pandas.DataFrame.interpolate

DataFrame.interpolate(method='linear', axis=0, limit=None, inplace=False, downcast='infer', **kwargs)

Interpolate values according to different methods.

Parameters :

method : {‘linear’, ‘time’, ‘values’, ‘index’ ‘nearest’, ‘zero’,

‘slinear’, ‘quadratic’, ‘cubic’, ‘barycentric’, ‘krogh’, ‘polynomial’, ‘spline’ ‘piecewise_polynomial’, ‘pchip’}

  • ‘linear’: ignore the index and treat the values as equally

    spaced. default

  • ‘time’: interpolation works on daily and higher resolution

    data to interpolate given length of interval

  • ‘index’: use the actual numerical values of the index

  • ‘nearest’, ‘zero’, ‘slinear’, ‘quadratic’, ‘cubic’, ‘barycentric’, ‘polynomial’ is passed to scipy.interpolate.interp1d with the order given both ‘polynomial’ and ‘spline’ requre that you also specify and order (int) e.g. df.interpolate(method=’polynomial’, order=4)

  • ‘krogh’, ‘piecewise_polynomial’, ‘spline’, and ‘pchip’ are all

    wrappers around the scipy interpolation methods of similar names. See the scipy documentation for more on their behavior: http://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html

axis : {0, 1}, default 0

  • 0: fill column-by-column
  • 1: fill row-by-row

limit : int, default None.

Maximum number of consecutive NaNs to fill.

inplace : bool, default False

Update the NDFrame in place if possible.

downcast : optional, ‘infer’ or None, defaults to ‘infer’

Downcast dtypes if possible.

Returns :

Series or DataFrame of same shape interpolated at the NaNs

See also

reindex, replace, fillna

Examples

# Filling in NaNs: >>> s = pd.Series([0, 1, np.nan, 3]) >>> s.interpolate() 0 0 1 1 2 2 3 3 dtype: float64