pandas@8bb2cc1 (original) (raw)

`@@ -6081,89 +6081,191 @@ def replace(self, to_replace=None, value=None, inplace=False, limit=None,

6081

6082

`_shared_docs['interpolate'] = """

6083

``` Please note that only method='linear' is supported for


`6084`

``

`-

DataFrames/Series with a MultiIndex.

`

``

`6084`

`+

DataFrame/Series with a MultiIndex.

`

`6085`

`6085`

``

`6086`

`6086`

` Parameters

`

`6087`

`6087`

` ----------

`

`6088`

``

`-

method : {'linear', 'time', 'index', 'values', 'nearest', 'zero',

`

`6089`

``

`-

'slinear', 'quadratic', 'cubic', 'barycentric', 'krogh',

`

`6090`

``

`-

'polynomial', 'spline', 'piecewise_polynomial',

`

`6091`

``

`-

'from_derivatives', 'pchip', 'akima'}

`

``

`6088`

`+

method : str, default 'linear'

`

``

`6089`

`+

Interpolation technique to use. One of:

`

`6092`

`6090`

``

`6093`

``

`-

* 'linear': ignore the index and treat the values as equally

`

``

`6091`

`+

* 'linear': Ignore the index and treat the values as equally

`

`6094`

`6092`

` spaced. This is the only method supported on MultiIndexes.

`

`6095`

``

`-

default

`

`6096`

``

`-

* 'time': interpolation works on daily and higher resolution

`

`6097`

``

`-

data to interpolate given length of interval

`

`6098`

``

`-

* 'index', 'values': use the actual numerical values of the index

`

`6099`

``

`-

* 'nearest', 'zero', 'slinear', 'quadratic', 'cubic',

`

`6100`

``

`-

'barycentric', 'polynomial' is passed to

`

`6101`

``

`` -

:class:`scipy.interpolate.interp1d`. Both 'polynomial' and

``

`6102`

``

`` -

'spline' require that you also specify an `order` (int),

``

``

`6093`

`+

* 'time': Works on daily and higher resolution data to interpolate

`

``

`6094`

`+

given length of interval.

`

``

`6095`

`+

* 'index', 'values': use the actual numerical values of the index.

`

``

`6096`

`+

* 'pad': Fill in NaNs using existing values.

`

``

`6097`

`+

* 'nearest', 'zero', 'slinear', 'quadratic', 'cubic', 'spline',

`

``

`6098`

`+

'barycentric', 'polynomial': Passed to

`

``

`6099`

`` +

`scipy.interpolate.interp1d`. Both 'polynomial' and 'spline'

``

``

`6100`

`` +

require that you also specify an `order` (int),

``

`6103`

`6101`

```  e.g. ``df.interpolate(method='polynomial', order=4)``.

6104

These use the actual numerical values of the index.

6105

'krogh', 'piecewise_polynomial', 'spline', 'pchip' and 'akima'

6106

are all wrappers around the scipy interpolation methods of

6107

similar names. These use the actual numerical values of the

6108

index. For more information on their behavior, see the

6109

`` -

`scipy documentation

6110

`` -

http://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation`__

6111

`` -

and `tutorial documentation

6112

`` -

http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html`__

6113

'from_derivatives' refers to

6114

`` -

:meth:scipy.interpolate.BPoly.from_derivatives which

6102

These use the numerical values of the index.

6103

'krogh', 'piecewise_polynomial', 'spline', 'pchip', 'akima':

6104

Wrappers around the SciPy interpolation methods of similar

6105

`` +

names. See Notes.

6106

'from_derivatives': Refers to

6107

`` +

scipy.interpolate.BPoly.from_derivatives which

6115

6108

` replaces 'piecewise_polynomial' interpolation method in

6116

scipy 0.18

6109

scipy 0.18.

6117

6110

6118

6111

` .. versionadded:: 0.18.1

6119

6112

6120

6113

` Added support for the 'akima' method.

6121

6114

` Added interpolate method 'from_derivatives' which replaces

6122

'piecewise_polynomial' in scipy 0.18; backwards-compatible with

6123

scipy < 0.18

6124

-

6125

axis : {0, 1}, default 0

6126

0: fill column-by-column

6127

1: fill row-by-row

6128

limit : int, default None.

6129

Maximum number of consecutive NaNs to fill. Must be greater than 0.

6130

limit_direction : {'forward', 'backward', 'both'}, default 'forward'

6131

limit_area : {'inside', 'outside'}, default None

6132

None: (default) no fill restriction

6133

'inside' Only fill NaNs surrounded by valid values (interpolate).

6134

'outside' Only fill NaNs outside valid values (extrapolate).

6115

'piecewise_polynomial' in SciPy 0.18; backwards-compatible with

6116

SciPy < 0.18

6135

6117

6118

axis : {0 or 'index', 1 or 'columns', None}, default None

6119

Axis to interpolate along.

6120

limit : int, optional

6121

Maximum number of consecutive NaNs to fill. Must be greater than

6122

6123

inplace : bool, default False

6124

Update the data in place if possible.

6125

limit_direction : {'forward', 'backward', 'both'}, default 'forward'

6136

6126

` If limit is specified, consecutive NaNs will be filled in this

6137

6127

` direction.

6128

`` +

limit_area : {None, 'inside', 'outside'}, default None

6129

If limit is specified, consecutive NaNs will be filled with this

6130

restriction.

6131

+

6132


 * ``None``: No fill restriction.

6133

'inside': Only fill NaNs surrounded by valid values

6134

(interpolate).

