Skip to contents

Restriction for multiple dependent and independent variables

rROC.data.frame.Rd

Restriction for multiple dependent and independent variables. Traverses all dependent variables and within all independent variables. Then calculates the rROC in the sense of dependent ~ independent.

Can save intermediate results to disk, to avoid re-calculating for crash-reasons or to save time when re-running the same analysis.

Can return plot_density_rROC_empirical for every combination.

Usage

# S3 method for class 'data.frame'
rROC(
  x,
  independent_vars,
  dependent_vars = NULL,
  y = NULL,
  save_path = NULL,
  parallel_permutations = TRUE,
  n_permutations = 10000,
  save_intermediate = TRUE,
  load_existing_intermediate = TRUE,
  positive_label = 1,
  verbose = TRUE,
  do_plots = FALSE,
  fix_seed = 0,
  ...
)

Arguments

x

A data.frame containing all dependent and independent variables as columns.

independent_vars

A character vector of independent variable column names. If NULL, all columns except dependent_vars are used.

dependent_vars

A character vector of dependent variable column names. If NULL, y must be given.

y

Either a vector of dependent variable values or a list of length 1 of a vector of dependent variable values. If NULL, dependent_vars must be given.

save_path

Path to save the results to. Intermediate results are saved into the directory file.path(save_path, "_partial_directory").

parallel_permutations

boolean: If TRUE, the permutation will be done via future.apply::future_lapply, otherwise by base::lapply

n_permutations

How many permutations should be done

save_intermediate

Should intermediate results be saved to disk? If TRUE, every combination by itself is saved into file.path(save_path, "_partial_directory").

load_existing_intermediate

Should the earlier saved intermediate results in the folder file.path(save_path, "_partial_directory") be loaded?

positive_label

Label for the positive class. All other values of response are regarded as negative cases.

verbose

Should progress be printed?

do_plots

Should the plot_density_rROC_empirical be calculated and returned?

fix_seed

boolean: If not FALSE, the seed for each permutation will be set by set.seed(fix_seed + permutation_i)

...

Arguments passed on to simple_rROC_permutation

return_proc

  1. Should pROC::roc() be returned for the full dataset? 2) Should pROC::roc() be returned on each of the part datasets? Only works with get_all_aucs_fun=get_all_aucs after get_all_aucs_norecalculation() does not calculate the ROC curves for each restriction separately.

Value

A list of lists of simple_rROC_permutation and plot results. It is structured as follows: