Source Code for Module mvpa.measures.corrcoef

 1  # emacs: -*- mode: python; py-indent-offset: 4; indent-tabs-mode: nil -*- 
 2  # vi: set ft=python sts=4 ts=4 sw=4 et: 
 3  ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ## 
 4  # 
 5  #   See COPYING file distributed along with the PyMVPA package for the 
 6  #   copyright and license terms. 
 7  # 
 8  ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ## 
 9  """FeaturewiseDatasetMeasure of correlation with the labels.""" 
10   
11  __docformat__ = 'restructuredtext' 
12   
13  from mvpa.base import externals 
14   
15  import numpy as N 
16   
17  if externals.exists('scipy', raiseException=True): 
18      # TODO: implement corrcoef optionally without scipy, e.g. N.corrcoef 
19      from scipy.stats import pearsonr 
20   
21  from mvpa.measures.base import FeaturewiseDatasetMeasure 
22   
23 -class CorrCoef(FeaturewiseDatasetMeasure): 
24      """`FeaturewiseDatasetMeasure` that performs correlation with labels 
25   
26      XXX: Explain me! 
27      """ 
28   
29 -    def __init__(self, pvalue=False, attr='labels', **kwargs): 
30          """Initialize 
31   
32          :Parameters: 
33            pvalue : bool 
34              Either to report p-value of pearsons correlation coefficient 
35              instead of pure correlation coefficient 
36            attr : basestring 
37              What attribut to correlate with 
38          """ 
39          # init base classes first 
40          FeaturewiseDatasetMeasure.__init__(self, **kwargs) 
41   
42          self.__pvalue = int(pvalue) 
43          self.__attr = attr 
44   
45   
46 -    def _call(self, dataset): 
47          """Computes featurewise scores.""" 
48   
49          attrdata = eval('dataset.' + self.__attr) 
50          samples = dataset.samples 
51          pvalue_index = self.__pvalue 
52          result = N.empty((dataset.nfeatures,), dtype=float) 
53   
54          for ifeature in xrange(dataset.nfeatures): 
55              samples_ = samples[:, ifeature] 
56              corr = pearsonr(samples_, attrdata) 
57              corrv = corr[pvalue_index] 
58              # Should be safe to assume 0 corr_coef (or 1 pvalue) if value 
59              # is actually NaN, although it might not be the case (covar of 
60              # 2 constants would be NaN although should be 1) 
61              if N.isnan(corrv): 
62                  if N.var(samples_) == 0.0 and N.var(attrdata) == 0.0 \ 
63                     and len(samples_): 
64                      # constant terms 
65                      corrv = 1.0 - pvalue_index 
66                  else: 
67                      corrv = pvalue_index 
68              result[ifeature] = corrv 
69   
70          return result 
71