SurvMetrics: An R package for Predictive Evaluation Metrics in Survival Analysis

Recently, survival models have found vast applications in biostatistics, bioinformatics, reliability engineering, finance and related fields. But there are few R packages focusing on evaluating the predictive power of survival models. This lack of handy software on evaluating survival predictions hinders further applications of survival analysis for practitioners. In this research, we want to fill this gap by providing an "all-in-one" R package which implements most predictive evaluation metrics in survival analysis. In the proposed SurvMetrics R package, we implement concordance index for both untied and tied survival data; we give a new calculation process of Brier score and integrated Brier score; we also extend the applicability of integrated absolute error and integrated square error for real data. For models that can output survival time predictions, a simplified metric called mean absolute error is also implemented. In addition, we test the effectiveness of all these metrics on simulated and real survival data sets. The newly developed SurvMetrics R package is available on CRAN at https://CRAN.R-project.org/package=SurvMetrics and GitHub at https://github.com/skyee1/SurvMetrics .