6135

'outside': Only fill NaNs outside valid values (extrapolate).

6138

6136

6139

6137

` .. versionadded:: 0.21.0

6140

inplace : bool, default False

6141

Update the NDFrame in place if possible.

6138

+

6142

6139

` downcast : optional, 'infer' or None, defaults to None

6143

6140

` Downcast dtypes if possible.

6144

kwargs : keyword arguments to pass on to the interpolating function.

6141

**kwargs

6142

Keyword arguments to pass on to the interpolating function.

6145

6143

6146

6144

` Returns

6147

6145

` -------

6148

Series or DataFrame of same shape interpolated at the NaNs

6146

Series or DataFrame

6147

Returns the same object type as the caller, interpolated at

6148


 some or all ``NaN`` values

6149

6150

` See Also

6151

` --------

6152

reindex, replace, fillna

6152

fillna : Fill missing values using different methods.

6153

scipy.interpolate.Akima1DInterpolator : Piecewise cubic polynomials

6154

(Akima interpolator).

6155

scipy.interpolate.BPoly.from_derivatives : Piecewise polynomial in the

6156

Bernstein basis.

6157

scipy.interpolate.interp1d : Interpolate a 1-D function.

6158

scipy.interpolate.KroghInterpolator : Interpolate polynomial (Krogh

6159

interpolator).

6160

scipy.interpolate.PchipInterpolator : PCHIP 1-d monotonic cubic

6161

interpolation.

6162

scipy.interpolate.CubicSpline : Cubic spline data interpolator.

6163

+

6164

Notes

6165

6166

The 'krogh', 'piecewise_polynomial', 'spline', 'pchip' and 'akima'

6167

methods are wrappers around the respective SciPy implementations of

6168

similar names. These use the actual numerical values of the index.

6169

For more information on their behavior, see the

6170

`` +

`SciPy documentation

6171

`` +

http://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation`__

6172

`` +

and `SciPy tutorial

6173

`` +

http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html`__.

6153

6174

6154

6175

` Examples

6155

6176

` --------

6156

-

6157

Filling in NaNs

6177


 Filling in ``NaN`` in a :class:`~pandas.Series` via linear

6178

interpolation.

6158

6179

6159

6180

` >>> s = pd.Series([0, 1, np.nan, 3])

6181

s

6182

0 0.0

6183

1 1.0

6184

2 NaN

6185

3 3.0

6186

dtype: float64

6160

6187

` >>> s.interpolate()

6161

0 0

6162

1 1

6163

2 2

6164

3 3

6188

0 0.0

6189

1 1.0

6190

2 2.0

6191

3 3.0

6165

6192

` dtype: float64

6166

6193

6194


 Filling in ``NaN`` in a Series by padding, but filling at most two

6195


 consecutive ``NaN`` at a time.

6196

+

6197

s = pd.Series([np.nan, "single_one", np.nan,

6198

... "fill_two_more", np.nan, np.nan, np.nan,

6199

... 4.71, np.nan])

6200

s

6201

0 NaN

6202

1 single_one

6203

2 NaN

6204

3 fill_two_more

6205

4 NaN

6206

5 NaN

6207

6 NaN

6208

7 4.71

6209

8 NaN

6210

dtype: object

6211

s.interpolate(method='pad', limit=2)

6212

0 NaN

6213

1 single_one

6214

2 single_one

6215

3 fill_two_more

6216

4 fill_two_more

6217

5 fill_two_more

6218

6 NaN

6219

7 4.71

6220

8 4.71

6221

dtype: object

6222

+

6223


 Filling in ``NaN`` in a Series via polynomial interpolation or splines:

6224

Both 'polynomial' and 'spline' methods require that you also specify

6225


 an ``order`` (int).

6226

+

6227

s = pd.Series([0, 2, np.nan, 8])

6228

s.interpolate(method='polynomial', order=2)

6229

0 0.000000

6230

1 2.000000

6231

2 4.666667

6232

3 8.000000

6233

dtype: float64

6234

+

6235

Fill the DataFrame forward (that is, going down) along each column

6236

using linear interpolation.

6237

+

6238

Note how the last entry in column 'a' is interpolated differently,

6239

because there is no entry after it to use for interpolation.

6240


 Note how the first entry in column 'b' remains ``NaN``, because there

6241

is no entry befofe it to use for interpolation.

6242

+

6243

df = pd.DataFrame([(0.0, np.nan, -1.0, 1.0),

6244

... (np.nan, 2.0, np.nan, np.nan),

6245

... (2.0, 3.0, np.nan, 9.0),

6246

... (np.nan, 4.0, -4.0, 16.0)],

6247

... columns=list('abcd'))

6248

df

6249

a b c d

6250

0 0.0 NaN -1.0 1.0

6251

1 NaN 2.0 NaN NaN

6252

2 2.0 3.0 NaN 9.0

6253

3 NaN 4.0 -4.0 16.0

6254

df.interpolate(method='linear', limit_direction='forward', axis=0)

6255

a b c d

6256

0 0.0 NaN -1.0 1.0

6257

1 1.0 2.0 -2.0 5.0

6258

2 2.0 3.0 -3.0 9.0

6259

3 2.0 4.0 -4.0 16.0

6260

+

6261

Using polynomial interpolation.

6262

+

6263

df['d'].interpolate(method='polynomial', order=2)

6264

0 1.0

6265

1 4.0

6266

2 9.0

6267

3 16.0

6268

Name: d, dtype: float64

6167

6269

` """

6168

6270

6169

6271

`@Appender(_shared_docs['interpolate'] % _shared_doc_kwargs)