vgislw

a matrix or vector of log weights. For for computing LOO <code>lw =
-log_lik</code> (see <code><a href="/link/extract_log_lik?package=loo&version=0.1.0" rd-options="" data-mini-rdoc="loo::extract_log_lik">extract_log_lik</a></code>) and is an $S$ by $N$
matrix where $S$ is the number of simulations and $N$ is the number
of 

the proportion of importance weights to use for the generalized
Pareto fit. The <code>100*wcp</code>% largest weights are used as the sample
from which to estimate the parameters of the generalized Pareto
distribution.

when used for the generalized pareto fit, the largest
<code>100*wcp</code>% of the importance weights are modified according to
<code>pmax(x, max(x) - thresh)</code> for numerical stability.

thresh

maximum allowed value for the generalized Pareto shape parameter
$k$.

kmax

for truncating very large weights to $S$^<code>wtrunc</code>. Set
to zero for no truncation.

wtrunc

the number of cores to use for parallelization.

cores

We efficiently approximate leave-one-out cross-validation (LOO)
  using very good importance sampling (VGIS), a new procedure for regularizing
  importance weights. As a byproduct of our calculations, we also obtain
  approximate standard errors for estimated predictive errors, and for the
  comparison of predictive errors between two models. We also compute the
  widely applicable information criterion (WAIC).

vgislw: VGIS: Very good importance sampling

Description

Usage

Arguments

Value

Details

See Also