grouby( ... , as_index=False) convert "float64" to "object" type.

[ronenabr]

In [1]: import pandas as pd 

In [2]: df = pd.DataFrame({"c1" : [1,2,6,6,8]})

In [3]: df["c2"] = df.c1/2.0

In [4]: df.groupby("c2").mean().reset_index().c2
Out[4]: 
0    0.5
1    1.0
2    3.0
3    4.0
Name: c2, dtype: float64

In [5]: df.groupby("c2", as_index=False).mean().c2
Out[5]: 
0    0.5
1      1
2      3
3      4
Name: c2, dtype: object

Note the difference in dtype between the two cases.
I think the 2nd case shouldn't be different then the 1st case.

[ronenabr]

(On pandas 0.11.0)

[jreback]

this is a bug, thanks

as a work around, you can do:

In [11]: df.groupby("c2", as_index=False).mean().convert_objects().c2
Out[11]: 
0    0.5
1    1.0
2    3.0
3    4.0
Name: c2, dtype: float64

[jreback]

thanks for the report! closed by #3613