Computes pointwise log predictive density contributions from a matrix of
log predictive densities/probabilities for posterior draws. When PSIS
log-weights are supplied, they are used to form a weighted log-sum-exp per
observation.
Usage
ptw_log_pred_density(ylp, psis_log_weights = NULL)
Arguments
- ylp
A numeric matrix of log predictive densities/probabilities with
dimensions draws x observations.
- psis_log_weights
Optional numeric matrix of normalized PSIS log
weights with the same dimensions as ylp. Each column must sum to 1 on
the probability scale.
Value
A numeric vector of length ncol(ylp) with pointwise log
predictive density values.
Examples
ylp <- matrix(log(c(0.2, 0.4, 0.3, 0.8)), nrow = 2)
ptw_log_pred_density(ylp)
#> [1] -1.203973 -0.597837
lw <- matrix(log(c(0.7, 0.3, 0.6, 0.4)), nrow = 2)
ptw_log_pred_density(ylp, lw)
#> [1] -1.3470736 -0.6931472
