Extract Log Likelihood from Model Fit

Offline calculations of log likelihood(s) from a bmoe model fit.

Usage

extract_log_lik(object, new_data)

extract_pointwise_log_lik(object, new_data)

extract_componentwise_pointwise_log_lik(object, new_data)

Arguments

object

Object that inherits from bmoe_fit.

new_data

data frame. Similar structure to original data argument.

Setting new_data = NULL signals for the observed data to be used.

Value

• extract_log_lik()

  • Scalar log likelihood for each MCMC iteration.

  • dim(.) = c(n_iters, n_chains, 1).

• extract_pointwise_log_lik()

  • Pointwise likelihood is sum of each component contribution, weighted by the allocation probabilities.

  • dim(.) = c(n_iters, n_chains, n_s_new, n_y, n_k).

• extract_componentwise_pointwise_log_lik()

  • Evaluated at each component (expert) and each observation in y.

  • dim(.) = c(n_iters, n_chains, n_s_new, n_y, n_k).

See also

The loo package for using these quantities in model comparison.