The unnormalized log density behind an MCMC posterior's log_prob() method

Neither surrogate gives the evidence \(p(x)\), so normalize = TRUE has nothing to normalize by and says so rather than returning a number that looks like a density.

Usage

mcmc_log_prob(post, theta, x, warn, what)

Arguments

post

An MCMC-sampled nsbi_posterior.

theta

Parameter values to evaluate.

x

Observation to condition on, or NULL for the posterior's own.

warn

Warn that normalize is being ignored (the caller decides, since only it can see whether the argument was actually supplied).

what

The method name to use in that warning, "NLE" or "NRE".