crps

BMCI.crps(y_obs, x_true, x2_max=-1.0)[source]

Compute the Continuous Ranked Probability Score.

This function approximates the cumulative probability density of the posterior distribution for the observations in y_obs and computes the continuously ranked probability score (CRPS):

\[CRPS(\mathbf{y}, x) = \int_{-\infty}^\infty (F_{x | \mathbf{y}}(x') - \mathrm{1}_{x < x'})^2 \: dx'\]
Parameters:
  • y_obs (numpy.ndarray) – Array of shape (n, m) containing the n observations for which to perform the retrieval.

  • x_true (numpy.ndarray) – 1-D array containing the n x values to test the predictions against.