Skip to contents

Helpers for computing ROC data from the subset dataset

compute_roc.Rd

Helpers for computing ROC data from the subset dataset

Usage

assert_compute_roc_data(r, with_ci)

compute_roc_data(predictor, response, do_bootstrap, ci_points)

Arguments

r

[list(data.frame())]

dataframe resulting from compute_roc_data

with_ci

[logical(1)]

Indicates if CI is included in the result

predictor

[numeric(n)]

The scores of the predictor

response

[factor(n)]

The response value

do_bootstrap

[logical(1)] Calculate confidence intervals for sensitivity and specificity

ci_points

[list(spec = numeric(), thr = numeric())] Points at which 95% confidence intervals for sensitivity and specificity will be calculated. Depending on the entry CI will be calculated at defined specificity points or threshold points.

Value

[list(data.frame())]

A list with entries:

roc_curve

[data.frame()]

With columns:

  • specificity [numeric()]: Sensitivity

  • sensitivity [numeric()]: Specificity

  • threshold [numeric()]: Threshold

  • auc [numeric(3)]: A numeric vector of length 3 c(LOWER AUC CI, AUC, UPPER AUC CI)

  • direction [character(1)]: The direction of the comparisons < or > according to levels

  • levels [character(2)]: The sorted levels of the response variable according to direction

roc_ci

[data.frame()]

With columns:

  • ci_specificity [numeric()]: Specificity value

  • ci_lower_specificity [numeric()]: Specificity lower confidence interval

  • ci_upper_specificity [numeric()]: Specificity upper confidence interval

  • ci_sensitivity [numeric()]: Sensitivity value

  • ci_lower_sensitivity [numeric()]: Sensitivity lower confidence interval

  • ci_upper_sensitivity [numeric()]: Sensitivity upper confidence interval

  • threshold [numeric()]: Threshold

roc_optimal_cut

[data.frame()]

With columns:

  • optimal_cut_title [character()]: Name of the optimal cut

  • optimal_cut_specificity [numeric()]: Sensitivity at the optimal cut point

  • optimal_cut_lower_specificity [numeric()]: Lower Confidence interval of sensitivity

  • optimal_cut_upper_specificity [numeric()]: Upper Confidence interval of sensitivity

  • optimal_cut_sensitivity [numeric()]: Specificity at the optimal cut point

  • optimal_cut_lower_sensitivity [numeric()]: Lower Confidence interval of sensitivity

  • optimal_cut_upper_sensitivity [numeric()]: Upper Confidence interval of sensitivity

  • optimal_cut_threshold [numeric()]: Threshold of optimal cut

Details

  • Computing CIs for sensitivity and specifity usually implies using bootstrap which can be too expensive, therefore the option of not running calculating them when the function is invoked is included.

  • Response levels are selected alphabetically being case the first one alphabetically the comparison direction is selected automatically by pROC::roc() and related functions.

  • CIs are expected to be 95% CIs

Functions

  • assert_compute_roc_data(): Assert compute_roc result data.frame types are correct

  • compute_roc_data(): Compute ROC analysis