使用 pandas groupby 但保持其他列的顺序，然后在组内移动

Question

Say I have a pandas dataframe df as shown below -

    a   b
0   1   23
1   2   67
2   1   98
3   1   45
4   2   64
5   3   76

I want to use groupby so the resulting dataframe is as follows -

    a   b
0   1   23
1   1   98
2   1   45
3   2   67
4   2   64
5   3   76

I tried df.groupby(['a', 'b']).count().reset_index() but it changes the order of rows in 'b' Next, how do I shift within a group? For example I want to groupby 'a' and shift 'b' by 1 position. So the grouped and shifted dataframe would like this -

    a   b
0   1   NA
1   1   23
2   1   98
3   2   NA
4   2   67
5   3   NA

Answer 1

First you want the df to be sorted on a :

In [4501]: df = df.sort_values('a')

In [4502]: df
Out[4502]: 
   a   b
0  1  23
2  1  98
3  1  45
1  2  67
4  2  64
5  3  76

Now, you want to shift it by 1 per group:

In [4504]: df['b'] = df.groupby('a')['b'].shift()

In [4505]: df
Out[4505]: 
   a     b
0  1   NaN
2  1  23.0
3  1  98.0
1  2   NaN
4  2  67.0
5  3   NaN