The function logs_nbmix computes the logarithmic score of forecasts
based on mixtures of negative binomial (or Poisson) distributions via
dnbmix(). This is different from the kernel density estimation
approach available via scores_sample().
logs_nbmix(observed, means, size)a vector of observed counts during the simulation period.
a n.ahead x n.sim matrix of means.
the dispersion parameter of the dnbinom() distribution
or NULL (Poisson forecasts). Can also be time-varying (of length
n.ahead).
a vector of log-scores for the observed counts.
scores_sample() for an alternative approach of calculating
the logarithmic score from simulation-based forecasts