integration.compute_multi_dim_pod

compute_multi_dim_pod(
    poi_grid,
    nuisance_ranges,
    model,
    X_train,
    residuals,
    bandwidth,
    dist_info,
    threshold,
    n_mc_samples=3000,
)

Performs Monte Carlo integration over continuous nuisance parameters to calculate the marginal Probability of Detection (PoD) for a grid of Parameters of Interest (PoI).

Based on the methodology from Malkiel et al. (2026), this function evaluates the flexible multi-dimensional surrogate model across randomly sampled nuisance parameter realizations, calculates the probability of detection at each realization using local variance standardisation, and aggregates them.

Parameters

Name	Type	Description	Default
poi_grid	np.ndarray	The evaluation grid of the Parameters of Interest shape (N, n_pois).	required
nuisance_ranges	Dict[str, Tuple[float, float]]	The min/max bounds for each nuisance.	required
model	Any	Fitted multi-dimensional surrogate model (predicts mean response).	required
X_train	np.ndarray	Original training data (N_train, n_total_vars).	required
residuals	np.ndarray	Residuals from the model fit.	required
bandwidth	float	Local kernel smoothing bandwidth.	required
dist_info	Tuple[str, Tuple]	Residual error distribution (name, params).	required
threshold	float	Signal detection threshold.	required
n_mc_samples	int	Number of Monte Carlo draws per PoI grid point.	3000

Returns

Name	Type	Description
	Tuple[np.ndarray, np.ndarray]	Tuple[np.ndarray, np.ndarray]: - pod_integrated: Integrated PoD values across the poi_grid. - mean_integrated: Expected mean response across the poi_grid.