variogram_score_multivariate.Rd

% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/metrics-multivariate-sample.R
\name{variogram_score_multivariate}
\alias{variogram_score_multivariate}
\title{Variogram score for multivariate forecasts}
\usage{
variogram_score_multivariate(
  observed,
  predicted,
  mv_group_id,
  w = NULL,
  w_vs = NULL,
  p = 0.5
)
}
\arguments{
\item{observed}{A vector with observed values of size n}

\item{predicted}{nxN matrix of predictive samples, n (number of rows) being
the number of data points and N (number of columns) the number of Monte
Carlo samples. Alternatively, if n = 1, \code{predicted} can just be a vector
of size n.}

\item{mv_group_id}{Numeric vector of length n with ids indicating the
grouping of predicted values. Conceptually, each row of the \code{predicted}
matrix \emph{could} be seen as a separate (univariate) forecast.
The grouping id then groups several of those forecasts together, treating
them as a single multivariate forecast.}

\item{w}{Optional numeric vector of weights for forecast samples
(length equal to the number of columns of \code{predicted}).
If \code{NULL} (the default), equal weights are used.}

\item{w_vs}{Optional non-negative weight matrix for the
pairwise comparisons between dimensions. Entry \code{w_vs[i, j]}
controls the importance of the pair (i, j) in the score.
Must be a symmetric square matrix with dimensions equal to
the number of targets within each multivariate group.
If \code{NULL} (the default), all pairs are weighted equally.}

\item{p}{Numeric, order of the variogram score.
Typical choices are 0.5 (default, more robust) and 1.}
}
\value{
A named numeric vector of scores, one per multivariate
group. Lower values are better.
}
\description{
Compute the variogram score for multivariate forecasts.
The variogram score (Scheuerer and Hamill, 2015) evaluates the
dependence structure of multivariate forecasts by comparing
predicted pairwise differences against observed pairwise
differences.

The score is computed using
\code{\link[scoringRules:scores_sample_multiv]{scoringRules::vs_sample()}}.
}
\references{
Scheuerer, M. and Hamill, T.M. (2015). Variogram-Based
Proper Scoring Rules for Probabilistic Forecasts of
Multivariate Quantities. \emph{Monthly Weather Review}, 143,
1321-1334.
}
\keyword{metric}