Wilkinson's GABC log-likelihood — gabc_log_likelihood • rmorie

Monte Carlo estimate of the GABC likelihood of Wilkinson (2014) eq. (1) and the bootstrap variance that is used as the GP's nugget.

Usage

gabc_log_likelihood(
  sim,
  obs,
  theta,
  n_sim = 50L,
  epsilon = 1,
  summary = NULL,
  kernel = "gaussian",
  seed = 0L,
  bootstrap = 25L
)

Arguments

sim: Simulator, signature (theta, rng).
obs: Observed data.
theta: Parameter vector.
n_sim: Number of Monte Carlo samples.
epsilon: Acceptance tolerance.
summary: Optional summary function.
kernel: Acceptance kernel, "gaussian" or "uniform".
seed: Random seed.
bootstrap: Number of bootstrap replications for the nugget.

Value

List log_lik, nugget_variance.

Examples

sim <- function(theta, e) theta + c(-0.5, 0, 0.5)
gabc_log_likelihood(sim, obs = c(0.5, 1, 1.5), theta = 1, n_sim = 5L, epsilon = 0.5)$log_lik
#> [1] 0