Find upper/lower thresholds from ROC curves
trinaryROCRoots.RdFit a smoothed ROC curve, find bounds for threshold and report partial AUC statistics
Usage
trinaryROCRoots(
ins,
max.sens = 0.95,
smoothMethod = "binormal",
sdMultiplier = 2,
maxTPQuantile = 0.3,
...
)Arguments
- ins
A data frame where the first column contains the the observed binary outcomes (e.g., 0 for absence and 1 for presence) and the second column contains predictor values (e.g., predicted probabilities/suitability). This input is used to calculate the ROC curve, derive thresholds, and assess model performance.
- max.sens
value of sensitivity to use if no upper limit is found based on derivatives. default is 0.95
- smoothMethod
default is 'binormal'
- sdMultiplier
A numeric value that controls the adjustment of the lower threshold for presence predictions. It multiplies the standard deviation of the predicted values at presence locations to define this lower threshold. If this threshold results in a negative value, it is adjusted to zero by default. Default is `2`.
- maxTPQuantile
A numeric value representing the quantile of the predicted values at presence locations used to determine the upper threshold for presence predictions. For example, a value of `0.3` means the function uses the 30th percentile of the predicted presence values as the upper threshold, ensuring that the model maintains at least 70 sensitivity. Default is `0.3`.
- ...
options to be passed to `pROC::smooth`