我有一个具有三个索引级别的数据框，我希望计算一个值偏离平均值的程度。但是根据我的指数，我对不同的群体有不同的意思。这是我尝试过的：
In [4]: df['count'].groupby(level=[0,1,2]).apply(lambda x: x-np.mean(x))
但是，我收到一个错误，我在下面插入了堆栈跟踪。我不确定为什么会出现这样的问题。
Exception Traceback (most recent call last)
<ipython-input-4-678992689ff2> in <module>()
----> 1 df['count'].groupby(level=[0,1,2]).apply(lambda x: x-np.mean(x))
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\groupby.pyc in apply(self, func, *args, **kwargs)
713 # ignore SettingWithCopy here in case the user mutates
714 with option_context('mode.chained_assignment',None):
--> 715 return self._python_apply_general(f)
716
717 def _python_apply_general(self, f):
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\groupby.pyc in _python_apply_general(self, f)
720
721 return self._wrap_applied_output(keys, values,
--> 722 not_indexed_same=mutated)
723
724 def aggregate(self, func, *args, **kwargs):
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\groupby.pyc in _wrap_applied_output(self, keys, values, not_indexed_same)
2520 if isinstance(values[0], (Series, dict)):
2521 return self._concat_objects(keys, values,
-> 2522 not_indexed_same=not_indexed_same)
2523 elif isinstance(values[0], DataFrame):
2524 # possible that Series -> DataFrame by applied function
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\groupby.pyc in _concat_objects(self, keys, values, not_indexed_same)
1258
1259 if isinstance(result, Series):
-> 1260 result = result.reindex(ax)
1261 else:
1262 result = result.reindex_axis(ax, axis=self.axis)
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\series.pyc in reindex(self, index, **kwargs)
2266 @Appender(generic._shared_docs['reindex'] % _shared_doc_kwargs)
2267 def reindex(self, index=None, **kwargs):
-> 2268 return super(Series, self).reindex(index=index, **kwargs)
2269
2270 @Appender(generic._shared_docs['fillna'] % _shared_doc_kwargs)
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\generic.pyc in reindex(self, *args, **kwargs)
1960 # perform the reindex on the axes
1961 return self._reindex_axes(axes, level, limit, tolerance,
-> 1962 method, fill_value, copy).__finalize__(self)
1963
1964 def _reindex_axes(self, axes, level, limit, tolerance, method,
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\generic.pyc in _reindex_axes(self, axes, level, limit, tolerance, method, fil
l_value, copy)
1974 new_index, indexer = ax.reindex(
1975 labels, level=level, limit=limit, tolerance=tolerance,
-> 1976 method=method)
1977
1978 axis = self._get_axis_number(a)
C:\Users\bchandra\AppData\Local\Continuum\Anaconda2\lib\site-packages\pandas\core\index.pyc in reindex(self, target, method, level, limit, tolerance)
5280 else:
5281 raise Exception(
-> 5282 "cannot handle a non-unique multi-index!")
5283
5284 if not isinstance(target, MultiIndex):
Exception: cannot handle a non-unique multi-index!
我的数据框看起来像这样：
Count
Cat SubCat
1 2 7
1 2 5
1 3 4
1 3 3
4 5 2
4 5 1
4 7 0
4 7 -1
为简单起见，假设我的索引是 2 级而不是 3 级。我想做的是按 (cat,Sub) 分组，这意味着 (category,subcategory)。
然后在我按 cat=1, sub=2 分组的第一种情况下找到所有组的平均值，这将是 7+5/2=6。然后我想分别找到7-6和5-6。
所以像df.groupby(level=[0,1]).apply(lambda x: x-np.mean(x))
一些在我的电脑上显示错误的虚拟代码（Pandas 版本 0.17.1）：
index=[]
[index.append(x) for y in range(25) for x in np.arange(2)]
subindex=[]
[subindex.append(10*x) for y in range(25) for x in np.arange(2)]
sample=pd.DataFrame({'count':np.arange(2*25),'cat':index,'sub':subindex,'date':np.random.randint(2*25)})
sample.set_index(['cat','sub'],inplace=True)
sample['count'].groupby(level=[0,1]).apply(lambda x: x-np.mean(x))