Haghish, E. F. (2022). adjROC: Computing Sensitivity at a Fix Value of Specificity and Vice Versa [Computer software]. https://CRAN.R-project.org/package=adjROC.
adjROC is an R package for computing adjusted sensitivity and specificity at particular thresholds. There are
methods for estimating the best balance betwen sensitivity and specificity. However, in clinical settings,
there might be an interest in calculating the sensitivity based on a particular fixed value of specificity or
in contrast, calculating specificity for a particular value of sensitivity which is of interest.
For a screening test, specificity of 0.95 might be too high and lower values of specificity may also be acceptable.
In another settings, researchers might wish to know the mount of specificity, while keeping sensitivity high. And finally,
in some situations, a roughly equal value might be desired. Depending on the application, adjROC package allows
users to calculate:
- Specificity at a particular fix value of specificity
- Specificity at the specified value of sensitivity
- The crossing point (meeting point) between sensitivity and specificity curves
and on top of these, it can also visualize the curves and the selected cutoff threshold.
adjROC provides a procedure to compare the performance of two models based on a specific performance metric and using a bootstrap test, evaluates whether the difference in performance of the two models is due to chance or not, and provides a confidence interval accordingly. This allows researchers to not only report AUC, AUCPR, or other performance metrics, but also examine the significant differences between the models.