TypeError：对2列进行分组时，插入列的索引与框架索引不兼容

Question

我有一个看起来像这样的数据集（+ 其他一些列）：

Value         Theme       Country
-1.975767     Weather     China
-0.540979     Fruits      China
-2.359127     Fruits      China
-2.815604     Corona      Brazil
-0.929755     Weather     UK
-0.929755     Weather     UK

我想找到按主题和国家分组后的值的标准偏差（如calculate standard deviation by grouping two columns 此处所述）

df = pd.read_csv('./Brazil.csv')
df['std'] = df.groupby(['themes', 'country'])['value'].std()

但是，目前，我收到此错误：

File /usr/local/Cellar/ipython/8.0.1/libexec/lib/python3.10/site-packages/pandas/core/frame.py:3656, in DataFrame.__setitem__(self, key, value)
   3653     self._setitem_array([key], value)
   3654 else:
   3655     # set column
-> 3656     self._set_item(key, value)

File /usr/local/Cellar/ipython/8.0.1/libexec/lib/python3.10/site-packages/pandas/core/frame.py:3833, in DataFrame._set_item(self, key, value)
   3823 def _set_item(self, key, value) -> None:
   3824     """
   3825     Add series to DataFrame in specified column.
   3826 
   (...)
   3831     ensure homogeneity.
   3832     """
-> 3833     value = self._sanitize_column(value)
   3835     if (
   3836         key in self.columns
   3837         and value.ndim == 1
   3838         and not is_extension_array_dtype(value)
   3839     ):
   3840         # broadcast across multiple columns if necessary
   3841         if not self.columns.is_unique or isinstance(self.columns, MultiIndex):

File /usr/local/Cellar/ipython/8.0.1/libexec/lib/python3.10/site-packages/pandas/core/frame.py:4534, in DataFrame._sanitize_column(self, value)
   4532 # We should never get here with DataFrame value
   4533 if isinstance(value, Series):
-> 4534     return _reindex_for_setitem(value, self.index)
   4536 if is_list_like(value):
   4537     com.require_length_match(value, self.index)

File /usr/local/Cellar/ipython/8.0.1/libexec/lib/python3.10/site-packages/pandas/core/frame.py:10985, in _reindex_for_setitem(value, index)
  10981     if not value.index.is_unique:
  10982         # duplicate axis
  10983         raise err
> 10985     raise TypeError(
  10986         "incompatible index of inserted column with frame index"
  10987     ) from err
  10988 return reindexed_value

TypeError: incompatible index of inserted column with frame index

Answer 1

使用DataFrame.expanding 并通过DataFrame.droplevel 删除新列的第一级应该是更简单的解决方案：

df['std']  = (df.groupby(['Theme', 'Country'])['Value']
                .expanding()
                .std()
                .droplevel([0,1]))
print (df)
      Value    Theme Country       std
0 -1.975767  Weather   China       NaN
1 -0.540979   Fruits   China       NaN
2 -2.359127   Fruits   China  1.285625
3 -2.815604   Corona  Brazil       NaN
4 -0.929755  Weather      UK       NaN
5 -0.929755  Weather      UK  0.000000

Answer 2

您可以使用rolling 方法计算每个组的累积标准差。

代码

import pandas as pd

# Create a sample dataframe
import io
text_csv = '''Value,Theme,Country
-1.975767,Weather,China
-0.540979,Fruits,China
-2.359127,Fruits,China
-2.815604,Corona,Brazil
-0.929755,Weather,UK
-0.929755,Weather,UK'''
df = pd.read_csv(io.StringIO(text_csv))

# Calculate cumulative standard deviations
df_std = df.groupby(['Theme', 'Country'], as_index=False)['Value'].rolling(len(df), min_periods=1).std()

# Merge the original df with the cumulative std values
df_std = df.join(df_std.drop(['Theme', 'Country'], axis=1).rename(columns={'Value': 'CorrectedStd'}))

输出

	Value	Theme	Country	CorrectedStd
0	-1.97577	Weather	China	nan
1	-0.540979	Fruits	China	nan
2	-2.35913	Fruits	China	1.28562
3	-2.8156	Corona	Brazil	nan
4	-0.929755	Weather	UK	nan
5	-0.929755	Weather	UK	0