IndicatorCrossValidation

IndicatorCrossValidation(
    actual: ndarray,
    estimate: ndarray,
    variance: ndarray,
    thresholds: list[float],
    cdf: ndarray,
    pit: ndarray,
)

Cross-validation of multiple indicator kriging: estimate is the E-type mean, variance the conditional variance, cdf the (samples, thresholds) corrected probabilities and pit F*(actual), the probability of the sample's own distribution not exceeding its value; NaN where a sample had too few neighbors.

slope property

slope

Slope of the regression of actual on estimated values; below 1 means conditional bias.

standardized_squared_error property

standardized_squared_error

Mean of error² / variance; near 1 when the kriging variance is calibrated.

brier property

brier

Mean squared difference between each threshold's probability and indicator; 0 is perfect.

goodness property

goodness

Goodness statistic, 1 - ∫ (3a(p) - 2)(accuracy(p) - p) dp with a(p) = 1 where accurate; 1 is perfect, and an interval too narrow costs twice one too wide.

accuracy

accuracy(p)

Fraction of samples inside their symmetric p-probability interval; p or above when accurate.

Parameters:

Name	Type	Description	Default
p	float or array_like	Probabilities in [0, 1].	required