causallib.evaluation.plots.data_extractors module

Plot data extractors.

The responsibility of these classes is to extract the data from the EvaluationResults objects to match the requested plot.

class causallib.evaluation.plots.data_extractors.BaseEvaluationPlotDataExtractor(evaluation_results: causallib.evaluation.results.EvaluationResults)[source]

Extractor to get plot data from EvaluationResults.

Subclasses also have a plot_names property.

cv_by_phase(phase='train')[source]

Get the cross-validation indices of all folds for a given phase.

Parameters: phase (str, optional) – Requested phase: “train” or “valid. Defaults to “train”.
Returns: _description_
Return type: List

abstract get_data_for_plot(plot_name, phase='train')[source]: Get data for plot with name plot_name.

class causallib.evaluation.plots.data_extractors.BinaryOutcomePlotDataExtractor(evaluation_results: causallib.evaluation.results.EvaluationResults)[source]

Extractor to get plot data from OutcomeEvaluatorPredictions.

Note that the available plots are different if the outcome predictions are binary/classification or continuous/regression.

get_data_for_plot(plot_name, phase='train')[source]

Retrieve the data needed for each provided plot. Plot interfaces are at the plots module.

class causallib.evaluation.plots.data_extractors.ContinuousOutcomePlotDataExtractor(evaluation_results: causallib.evaluation.results.EvaluationResults)[source]

Extractor to get plot data from OutcomeEvaluatorPredictions.

Note that the available plots are different if the outcome predictions are binary/classification or continuous/regression.

get_data_for_plot(plot_name, phase='train')[source]

Retrieve the data needed for each provided plot. Plot interfaces are at the plots module.

plot_names = frozenset({'common_support', 'continuous_accuracy', 'residuals'})

class causallib.evaluation.plots.data_extractors.PropensityPlotDataExtractor(evaluation_results: causallib.evaluation.results.EvaluationResults)[source]

Extractor to get plot data from PropensityEvaluatorPredictions.

get_data_for_plot(plot_name, phase='train')[source]

Retrieve the data needed for each provided plot. Plot interfaces are at the plots.py module.

Parameters

plot_name (str) – Plot name.
fold_predictions (list[PropensityEvaluatorPredictions]) – Predictions for each fold.
list[np.ndarray] (cv) – Indices (in iloc positions) of each fold.

Returns

Plot data

Return type

tuple

plot_names = frozenset({'calibration', 'covariate_balance_love', 'covariate_balance_slope', 'pr_curve', 'roc_curve', 'weight_distribution'})

class causallib.evaluation.plots.data_extractors.WeightPlotDataExtractor(evaluation_results: causallib.evaluation.results.EvaluationResults)[source]

Extractor to get plot data from WeightEvaluatorPredictions.

get_data_for_plot(plot_name, phase='train')[source]

Retrieve the data needed for each provided plot.

Plot functions are in plots module.

plot_names = frozenset({'covariate_balance_love', 'covariate_balance_slope', 'weight_distribution'})