In [4]: df = pd.DataFrame({'A':[1,1,2,2],'B':[1,2,3,4]})

In [5]: df.groupby('A').transform('max')
Out[5]: 
   B
0  2
1  2
2  4
3  4

Might be nice to have a deferred .transform() object (notice it doesn't take any args).

In [7]: df.groupby('A').transform().max()

Pretty sure this could be back-compat.

Comment From: jbrockmendel

It isn't clear to me that this would simplify things. Unless that changes, I'm -1.

Comment From: rhshadrach

Agreed - closing.