Fix docs in groupby.tail

[kidphys]

The old docs is wrong where head() & tail() return the same result.
Change input of the example to see group data clearer.

[TomAugspurger]

I get a different index than your result:

In [9]: df.groupby("A", as_index=False).tail(1)
Out[9]:
   A  B
1  a  2
3  b  2

In [10]: df.groupby("B", as_index=False).head(1)
Out[10]:
   A  B
0  a  1
1  a  2

Is that a copy-paste error, or a bug?

[jorisvandenbossche]

you did a groupby 'B' instead of 'A'

[jorisvandenbossche]

So the index in the first example is indeed not correct, but for the second example it is OK (there Tom used a wrong groupby key)

[jreback]

@kidphys ping when you update this.

[jorisvandenbossche]

I think the index here is not OK:

In [16]: df.groupby('A', as_index=False).tail(1)
Out[16]:
   A  B
1  a  2
3  b  2

So 1,3 instead of 0,2

[kidphys]

@jorisvandenbossche you're right, updated.

[kidphys]

I just realize in my case index will be displayed no matter as_index is False or True. In the original doc, one call to groupby doesn't have as_index=False too. Is this a bug?

[jorisvandenbossche]

@kidphys head and filter are regarded as 'filter' operations, and not 'aggregation' operations, so in that case as_index is ignored (see the note at the end of this section: http://pandas.pydata.org/pandas-docs/stable/groupby.html#filtration). So, it is not a bug, but it also shouldn't be in the docs, as this can only cause confusion that it has no effect.

[jorisvandenbossche]

@kidphys so can you remove the as_index=False? And can you then also squash the commits? Apart from that, ready to merge!

[shoyer]

Thanks @kidphys