python - 向多索引数据帧添加级别

Question

我有一个多索引数据框:

iterables = [['bar', 'baz', 'foo', 'qux'], ['one', 'two']]
Index = pd.MultiIndex.from_product(iterables, names=['first', 'second'])
s  = pd.DataFrame(np.random.randn(8,2), index=Index)

它产生如下数据帧:

                     0         1
first second                    
bar   one     0.619602 -0.422137
      two    -0.372906  0.581697
baz   one    -0.968921 -0.014957
      two    -0.470649 -1.706410
foo   one     0.834609 -0.600675
      two     0.005306  0.109989
qux   one    -0.713642 -0.173100
      two    -1.155766 -0.365946

现在，在 multiIndex 的“第二”级别下，我想添加“三”，并使其等于“一”和“二”之间的差

在下面，A1 将等于 1.674156-(-1.061293)

A2 将等于 -1.380391-(-0.620890)

像这样:

                     0         1
first second                    
bar   one     1.674156 -1.380391
      two    -1.061293 -0.620890
      three   A1        A2
baz   one     0.839065 -1.985679
      two    -2.086971 -1.415384
      three  
foo   one    -1.673192 -0.559783
      two     0.135445 -1.101833
      three  
qux   one    -0.605042  1.814256
      two     0.182851 -1.819808
      three  

我该怎么做？我知道我可以取消 level1 的堆栈，进行比较，然后堆栈回来。只是好奇有更好的解决方案吗？

Answer 1

使用堆栈/取消堆栈/分配

d = s.stack().unstack(1)
d.assign(three=d.one - d.two).stack().unstack(1)

                     0         1
first second                    
bar   one     0.877453  0.214777
      two    -0.741437  0.185339
      three   1.618890  0.029437
baz   one     1.358314 -0.315129
      two     1.881061  0.253034
      three  -0.522747 -0.568162
foo   one     1.663033 -0.879386
      two    -0.658539 -0.331162
      three   2.321572 -0.548224
qux   one    -0.171216 -0.510144
      two     0.855199 -0.653881
      three  -1.026415  0.143737