es.crps: CRPS for multivariate-normal predictive density

In ClairBee/SX.weather: Functions to replicate & expand Sichun Xie MSc report on multi-model ensembkes in weather forecasting

Description

CRPS (energy score) for a single fitted multivariate-normal predictive density or ensemble forecast.

Usage

es.crps(o, mu, sig, efc, k = 1000)

Arguments

o	Vector of observations
mu	Vector of means of fitted model
sig	Covariance matrix or variance of fitted model - NOT the standard deviation
efc	Matrix of multivariate forecasts, each column an ensemble member.
k	Number of samples to use in Monte Carlo approximation. Default is 1000.

Details

EITHER parameter 'efc' (for an ensemble forecast) OR parameters 'mu', 'sig' and 'k' (for a fitted univariate or multivariate normal density) must be supplied

Value

Energy score (multivariate CRPS)

ClairBee/SX.weather documentation built on May 6, 2019, 11:51 a.m.