cases: Stratified Evaluation of Subgroup Classification Accuracy

cases is an R package to simultaneously assess classification accuracy of multiple classifiers in several subgroups (strata). For instance, it allows to asses the accuracy of multiple candidate (index) diagnostic tests which is often measured with

• sensitivity (accuracy in the diseased subgroup) and
• specificity (accuracy in the healthy subgroup).

A widespread goal in diagnostic accuracy studies a so-called co-primary analysis of these two endpoints, i.e. to show a significant benefit (compared to some benchmark) in sensitivity and specificity for at least one of the candidate classifiers. The package implements different methods for multiplicity adjustment for that purpose (e.g. Bonferroni, maxT, pairs bootstrap).

Installation

You can install the development version of cases from GitHub with:

# install.packages("remotes")
remotes::install_github('maxwestphal/cases', build_vignettes = TRUE)

Usage

A vignette which explains the basic functionality of the cases package can be displayed as follows:

vignette(topic="package_overview", package = "cases")

The following vignette shows an exemplary usage of the package in the context of biomarker assessment and prediction model evaluation:

vignette(topic="example_wdbc", package = "cases")

References

• Westphal, Max, and Antonia Zapf. “Statistical Inference for Diagnostic Test Accuracy Studies with Multiple Comparisons.” arXiv preprint arXiv:2105.13469 (2021).