cascade.model.CASCADE.explain

CASCADE.explain(adata, ctfact, batch_size=128, n_devices=1, design=None)

Explain counterfactual outcome with individual components

Parameters:

• adata (AnnData) – Factual dataset

• ctfact (AnnData) – Counterfactual prediction from counterfactual()

• batch_size (int) – Batch size

• n_devices (int) – Number of GPU devices to use

• design (IntervDesign | None) – Optional intervention design module from design()

Return type:

AnnData

Returns:

Dataset with the following explanation components –

• layers["X_nil"]: Baseline expression without any effect

• layers["X_ctrb_i"]: Contribution from intervention

• layers["X_ctrb_s"]: Contribution from covariates

• layers["X_ctrb_z"]: Contribution from latent

• layers["X_ctrb_ptr"]: Contribution from parents

• layers["X_tot"]: Total counterfactual prediction

All having shape (n_obs, n_vars, n_particles)