`causica.datasets.interventional_data`¶

Module Contents¶

`InterventionData`	Dataclass to hold the data associated with an intervention
`CounterfactualData`	Dataclass to hold the data associated with a counterfactual

class causica.datasets.interventional_data.InterventionData[source]¶

Dataclass to hold the data associated with an intervention

This represents one intervention and many samples from the intervened distribution

The class also stores sampled_nodes, i.e. the ones that are neither intervened or conditioned on

Parameters:¶

intervention_data: A TensorDict with all the nodes (including intervened/conditioned)
intervention_values: A dictionary of node names to 1D numpy arrays of the intervened values
condition_values: A dictionary of node names to 1D numpy arrays of the conditioned values

class causica.datasets.interventional_data.CounterfactualData[source]¶

Dataclass to hold the data associated with a counterfactual

This represents one intervention and reference and many samples from the intervened and reference distributions

The class also stores sampled_nodes, i.e. the ones that are neither intervened or conditioned on

Parameters:¶

counterfactual_data: A TensorDict with all of the node values (including intervened) of counterfactual data
factual_data: A TensorDict with all of the node values of the base observations used for the counterfactuals data. This refers to the observations in “What would have happened (CFs) if I would have done (intervention) given I observed (base observation).
intervention_values: A dictionary of node names to 1D numpy arrays of the intervened values