Skip to contents

Fit an ABC rejection-sampling posterior for one subject

Source: R/abc.R
mbmm_abc_fit.Rd

Draws kinetic parameter vectors from the prior, simulates the mechanistic trajectory, evaluates a censoring-aware discrepancy, and retains the lowest-discrepancy candidates.

Usage

mbmm_abc_fit(
  id,
  obs,
  simulate_fn,
  prior_sample_fn,
  n_draws = 10000,
  n_keep = NULL,
  accept_frac = 0.025,
  distance_fn = NULL,
  seed = NULL,
  failure = c("discard", "error")
)

Arguments

id

Subject identifier.

obs

Subject observation data.frame containing time and y, and optionally censored and lod.

simulate_fn

Function with signature simulate_fn(theta_row, times).

prior_sample_fn

Function with signature prior_sample_fn(n) returning a data.frame.

n_draws

Number of prior draws.

n_keep

Optional number of particles to retain.

accept_frac

Optional fraction of valid draws to retain. Used when n_keep = NULL.

distance_fn

Optional custom function with signature distance_fn(theta_row, obs, pred).

seed

Optional random seed.

failure

Either "discard" or "error". With "discard", failed simulations receive infinite discrepancy.

Value

A data.frame of accepted particles with distance and id columns. ABC diagnostic information is stored as attributes.

Details

The number retained can be specified directly using n_keep, or by an acceptance fraction using accept_frac. The manuscript analyses generally used a 2.5 percent acceptance fraction before donor balancing.