Rescale data with emergent constraints#

Rescale label data using a single emergent constraint.


This diagnostic uses an emergent relationship between data marked as var_type=label (Y axis) and var_type=feature (X axis) together with an observation of the X axis (var_type=prediction_input and var_type=prediction_input_error) to calculate factors that are necessary to rescale each input point so that it matches the constraint. The rescaling is applied to data marked as var_type=label_to_rescale. All data needs the attribute tag which needs to be identical for label, prediction_input, prediction_input_error and label_to_rescale. Only a single tag for feature is possible.


Manuel Schlund (DLR, Germany)



Configuration options in recipe#

group_by_attributes: list of str, optional (default: [‘dataset’])

List of attributes used to separate different input points.

ignore: list of dict, optional

Ignore specific datasets by specifying multiple dict s of metadata.

legend_kwargs: dict, optional

Optional keyword arguments of matplotlib.pyplot.legend() (affects only plots with legends).

pattern: str, optional

Pattern matched against ancestor file names.

plot_emergent_relationship: dict, optional

If given, plot emergent relationship between X and Y data. Specify additional keyword arguments by plot_kwargs and plot appearance options by pyplot_kwargs (processed as functions of matplotlib.pyplot). Use {} to plot with default settings.

plot_kwargs_for_groups: dict, optional

Specify additional keyword arguments (values) for the different points defined by group_by_attributes (keys) used in plots.

savefig_kwargs: dict, optional

Keyword arguments for matplotlib.pyplot.savefig().

seaborn_settings: dict, optional

Options for seaborn.set_theme() (affects all plots).