Compute Performance Measures

This function provides a modular and extensible way to compute performance measures (PM) for Data-Generating Mechanisms (DGMs). It handles different types of measures and automatically determines the required arguments for each measure function.

Usage

compute_single_measure(
  dgm_name,
  measure_name,
  method,
  method_setting,
  conditions,
  measure_fun,
  measure_mcse_fun,
  power_test_type = "p_value",
  estimate_col = "estimate",
  true_effect_col = "mean_effect",
  ci_lower_col = "ci_lower",
  ci_upper_col = "ci_upper",
  p_value_col = "p_value",
  bf_col = "BF",
  convergence_col = "convergence",
  power_threshold_p_value = 0.05,
  power_threshold_bayes_factor = 10,
  method_replacements = NULL,
  n_repetitions = 1000,
  overwrite = FALSE,
  path = NULL,
  ...
)

Arguments

dgm_name

Character string specifying the DGM name

measure_name

Name of the measure to compute (e.g., "bias", "mse")

method

Character vector of method names

method_setting

Character vector of method settings, must be same length as method

conditions

Data frame of conditions from dgm_conditions()

measure_fun

Function to compute the measure

measure_mcse_fun

Function to compute the MCSE for the measure

power_test_type

Character vector specifying the test type for power computation: "p_value" (default) or "bayes_factor" for each method. If a single value is provided, it is repeated for all methods.

estimate_col

Character string specifying the column name containing parameter estimates. Default is "estimate"

true_effect_col

Character string specifying the column name in conditions data frame containing true effect sizes. Default is "mean_effect"

ci_lower_col

Character string specifying the column name containing lower confidence interval bounds. Default is "ci_lower"

ci_upper_col

Character string specifying the column name containing upper confidence interval bounds. Default is "ci_upper"

p_value_col

Character string specifying the column name containing p-values. Default is "p_value"

bf_col

Character string specifying the column name containing Bayes factors. Default is "BF"

convergence_col

Character string specifying the column name containing convergence indicators. Default is "convergence"

power_threshold_p_value

Numeric threshold for power computation with p-values. Default is 0.05 (reject H0 if p < 0.05).

power_threshold_bayes_factor

Numeric threshold for power computation with Bayes factors. Default is 10 (reject H0 if BF > 10)

method_replacements

Named list of replacement method specifications. Each element should be named with the "method-method_setting" combination (e.g., "RMA-default") and contain a named list with:

• method: Character vector of replacement method names

• method_setting: Character vector of replacement method settings (same length as methods)

• power_test_type: Optional character vector of power test types for each replacement method (same length as methods). If not specified, uses the main power_test_type parameter

If multiple elements are specified within the vectors, these replacements are applied consecutively in case the previous replacements also failed to converge. Defaults to NULL, i.e., omitting repetitions without converged results on method-by-method basis.

n_repetitions

Number of repetitions in each condition. Necessary method replacement. Defaults to 1000.

overwrite

Logical indicating whether to overwrite existing results. If FALSE (default), will skip computation for method-measure combinations that already exist

path

Character string specifying the directory path where the datasets/results/measures should be saved. Defaults to the location specified via PublicationBiasBenchmark.get_option("simulation_directory"). The objects are stored in dgm_name/datasets, dgm_name/results, dgm_name/measures subfolders.

...

Additional arguments passed to measure functions

Value

TRUE upon successfully computation of the results file