API: kdeplot fails with NaNs.

[TomAugspurger]

This is inconsistent with the other plotting methods:

In [65]: df = pd.DataFrame(np.random.uniform(size=(100, 4)))

In [66]: df.loc[0, 0] = np.nan

In [67]: df.plot(kind='kde')
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-67-9a78a7983298> in <module>()
----> 1 df.plot(kind='kde')

/Users/tom/Envs/py3/lib/python3.4/site-packages/pandas/pandas/tools/plotting.py in plot_frame(frame, x, y, subplots, sharex, sharey, use_index, figsize, grid, legend, rot, ax, style, title, xlim, ylim, logx, logy, xticks, yticks, kind, sort_columns, fontsize, secondary_y, layout, **kwds)
   2362                              secondary_y=secondary_y, layout=layout, **kwds)
   2363 
-> 2364     plot_obj.generate()
   2365     plot_obj.draw()
   2366     return plot_obj.result

/Users/tom/Envs/py3/lib/python3.4/site-packages/pandas/pandas/tools/plotting.py in generate(self)
    913         self._compute_plot_data()
    914         self._setup_subplots()
--> 915         self._make_plot()
    916         self._add_table()
    917         self._make_legend()

/Users/tom/Envs/py3/lib/python3.4/site-packages/pandas/pandas/tools/plotting.py in _make_plot(self)
   1915                 kwds['style'] = style
   1916 
-> 1917             artists = plotf(ax, y, column_num=i, **kwds)
   1918             self._add_legend_handle(artists[0], label)
   1919 

/Users/tom/Envs/py3/lib/python3.4/site-packages/pandas/pandas/tools/plotting.py in plotf(ax, y, style, column_num, **kwds)
   1960         def plotf(ax, y, style=None, column_num=None, **kwds):
   1961             if LooseVersion(spv) >= '0.11.0':
-> 1962                 gkde = gaussian_kde(y, bw_method=self.bw_method)
   1963             else:
   1964                 gkde = gaussian_kde(y)

/Users/tom/Envs/py3/lib/python3.4/site-packages/scipy/stats/kde.py in __init__(self, dataset, bw_method)
    186 
    187         self.d, self.n = self.dataset.shape
--> 188         self.set_bandwidth(bw_method=bw_method)
    189 
    190     def evaluate(self, points):

/Users/tom/Envs/py3/lib/python3.4/site-packages/scipy/stats/kde.py in set_bandwidth(self, bw_method)
    496             raise ValueError(msg)
    497 
--> 498         self._compute_covariance()
    499 
    500     def _compute_covariance(self):

/Users/tom/Envs/py3/lib/python3.4/site-packages/scipy/stats/kde.py in _compute_covariance(self)
    507             self._data_covariance = atleast_2d(np.cov(self.dataset, rowvar=1,
    508                                                bias=False))
--> 509             self._data_inv_cov = linalg.inv(self._data_covariance)
    510 
    511         self.covariance = self._data_covariance * self.factor**2

/Users/tom/Envs/py3/lib/python3.4/site-packages/scipy/linalg/basic.py in inv(a, overwrite_a, check_finite)
    352 
    353     if check_finite:
--> 354         a1 = np.asarray_chkfinite(a)
    355     else:
    356         a1 = np.asarray(a)

/Users/tom/Envs/py3/lib/python3.4/site-packages/numpy/lib/function_base.py in asarray_chkfinite(a, dtype, order)
    593     if a.dtype.char in typecodes['AllFloat'] and not np.isfinite(a).all():
    594         raise ValueError(
--> 595                 "array must not contain infs or NaNs")
    596     return a
    597 

ValueError: array must not contain infs or NaNs

The default should be to drop missing observations.

[onesandzeroes]

Missing values are ignored by all the DataFrame.plot() kinds, and there's no arg/switch to not ignore them, right? So we can just drop missing vals without checking any args. Seems pretty easy if so, I'll get a PR going.

[TomAugspurger]

Closed by #8196