statistics Package

statistics Package

compute_CI Module

WORC.statistics.compute_CI.compute_confidence(metric, N_train, N_test, alpha=0.95)[source]

Function to calculate the adjusted confidence interval for cross-validation. metric: numpy array containing the result for a metric for the different cross validations (e.g. If 20 cross-validations are performed it is a list of length 20 with the calculated accuracy for each cross validation) N_train: Integer, number of training samples N_test: Integer, number of test_samples alpha: float ranging from 0 to 1 to calculate the alpha*100% CI, default 0.95

WORC.statistics.compute_CI.compute_confidence_bootstrap(bootstrap_metric, test_metric, N_1, alpha=0.95)[source]

Function to calculate confidence interval for bootstrapped samples. metric: numpy array containing the result for a metric for the different bootstrap iterations test_metric: the value of the metric evaluated on the true, full test set alpha: float ranging from 0 to 1 to calculate the alpha*100% CI, default 0.95

WORC.statistics.compute_CI.compute_confidence_logit(metric, N_train, N_test, alpha=0.95)[source]

Function to calculate the adjusted confidence interval metric: numpy array containing the result for a metric for the different cross validations (e.g. If 20 cross-validations are performed it is a list of length 20 with the calculated accuracy for each cross validation) N_train: Integer, number of training samples N_test: Integer, number of test_samples alpha: float ranging from 0 to 1 to calculate the alpha*100% CI, default 95%

delong Module

Adopted from https://github.com/yandexdataschool/roc_comparison.

WORC.statistics.delong.calc_pvalue(aucs, sigma)[source]

Computes log(10) of p-values.

Args:

aucs: 1D array of AUCs sigma: AUC DeLong covariances

Returns:

log10(pvalue)

WORC.statistics.delong.compute_ground_truth_statistics(ground_truth, sample_weight=None)[source]
WORC.statistics.delong.compute_midrank(x)[source]

Computes midranks.

Args:

x - a 1D numpy array

Returns:

array of midranks

WORC.statistics.delong.compute_midrank_weight(x, sample_weight)[source]

Computes midranks.

Args:

x - a 1D numpy array

Returns:

array of midranks

WORC.statistics.delong.delong_roc_test(ground_truth, predictions_one, predictions_two)[source]

Computes log(p-value) for hypothesis that two ROC AUCs are different.

Args:

ground_truth: np.array of 0 and 1 predictions_one: predictions of the first model,

np.array of floats of the probability of being class 1

predictions_two: predictions of the second model,

np.array of floats of the probability of being class 1

WORC.statistics.delong.delong_roc_variance(ground_truth, predictions)[source]

Computes ROC AUC variance for a single set of predictions.

Args:

ground_truth: np.array of 0 and 1 predictions: np.array of floats of the probability of being class 1

WORC.statistics.delong.fastDeLong(predictions_sorted_transposed, label_1_count)[source]

Fast DeLong test computation.

The fast version of DeLong’s method for computing the covariance of unadjusted AUC. Args:

predictions_sorted_transposed: a 2D numpy.array[n_classifiers, n_examples]

sorted such as the examples with label “1” are first

Returns:

(AUC value, DeLong covariance)

Reference:
@article{sun2014fast,
title={Fast Implementation of DeLong’s Algorithm for

Comparing the Areas Under Correlated Receiver Oerating Characteristic Curves},

author={Xu Sun and Weichao Xu}, journal={IEEE Signal Processing Letters}, volume={21}, number={11}, pages={1389–1393}, year={2014}, publisher={IEEE}

}