Skip to contents

Negated CRPS for normal predictive distributions

Source: R/scores.R
crps_normal.Rd

Computes the Continuous Ranked Probability Score for a normal predictive distribution using scoringRules::crps_norm() and negates it so that higher values are better.

Usage

crps_normal(mu, sigma, x)

Arguments

mu

Numeric vector. Location parameters (conditional means).

sigma

Numeric vector. Scale parameters (conditional SDs, > 0).

x

Numeric vector. Realised observations.

Value

Numeric vector of CRPS values in (-Inf, 0] (negated loss).

Details

Calls scoringRules::crps_norm(y = x, mean = mu, sd = sigma) and negates. Use for GARCH(1,1)-norm forecasts where mu is the conditional mean and sigma is the conditional standard deviation.

Examples

if (requireNamespace("scoringRules", quietly = TRUE)) {
  crps_normal(mu = c(0, 1), sigma = c(1, 2), x = c(0.2, 1.3))
}
#> [1] -0.2495997 -0.4853088