Search code examples
pythonnumpypandasmedian

Median of a list with NaN values removed, in python

Is it possible to calculate the median of a list without explicitly removing the NaN's, but rather, ignoring them?

I want median([1,2,3,NaN,NaN,NaN,NaN,NaN,NaN]) to be 2, not NaN.

Solution

  • numpy 1.9.0 has the function nanmedian:

    nanmedian(a, axis=None, out=None, overwrite_input=False, keepdims=False)
    Compute the median along the specified axis, while ignoring NaNs.

    Returns the median of the array elements.

    .. versionadded:: 1.9.0

    E.g.

    >>> from numpy import nanmedian, NaN
>>> nanmedian([1,2,3,NaN,NaN,NaN,NaN,NaN,NaN])
2.0

    If you can't use version 1.9.0 of numpy, something like @Parker's answer will work; e.g.

    >>> import numpy as np
>>> x = np.array([1,2,3,NaN,NaN,NaN,NaN,NaN,NaN])
>>> np.median(x[~np.isnan(x)])
2.0

    or

    >>> np.median(x[np.isfinite(x)])
2.0

    (When applied to a boolean array, ~ is the unary operator notation for not.)