input
    is not already an `ndarray`, an error will be raised.
keepdims : bool, optional
    If this is set to True, the axes which are reduced are left
    in the result as dimensions with size one. With this option,
    the result will broadcast correctly against the input array.

Returns
-------
median : ndarray
    A new array holding the result is returned unless out is
    specified, in which case a reference to out is returned.
    Return data-type is `float64` for integers and floats smaller than
    `float64`, or the input data-type, otherwise.

See Also
--------
mean

Notes
-----
Given a vector ``V`` with ``N`` non masked values, the median of ``V``
is the middle value of a sorted copy of ``V`` (``Vs``) - i.e.
``Vs[(N-1)/2]``, when ``N`` is odd, or ``{Vs[N/2 - 1] + Vs[N/2]}/2``
when ``N`` is even.

Examples
--------
>>> import numpy as np
>>> x = np.ma.array(np.arange(8), mask=[0]*4 + [1]*4)
>>> np.ma.median(x)
1.5

>>> x = np.ma.array(np.arange(10).reshape(2, 5), mask=[0]*6 + [1]*4)
>>> np.ma.median(x)
2.5
>>> np.ma.median(x, axis=-1, overwrite_input=True)
masked_array(data=[2.0, 5.0],
             mask=[False, False],
       fill_value=1e+20)

r[