Methods for voting for a consensus model based on the metrics of fit_distributions

Usage

find_consensus_model(
  models,
  method = c("weighted_majority_vote", "rra"),
  metric = unique(sapply(models, `[[`, "metric")),
  distribution_prioritization = c("norm", "unif", "gamma", "gamma_flip"),
  weights = rev(sqrt(seq(1, 6))[1:length(metric)])
)

Arguments

models: a list of models, e.g. derived from fit_distributions
method: one of weighted_majority_vote and rra as a method of determining the best method. rra requires the package RobustRankAggreg
metric: metrics used to fit models. Metrics should be ordered in descending priority. The first metric in the vector will be used to return the consensus model for the distribution determined through voting.
distribution_prioritization: if method is weighted_majority_voting, a list of ranked distributions, to break ties
weights: required if method is weighted_majority_voting. weights of each metric to be multiplied by rankings. Weights should be in decreasing order. A higher weight results in a higher priority of the metric.

Value

a list of the best model for each metric and a consensus model representing the model with the consensus distribution and the lowest weighted metric. Each model is a list of the following data:

par: A character string denoting the region_id of the Histogram
dist: The distribution name
metric: The metric used to fit the distribution
value: The fitted value of the metric function
dens: A function that returns the density of the fitted distribution
seg_start: start index of the interval
seg_end: end index of the interval

Examples

if (FALSE) {
 data <- observations_to_histogram(rnorm(10000, mean = 20, sd = 10))
 data <- data$histogram_data
 models <- fit_distributions(data)
 find_consensus_models(models)

}