Pandas ewma does not work on Panel

I would expect to be able to perform an ewma over each minor axis within each major axis:

In [52]:

wp = pd.Panel(np.random.randn(2, 4, 5), items=['Item1', 'Item2'],
               minor_axis=pd.date_range('1/1/2000', periods=5),
               major_axis=['A', 'B', 'C', 'D'])
In [57]:

wp
Out[57]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 4 (major_axis) x 5 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: A to D
Minor_axis axis: 2000-01-01 00:00:00 to 2000-01-05 00:00:00
In [56]:

pd.ewma(wp, com=2)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-56-bb436def494f> in <module>()
----> 1 pd.ewma(wp, com=2)

/usr/local/lib/python2.7/dist-packages/pandas-0.13.0_285_gfcfaa7d-py2.7-linux-x86_64.egg/pandas/stats/moments.pyc in ewma(arg, com, span, halflife, min_periods, freq, time_rule, adjust)
    373         return result
    374 
--> 375     return_hook, values = _process_data_structure(arg)
    376     output = np.apply_along_axis(_ewma, 0, values)
    377     return return_hook(output)

/usr/local/lib/python2.7/dist-packages/pandas-0.13.0_285_gfcfaa7d-py2.7-linux-x86_64.egg/pandas/stats/moments.pyc in _process_data_structure(arg, kill_inf)
    329         values = arg
    330 
--> 331     if not issubclass(values.dtype.type, float):
    332         values = values.astype(float)
    333 

/usr/local/lib/python2.7/dist-packages/pandas-0.13.0_285_gfcfaa7d-py2.7-linux-x86_64.egg/pandas/core/generic.pyc in __getattr__(self, name)
   1802                 return self[name]
   1803             raise AttributeError("'%s' object has no attribute '%s'" %
-> 1804                                  (type(self).__name__, name))
   1805 
   1806     def __setattr__(self, name, value):

AttributeError: 'Panel' object has no attribute 'dtype'

Comment From: jreback

not implemented

Comment From: cancan101

Agreed. Does this seem like a reasonable feature? That being said, how would one go about performing such an ewma on a DataFrame or Series?

Comment From: jreback

you can use Panel.apply (newly enhanced in 0.13.1) to apply to the Frame (or Series), see here: http://pandas.pydata.org/pandas-docs/dev/whatsnew.html#enhancements.

I personally just use a dict iteration and build new objects with a function (I will switch to apply at some point though, cleaner syntax).

Then usually concat the results, that way you can control whether you build same / higher / or lower dimensional objects.

Comment From: jreback

closing as Panels are deprecated