Run FreqEstimationModel MCMC for one group

Usage

run_FreqEstimationModel(
  sample_matrix_list,
  COI,
  threads,
  seed,
  n_chains = 3L,
  no_traces_preburnin = 10000L,
  thinning_interval = 1L,
  moi_max = 8L
)

Arguments

sample_matrix_list: Output of create_FEM_input().
COI: Average complexity of infection.
threads: Number of threads.
seed: Random seed.
n_chains: Number of MCMC chains to run. At least two are needed to compute the Gelman-Rubin R-hat convergence diagnostic.
no_traces_preburnin: Number of MCMC traces retained per chain before burn-in is discarded. This is the dominant driver of memory use: the sampler pre-allocates a no_traces_preburnin x n_samples x n_haplotypes x n_chains array of doubles, so a 6-locus group (64 haplotypes) at the default reaches several GB.
thinning_interval: Number of Metropolis-Hastings updates performed per retained trace. Total sampler updates are no_traces_preburnin * thinning_interval, so halving the traces and doubling this runs the chain exactly as far while storing fewer draws. Note that reported ESS is bounded by the number of retained draws.

Value

A list with the population frequency table, MCMC runtime, marker names, alternate alleles, group size, mono-allelic loci, and a convergence_diag data frame.