如何基于二级标准从多索引的pandas.Series对象中提取? - python

我有一个pandas DataFrame样本对象,在下面对其进行分组相关计算。最后,我想查看Value1Value2之间的时间序列相关性

data = [
(1, 'alpha', 3, 101, 2),
(1, 'beta', 2, 102, 3),
(1, 'gamma', 5, 103, 4),
(2, 'alpha', 2.5, 101, 1),
(2, 'beta', 2.2, 105, 2),
(2, 'gamma', 5, 100, 0),
(3, 'alpha', 2.1, 102, 0),
(3, 'beta', 2.0, 102, 3.3),
(3, 'gamma', 5, 100, 2),
]

datapd = pandas.DataFrame(data, columns=('Time', 'ID', 'Value1', 'Value2', 'Value3'))
corrvals = datapd.groupby('Time').corr()

因此,当我查看corrvals['Value1']时,我只想选择所有Value2项目。但是,它们在Time之后。例如。 corrvals['Value1'].index.values显示:

array([(1, 'Value1'), (1, 'Value2'), (1, 'Value3'), (2, 'Value1'),
       (2, 'Value2'), (2, 'Value3'), (3, 'Value1'), (3, 'Value2'),
       (3, 'Value3')], dtype=object)

我如何在第二个元组中要求索引为Value2的所有值,而第一个元组中没有要求?

python大神给出的解决方案

您可以使用新的IndexSlice:

In [17]:
idx = pd.IndexSlice
corrvals.loc[idx[:,'Value2']]

Out[17]:
Time        
1     Value1    0.654654
      Value2    1.000000
      Value3    1.000000
2     Value1   -0.725288
      Value2    1.000000
      Value3    0.944911
3     Value1   -0.999569
      Value2    1.000000
      Value3   -0.121560
Name: Value2, dtype: float64

Slice:

In [18]:
corrvals.loc[slice(None),'Value2']

Out[18]:
Time        
1     Value1    0.654654
      Value2    1.000000
      Value3    1.000000
2     Value1   -0.725288
      Value2    1.000000
      Value3    0.944911
3     Value1   -0.999569
      Value2    1.000000
      Value3   -0.121560
Name: Value2, dtype: float64

或将axis=0传递给loc:

In [19]:
corrvals.loc(axis=0)[:,'Value2']

Out[19]:
               Value1  Value2    Value3
Time                                   
1    Value2  0.654654       1  1.000000
2    Value2 -0.725288       1  0.944911
3    Value2 -0.999569       1 -0.121560