scipp.groupby#

scipp.groupby(data, /, group, *, bins=None)#

Group dataset or data array based on values of specified labels.

This implements the “split-apply-combine” approach for data reduction. The returned object can be used to apply reduction operations (mean, sum, etc.) to each group.

Parameters:

data (DataArray | Dataset) – Input data to reduce.
group (Variable | str) – Name of labels to use for grouping or Variable to use for grouping
bins (Variable | None, default: None) – Optional bins for grouping label values.

Returns:

GroupByDataArray | GroupByDataset – GroupBy helper object.

See also

scipp.group: Group into bins based on unique values (creates binned data).
scipp.bin: Create binned data by binning along one or more coordinates.

Examples

Group by a string label coordinate and compute the mean:

>>> import scipp as sc
>>> data = sc.DataArray(
...     sc.array(dims=['x'], values=[1.0, 2.0, 3.0, 4.0, 5.0, 6.0], unit='K'),
...     coords={
...         'x': sc.arange('x', 6),
...         'label': sc.array(dims=['x'], values=['a', 'b', 'a', 'b', 'a', 'b'])
...     }
... )
>>> grouped = sc.groupby(data, 'label')
>>> grouped.mean('x')
<scipp.DataArray>
Dimensions: Sizes[label:2, ]
Coordinates:
* label                      string        <no unit>  (label)  ["a", "b"]
Data:
                            float64              [K]  (label)  [3, 4]

>>> grouped.sum('x')
<scipp.DataArray>
Dimensions: Sizes[label:2, ]
Coordinates:
* label                      string        <no unit>  (label)  ["a", "b"]
Data:
                            float64              [K]  (label)  [9, 12]

Group continuous values into bins and compute the sum:

>>> data = sc.DataArray(
...     sc.array(dims=['x'], values=[1.0, 2.0, 3.0, 4.0, 5.0], unit='K'),
...     coords={'x': sc.array(dims=['x'], values=[0.1, 0.5, 1.2, 1.8, 2.5], unit='m')}
... )
>>> bins = sc.linspace('x', 0.0, 3.0, num=4, unit='m')
>>> sc.groupby(data, 'x', bins=bins).sum('x')
<scipp.DataArray>
Dimensions: Sizes[x:3, ]
Coordinates:
* x                         float64              [m]  (x [bin-edge])  [0, 1, 2, 3]
Data:
                            float64              [K]  (x)  [3, 7, 5]

Coordinates not used for grouping are dropped if they depend on the reduced dimension:

>>> data = sc.DataArray(
...     sc.array(dims=['x'], values=[1.0, 2.0, 3.0, 4.0], unit='K'),
...     coords={
...         'x': sc.arange('x', 4),
...         'label': sc.array(dims=['x'], values=['a', 'b', 'a', 'b']),
...         'aux': sc.array(dims=['x'], values=[10, 20, 30, 40])
...     }
... )
>>> result = sc.groupby(data, 'label').mean('x')
>>> 'aux' in result.coords  # 'aux' depends on 'x' which is reduced
False
>>> 'label' in result.coords  # 'label' becomes a dimension coord
True

This Page

scipp.groupby#