benchmark_univariate_GMM_estimation: Launch the benchmark to compare statistical performances...
In bastienchassagnol/RGMMBench: A benchmark on Gaussian mixtures in R

benchmark_univariate_GMM_estimation

R Documentation

Launch the benchmark to compare statistical performances between packages

Description

Launch the benchmark to compare statistical performances between packages

Usage

benchmark_univariate_GMM_estimation(
  mixture_functions,
  sigma_values,
  mean_values,
  proportions,
  cores = getOption("mc.cores", parallel::detectCores()),
  id_scenario = NULL,
  prop_outliers = 0,
  nobservations = c(2000),
  Nbootstrap = 100,
  epsilon = 10^-4,
  itmax = 500,
  nstart = 10L,
  short_iter = 200,
  short_eps = 10^-2,
  prior_prob = 0.05,
  initialisation_algorithms = c("kmeans", "quantiles", "random", "hc", "rebmix")
)

benchmark_multivariate_GMM_estimation(
  mixture_functions,
  mean_values,
  proportions,
  sigma_values,
  id_scenario = NULL,
  cores = getOption("mc.cores", parallel::detectCores()),
  nobservations = c(2000),
  Nbootstrap = 100,
  epsilon = 10^-4,
  itmax = 500,
  nstart = 10L,
  short_iter = 200,
  short_eps = 10^-2,
  prior_prob = 0.05,
  initialisation_algorithms = c("kmeans", "random", "hc", "rebmix")
)

Arguments

`mixture_functions`	List of the packages to be compared (Id:name of the package, value: its options)
`sigma_values, mean_values, proportions`	the true parameters to be retrieved
`cores`	the number of cores to be used, by default all the available cores
`id_scenario`	Possibility to set it to another number than one, to uniquely identify them
`prop_outliers`	the proportion of outliers added in the simulation
`nobservations`	the number of observations drawn to generate the random sample
`Nbootstrap`	the number of bootstrap simulations and repetitions to perform
`epsilon, itmax`	respectively criterion threshold and maximal number of iterations to reach it
`nstart, short_iter, short_eps`	hyper-parameters to control the initialisation step
`prior_prob`	add minimal uncertainty on the cluster assignment returned by hierarchical clustering method
`initialisation_algorithms`	among 6 methods, which algorithms to be chosen for the initialisation phase

Value

a list with the simulated distributions of the estimates, some summary scores per parameter and aggregated measures as well as boxplot and Heatmap correlation representations of the estimates