熊猫创建新的数据框，从多个观测值中选择最大值

Question

I would like to make a new dataframe based on a max value from a column. However I have multiple observations from the same respondent and I only want to select the maximum value from the column value1 for each respondent. Here is a simplifyed example:

df:

   respondent  value1  value2
0           1       3      12
1           1       5      34
2           1       1      43
3           2       4      12
4           2       6      34
5           2       9      54
6           3       2      32
7           3       1       2
8           3       3      21

Here is the result I would like to have:

newdf:

  respondent  value1  value2
0           1       5      34
1           2       9      54
2           3       3      21

Any ideas?

Answer 1

The following achieves what you want and appears to be faster than @CT Zhu's answer:

In [30]:

df.loc[df.groupby('respondent').value1.idxmax().values]
Out[30]:
   respondent  value1  value2
1           1       5      34
5           2       9      54
8           3       3      21
In [31]:

%timeit df.loc[df.groupby('respondent').value1.idxmax().values]
%timeit df[df.groupby('respondent').value1.transform(lambda x: x==x.max())]
%timeit df.sort(['respondent', 'value1'], ascending=[1,0]).groupby('respondent').head(1)
100 loops, best of 3: 1.76 ms per loop
100 loops, best of 3: 2.99 ms per loop
100 loops, best of 3: 4.42 ms per loop

Also the above was achieved on pandas version 0.12.0 64-bit using python 3.3