Run main MCMC — run_mcmc • silverblaze

Run the main geographic profiling MCMC. Model parameters are taken from the current active parameter set, and MCMC parameters are passed in as arguments. All output is stored within the project.

run_mcmc(
  project,
  K = 3,
  burnin = 100,
  samples = 1000,
  auto_converge = TRUE,
  converge_test = 100,
  coupling_on = FALSE,
  cluster = NULL,
  pb_markdown = FALSE,
  store_raw = TRUE,
  create_maps = TRUE,
  silent = !is.null(cluster),
  beta_manual = NULL,
  rung_store = NULL
)

Arguments

project	an rgeoprofile_project, as produced by the function `rgeoprofile_project()`.
K	the number of sources.
burnin	the number of burn-in iterations.
samples	the number of sampling iterations.
auto_converge	whether convergence should be assessed automatically every `converge_test` iterations, leading to termination of the burn-in phase. If `FALSE` then the full `burnin` iterations are used.
converge_test	test for convergence every `convergence_test` iterations if `auto_converge` is being used.
coupling_on	whether to implement Metropolis coupling.
cluster	option to pass in a cluster environment (see package "parallel").
pb_markdown	whether to run progress bars in markdown mode, in which case they are updated once at the end to avoid large amounts of output.
store_raw	whether to store raw MCMC output in addition to summary output. Setting to FALSE can considerably reduce output size in memory.
create_maps	whether to create maps of posterior probability and geoprofile. Usually will want to create these maps, but the code runs much faster without this step, hence the option.
silent	whether to suppress all console output.
beta_manual	allows manual specification of thermodynamic powers used.
rung_store	Pick a rung whose output will be stored

Details

Both longitude and latitude values can be represented to a given precision level using the arguments precision_lon and precision_lat - for example, a precision of 0.01 means that values are rounded to the nearest hundredth of a degree. This allows the use of look-up tables for the likelihood calculation, which significantly speeds up the MCMC. Set to 0 to use exact values (up to C++ "double" precision) rather than using look-up tables.