create_inference_conditions: Create the inference conditions

In richelbilderbeek/pirouette: Create a Bayesian Posterior from a Phylogeny

Create the inference conditions

Description

Create the parameters to determine how to choose a model for the inference

Usage

create_inference_conditions(
  model_type = "generative",
  run_if = "always",
  do_measure_evidence = FALSE,
  os = rappdirs::app_dir()$os
)

Arguments

model_type	type of inference model supplied for an experiment. Possible values: generative: the inference model is (or is assumed to be) the inference model underlying the phylogeny candidate: the inference model is a candidate model, that competes with other models for having the most evidence (aka highest marginal likelihood)
run_if	the condition for an experiment's inference model to be run. Possible values: always: always best_candidate: if the inference model is the candidate model with the most evidence (aka highest marginal likelihood)
do_measure_evidence	boolean to indicate if the evidence (aka marginal likelihood) of an experiment must be measured
os	name of the operating system, can be mac, unix or win. Use check_os if the operating system is valid.

Value

the inference conditions

Author(s)

Giovanni Laudanno, Richèl J.C. Bilderbeek

Examples

if (beautier::is_on_ci()) {

  # Create the inference conditions parameter set
  if (rappdirs::app_dir()$os != "win") {
    # it does not work on Windows

    # Model type can be 'generative' or 'candidate'
    model_type <- "candidate"
    # Run condition can be 'always' or 'best_candidate'
    run_if <- "best_candidate"
    # Evidence (aka marginal likelihood) can be measured yes or no
    do_measure_evidence <- TRUE
  } else {
    # Model type can be 'generative' or 'candidate'
    model_type <- "generative"
    # Run condition can be 'always' or 'best_candidate'
    run_if <- "always"
    # Evidence (aka marginal likelihood) can be measured yes or no
    do_measure_evidence <- FALSE
  }

  inference_conditions <- create_inference_conditions(
    model_type = model_type,
    run_if = run_if,
    do_measure_evidence = do_measure_evidence
  )

  # Using the inference conditions, create a testing candidate experiment
  experiment <- create_test_cand_experiment(
    inference_conditions = inference_conditions
  )

  evidence_filename <- NA
  if (do_measure_evidence) evidence_filename <- get_temp_evidence_filename()

  # Use the experiment to create the full pirouette parameter set
  pir_params <- create_pir_params(
    alignment_params = create_test_alignment_params(),
    experiments = list(experiment),
    evidence_filename = evidence_filename
  )

  # Run that experiment on a continuous integration service,
  # only when BEAST2 is installed
  if (beautier::is_on_ci() &&
    beastier::is_beast2_installed() &&
    is_beast2_ns_pkg_installed()
  ) {
    pir_out <- pir_run(
      phylogeny = ape::read.tree(text = "((A:2, B:2):1, C:3);"),
      pir_params = pir_params
    )
    pir_plot(pir_out)
  }
}

richelbilderbeek/pirouette documentation built on Oct. 18, 2023, 3:09 p.m.