causalkit.refutation.refute_placebo_outcome#
- causalkit.refutation.refute_placebo_outcome(inference_fn, data, random_state=None, **inference_kwargs)[source]#
Generate random outcome (target) variables while keeping treatment and covariates intact. For binary outcomes, generates random binary variables with the same proportion. For continuous outcomes, generates random variables from a normal distribution fitted to the original data. A valid causal design should now yield θ ≈ 0 and a large p-value.