catenets.experiment_utils.torch_metrics module

abs_error_ATE(po: torch.Tensor, hat_te: torch.Tensor) → torch.Tensor

Average Treatment Effect. ATE measures what is the expected causal effect of the treatment across all individuals in the population. :param po: expected outcome. :param hat_te: estimated outcome.

sqrt_PEHE(po: torch.Tensor, hat_te: torch.Tensor) → torch.Tensor

Precision in Estimation of Heterogeneous Effect(PyTorch version). PEHE reflects the ability to capture individual variation in treatment effects. :param po: expected outcome. :param hat_te: estimated outcome.