将数据透视表彼此堆叠

时间:2019-09-11 08:53:12

标签: pandas python-2.7 dataframe pivot-table

我有一个包含以下列的数据框:

Index(u['Stock','EarnDate','Last','Settle','Change'],dtype='object')

EarnDate是反映下一个收益发布日期的日期。

我创建了一个数据透视表:

pivot = pd.pivot_table(df, index='EarnDate',columns='Stock'),dtype='object')

这给了我以下输出

                 Last         Settle              Chg
Stock          Stock1  Stock2 Stock1  Stock2   Stock1  Stock2
EarnDate                                          
2019-10-01        NaN  5.55      NaN  5.55        NaN   +1
2019-11-01      65.91  3.43    62.91  6.55       -.5    +2
2019-12-01      62.97  6.87    61.97  7.00       +.4    +3
2020-01-01      63.33  6.66    61.38  9.50       -.3    +4
2020-02-01      60.91  5.98    60.99  8.50       +.2    +5
2020-03-01      60.71  6.23    60.70  7.50       -.15   +6

我想做的是按Stock将Last,Settle,Chg,Chant字段分组,这样看起来像这样:

Stock                 Stock1                 Stock 2
                Last  Settle Chg          Last Settle Chg
EarnDate                                          
2019-10-01        NaN   NaN  NaN          5.55  5.55   +1
2019-11-01      65.91 62.91   -.5         3.43  6.55   +2 
2019-12-01      62.97 61.97   +.4         6.87  7.00   +3
2020-01-01      63.33 61.38   -.3         6.66  9.50   +4
2020-02-01      60.91 60.99   +.2         5.98  8.50   +5   
2020-03-01      60.71 60.70  -.15         6.23  7.50   +6 

我尝试了各种stack()/ unstack()组,但均未成功。有人可以带我回家吗?谢谢!

1 个答案:

答案 0 :(得分:1)

DataFrame.swaplevelDataFrame.reindex一起使用:

mux = pd.MultiIndex.from_product([['Stock1', 'Stock2'], ['Last', 'Settle', 'Chg']])
df = df.swaplevel(0,1, axis=1).reindex(mux, axis=1)
print (df)
           Stock1              Stock2           
             Last Settle   Chg   Last Settle Chg
2019-10-01    NaN    NaN   NaN   5.55   5.55   1
2019-11-01  65.91  62.91 -0.50   3.43   6.55   2
2019-12-01  62.97  61.97  0.40   6.87   7.00   3
2020-01-01  63.33  61.38 -0.30   6.66   9.50   4
2020-02-01  60.91  60.99  0.20   5.98   8.50   5
2020-03-01  60.71  60.70 -0.15   6.23   7.50   6

原因是如果使用DataFrame.sort_index在第二级中获得不同的列顺序:

df = df.swaplevel(0,1, axis=1).sort_index(axis=1, level=0)
print (df)
           Stock1               Stock2             
              Chg   Last Settle    Chg  Last Settle
2019-10-01    NaN    NaN    NaN      1  5.55   5.55
2019-11-01  -0.50  65.91  62.91      2  3.43   6.55
2019-12-01   0.40  62.97  61.97      3  6.87   7.00
2020-01-01  -0.30  63.33  61.38      4  6.66   9.50
2020-02-01   0.20  60.91  60.99      5  5.98   8.50
2020-03-01  -0.15  60.71  60.70      6  6.23   7.50