roc¶
ROC (Receiver Operating Characteristic) analysis for single-interval classification.
These tools provide the ability to quickly run receiver operating characteristic analyses on the output of machine-learning models applied to imaging data.
Classes:
| Name | Description |
|---|---|
Roc | Compute receiver operating characteristic curves for single-interval or forced-choice classification. |
Classes¶
Roc¶
Roc(*, input_values = None, binary_outcome = None, method = 'optimal_overall', forced_choice = None)Compute receiver operating characteristic curves for single-interval or forced-choice classification.
The Roc class is based on Tor Wager’s Matlab roc_plot.m function and allows a user to easily run different types of receiver operator characteristic curves. For example, one might be interested in single interval or forced choice.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
input_values | 1-D array/vector of continuous decision values (one per observation) | None | |
binary_outcome | vector of training labels | None | |
method | threshold-selection variant, one of 'optimal_overall', 'optimal_balanced', 'minimum_sdt_bias' | ‘optimal_overall’ | |
forced_choice | index indicating position for each unique subject (default=None) | None |
Attributes:
| Name | Type | Description |
|---|---|---|
binary_outcome | ||
forced_choice | ||
input_values | ||
method |
Methods:
| Name | Description |
|---|---|
calculate | Calculate ROC metrics for single-interval classification. |
plot | Create a ROC plot. |
summary | Display a formatted summary of ROC analysis. |
Methods¶
calculate¶
calculate(*, input_values = None, binary_outcome = None, criterion_values = None, method = 'optimal_overall', forced_choice = None, balanced_acc = False, tail = 2)Calculate ROC metrics for single-interval classification.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
input_values | 1-D array/vector of continuous decision values (one per observation) | None | |
binary_outcome | vector of training labels | None | |
criterion_values | (optional) criterion values for calculating fpr & tpr | None | |
method | threshold-selection variant, one of 'optimal_overall', 'optimal_balanced', 'minimum_sdt_bias' | ‘optimal_overall’ | |
forced_choice | index indicating position for each unique subject (default=None) | None | |
balanced_acc | balanced accuracy for single-interval classification (bool). THIS IS NOT COMPLETELY IMPLEMENTED BECAUSE IT AFFECTS ACCURACY ESTIMATES, BUT NOT P-VALUES OR THRESHOLD AT WHICH TO EVALUATE SENS/SPEC | False | |
tail | 2 | ‘two’ (two-tailed, default) or 1 |
plot¶
plot(*, method = 'gaussian', balanced_acc = False)Create a ROC plot.
Create a specific kind of ROC curve plot, based on input values along a continuous distribution and a binary outcome variable (logical)
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
method | type of plot, one of 'gaussian', 'observed' | ‘gaussian’ | |
balanced_acc | balanced accuracy for single-interval classification | False |
Returns:
| Type | Description |
|---|---|
| fig |
summary¶
summary()Display a formatted summary of ROC analysis.