使用 GroupBy 合并组内的两个数据框

Question

我有两个数据框，需要根据日期合并它们，但应该为每个组 (participant_id) 分别进行合并。

df1: 
   response_date  summary  epis_mark  participant_id
0     2012-01-04      0.0      False              13
1     2012-01-11      0.0      False              13
2     2012-01-19      0.0      False              13
3     2012-01-29      0.0      False              13
4     2012-02-02      0.0      False              13
0     2012-01-02      8.0       True              14
1     2012-01-10      5.0      False              14
2     2012-01-18      2.0      False              14
3     2012-01-24      1.0      False              14
4     2012-01-31      2.0      False              14
0     2012-01-07      4.0      False              17
1     2012-01-11      NaN      False              17
2     2012-01-18      4.0      False              17
3     2012-01-25      NaN      False              17
4     2012-02-01      NaN      False              17



  df2:
   response_date  summary  epis_mark  participant_id
0     2012-01-04     17.0       True              13
1     2012-01-11     18.0       True              13
2     2012-01-19     16.0       True              13
3     2012-01-29     15.0       True              13
4     2012-02-02     15.0       True              13
0     2012-01-02     12.0       True              14
1     2012-01-10      8.0       True              14
2     2012-01-18     21.0       True              14
3     2012-01-24     19.0       True              14
4     2012-01-31     20.0       True              14
0     2012-01-04      NaN      False              17
1     2012-01-11      NaN      False              17
2     2012-01-18      NaN      False              17
3     2012-01-25      NaN      False              17
4     2012-02-01      NaN      False              17

我需要获取一个数据帧 (wide)，其中每个 participant_id 在 response date 上独立完成合并。类似于：

>> pd.merge(df1[df1.participant_id == i], df2[df2.participant_id == i], on='response_date', how='outer')

但没有循环 i 和使用 groupby。

Answer 1

在 response_date 和 participant_id 上合并：

In [75]: pd.merge(df1, df2, on=['response_date', 'participant_id'], how='outer')
Out[75]: 
   response_date  summary_x epis_mark_x  participant_id  summary_y epis_mark_y
0     2012-01-04        0.0       False              13       17.0        True
1     2012-01-11        0.0       False              13       18.0        True
2     2012-01-19        0.0       False              13       16.0        True
3     2012-01-29        0.0       False              13       15.0        True
4     2012-02-02        0.0       False              13       15.0        True
5     2012-01-02        8.0        True              14       12.0        True
6     2012-01-10        5.0       False              14        8.0        True
7     2012-01-18        2.0       False              14       21.0        True
8     2012-01-24        1.0       False              14       19.0        True
9     2012-01-31        2.0       False              14       20.0        True
10    2012-01-07        4.0       False              17        NaN         NaN
11    2012-01-11        NaN       False              17        NaN       False
12    2012-01-18        4.0       False              17        NaN       False
13    2012-01-25        NaN       False              17        NaN       False
14    2012-02-01        NaN       False              17        NaN       False
15    2012-01-04        NaN         NaN              17        NaN       False

Answer 2

我不确定我是否理解正确。

您可以尝试以下方法：

pd.merge(df1, df2, on=['response date', 'participant_id'], how='outer')

使用 GroupBy 合并组内的两个数据框

Merge two dataframes within a group with GroupBy

python

merge

pandas

pandas-groupby