

 1.) The function is WRONG. It gives the same result as scipy.stats.stderr
 (using N-1 and not N), whereas scipy.stats.tsem uses N and gives the
 correct result.
 It should read (stats.stats.py):

 def stderr(a, axis=0):
     a, axis = _chk_asarray(a, axis)

     return samplestd(a,axis) / sqrt(a.shape[axis])

 the test should be changed to (stats/tests/test_stats.py):
     """sqrt(samplevar(testcase))/sqrt(4)"""
     assert_approx_equal(y,0.5590169944)

 2.) Docstring is ok.

 The function should be unified with tsem, and maybe renamed/aliased to
 "sampleerr" for consistency with "samplevar" and "samplestd". The unified
 function should have argument "axis" as sem, and arguments "limits" and
 "inclusive" as tsem, and maybe "ignorenans" (functionality similar to
 nanstd). The algorithm should use the std method of numpy, as tsem does.
 This unification should be done also for functions std, samplestd, var,
 samplevar, mean, etc.

-- 
Ticket URL: <http://projects.scipy.org/scipy/scipy/ticket/86#comment:2>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list